Word to Info Inc v. Google Inc

Filing 162

CLAIM CONSTRUCTION ORDER. Signed by Judge William H. Orrick on 07/12/2016. (jmdS, COURT STAFF) (Filed on 7/12/2016)

1

UNITED STATES DISTRICT COURT

2

NORTHERN DISTRICT OF CALIFORNIA

3
WORD TO INFO INC,

4

Case No. 15-cv-03486-WHO

Plaintiff,
5

CLAIM CONSTRUCTION ORDER
v.

6
GOOGLE INC.,

7

Defendant.
8
INTRODUCTION

9

Plaintiff Word to Info, Inc. (“WTI”) brings two separate lawsuits – one against defendant

10

United States District Court
Northern District of California

11

Facebook Inc. (“Facebook”) (Case No. 15-cv-03485-WHO) and the other against Google Inc.

12

(“Google”) (Case No. 15-cv-03486-WHO) – accusing both defendants of infringing the same

13

seven patents, United States Patent Nos. (1) 5,715,468 (“the ’468 patent”); (2) 6,138,087 (“the

14

’087 patent”); (3) 6,609,091 (“the ’091 patent”); (4) 7,349,840 (“the ’840 patent”); (5) 7,873,509

15

(“the ’509 patent”); (6) 8,326,603 (“the ’603 patent”); and (7) 8,688,436 (“the ’436 patent”). Each

16

of the patents in suit relates to natural language processing. The parties have asked me to construe

17

ten terms from the asserted claims. Based on the parties’ briefing, the tutorial on June 10, 2016,

18

and the arguments presented at the hearing on June 17, 2016, I construe the terms as set forth

19

below.1
BACKGROUND

20
21

The patents-in-suit relate to natural language processing. Each of the patents-in-suit was

22

invented by Robert Budzinski, each shares the same specification, and each besides the ’468 and

23

’436 patents shares the same abstract.
The ’468 patent (issued February 3, 1998) is titled “Memory System for Storing and

24
25
26
27
28

1

Defendants indicated in the parties’ joint claim construction statement that they intended to assert
that the terms “word sense number,” “clause implying word sense number,” and “lexically
processing” are indefinite. Jnt. Claim Constr. Stmt. at 2-3 (Dkt. No. 96); Jnt. Claim Constr. Stmt.
Ex. A at 1, 9, 16 (Dkt. No. 96-1). However, they do not argue indefiniteness in their claim
construction briefing, and I do not address the issue here.

1

Retrieving Experience and Knowledge with Natural Language.”
The ’087 patent (issued October 24, 2000) is titled “Memory System for Storing and

2
3

Retrieving Experience and Knowledge with Natural Language Utilizing State Representation

4

Data, Word Sense Numbers, Function Codes and/or Directed Graphs.”
The ’091 patent (issued August 19, 2003) is titled “Memory System for Storing and

5
6

Retrieving Experience and Knowledge with Natural Language Utilizing State Representation

7

Data, Word Sense Numbers, Function Codes and/or Directed Graphs.”
The ’840 patent (issued March 25, 2008) is titled “Memory System for Storing and

8
9

Retrieving Experience and Knowledge with Natural Language Utilizing State Representation
Data, Word Sense Numbers, Function Codes, Directed Graphs, Context Memory, and/or Purpose

11
United States District Court
Northern District of California

10

Relations.”
The ’509 patent (issued January 18, 2011) is titled “Memory System for Storing and

12
13

Retrieving Experience and Knowledge with Natural Language Utilizing State Representation

14

Data, Word Sense Numbers, Function Codes, Directed Graphs, Context Memory, and/or Purpose

15

Relations.”
The ’603 patent (issued December 4, 2012) is titled “Memory System for Storing and

16
17

Retrieving Experience and Knowledge with Natural Language Queries.”
The ’436 patent (issued April 1, 2014) is titled “Memory System for Storing and

18
19

Retrieving Experience and Knowledge by Utilizing Natural Language Responses.”
Budzinski filed the first of the seven applications underlying the patents-in-suit on

20
21

September 30, 1994. Opening Br. at 2 (Dkt. No. 101); Defs. Br. at 1 (Dkt. No. 103).2 During

22

prosecution of the ’468 patent, the examiner initially rejected certain claims as unpatentable over

23

European Patent Application Publication No. 0180888 to Katayama (“Katayama”) (Mead Decl.

24

Ex. A, Dkt. No. 103-2). See ’468 file at 38 (Webb Decl. Ex. 9, Dkt. No. 101-9). In a response

25

dated June 25, 1996, Budzinski argued that his invention – including the invention’s claimed

26

“word sense numbers” – was distinct from Katayama. See id. at 38 (“Word sense numbers are a

27
2

28

All “Dkt. No.” citations in this Order are to the docket in WTI v. Facebook, No. 15-cv-03485WHO.
2

1

novel, nonobvious invention not taught in Katayama. A word sense number makes it possible to

2

have new capabilities compared to the case dictionary, function memory, and semantic analysis

3

means of Katayama.”); see also id. at 38-65. The examiner subsequently allowed the claims. In

4

all subsequent applications for the patents-in-suit, Budzinski included a sentence in the abstract

5

stating, “A word sense number is an address to the meaning of a word.”

6

During prosecution of the ’509 and ’603 patents, the examiner initially rejected certain
claims as unpatentable over United States Patent No. 7,383,169 to Vanderwende (“Vanderwende”)

8

(Webb Decl. Ex. 13, Dkt. No. 101-13). See ’509 file at 4 (Webb Decl. Ex. 10, Dkt. No. 101-10);

9

’603 file at 13 (Webb Decl. Ex. 11, Dkt. No. 101-11). In responses dated September 11, 2009 (for

10

the ’509 patent) and March 2, 2012 (for the ’603 patent), Budzinski argued that his invention was

11
United States District Court
Northern District of California

7

distinct from Vanderwende, and the examiner subsequently allowed the claims. See ’509 file at 4-

12

76; ’603 file at 13-87.

13
14

LEGAL STANDARD
Claim construction is a matter of law. Markman v. Westview Instruments, Inc., 517 U.S.

15

370, 379 (1996). “The words of a claim are generally given their ordinary and customary meaning

16

as understood by a person of ordinary skill in the art when read in the context of the specification

17

and prosecution history.” Thorner v. Sony Computer Entm’t Am. LLC, 669 F.3d 1362, 1365 (Fed.

18

Cir. 2012). “There are only two exceptions to this general rule: (1) when a patentee sets out a

19

definition and acts as his own lexicographer, or (2) when the patentee disavows the full scope of a

20

claim term either in the specification or during prosecution.” Id.

21

A patentee acts as his or her own lexicographer when he or she “clearly set[s] forth a

22

definition of the disputed claim term other than its plain and ordinary meaning.” Id. (internal

23

quotation marks omitted); accord Akamai Techs., Inc. v. Limelight Networks, Inc., 805 F.3d 1368,

24

1375 (Fed. Cir. 2015). That is, the patentee must “clearly express an intent to redefine the term.”

25

Thorner, 669 F.3d at 1365 (internal quotation marks omitted).

26

The standard for when a patentee has disavowed the full scope of a claim term is “similarly

27

exacting.” Id. at 1366. “Disavowal requires that the specification or prosecution history make

28

clear that the invention does not include a particular feature or is . . . limited to a particular form of
3

1

the invention.” Hill-Rom Servs., Inc. v. Stryker Corp., 755 F.3d 1367, 1372 (Fed. Cir. 2014)

2

(internal quotation marks omitted). With respect to disavowal made during prosecution, it is only

3

a “clear and unmistakable disavowal [that] overcomes the heavy presumption that claim terms

4

carry their full ordinary and customary meaning.” Biogen Idec, Inc. v. GlaxoSmithKline LLC, 713

5

F.3d 1090, 1095 (Fed. Cir. 2013) (internal quotation marks omitted). “Where an applicant’s

6

statements are amenable to multiple reasonable interpretations, they cannot be deemed clear and

7

unmistakable.” 3M Innovative Properties Co. v. Tredegar Corp., 725 F.3d 1315, 1326 (Fed. Cir.

8

2013); accord Avid Tech., Inc. v. Harmonic, Inc., 812 F.3d 1040, 1045 (Fed. Cir. 2016). The

9

Federal Circuit “ha[s] thus consistently rejected [prosecution history disclaimer arguments] based
on prosecution statements too vague or ambiguous to qualify as a disavowal of claim scope.”

11
United States District Court
Northern District of California

10

Avid, 812 F.3d at 1045 (internal quotation marks omitted).

12
13
14
15
16
17
18
19
20
21
22
23
24
25
26

On the other hand, the Federal Circuit has found prosecution history disclaimer in a variety
of circumstances:
For example, we have held that disclaimer applies when the patentee
makes statements such as “the present invention requires” or “the
present invention is” or “all embodiments of the present invention
are.” We have also found disclaimer when the specification
indicated that for “successful manufacture” a particular step was
“required.” Andersen Corp. v. Fiber Composites, LLC, 474 F.3d
1361, 1367 (Fed. Cir. 2007) (“Those statements are not descriptions
of particular embodiments, but are characterizations directed to the
invention as a whole.”). We found disclaimer when the specification
indicated that the invention operated by “pushing (as opposed to
pulling) forces,” and then characterized the “pushing forces” as “an
important feature of the present invention.” We found disclaimer
when the patent repeatedly disparaged an embodiment as
“antiquated,” having “inherent inadequacies,” and then detailed the
“deficiencies that make it difficult” to use. Chicago Bd. Options
Exch., Inc. v. Int’l Sec. Exch., LLC, 677 F.3d 1361, 1372 (Fed. Cir.
2012) (“[T]he specification goes well beyond expressing the
patentee’s preference and its repeated derogatory statements about [a
particular embodiment] reasonably may be viewed as a
disavowal.”). Likewise, we found disclaimer limiting a claim
element to a feature of the preferred embodiment when the
specification described that feature as a “very important feature in an
aspect of the present invention” and disparaged alternatives to that
feature.
Hill-Rom, 755 F.3d at 1372 (some internal citations and alterations omitted).

27
28
4

When supported by clear and unmistakable statements of disavowal, “the scope of a

1
2

surrender of subject matter during prosecution is [not] limited to what is absolutely necessary to

3

avoid a prior art reference.” Norian Corp. v. Stryker Corp., 432 F.3d 1356, 1361 (Fed. Cir. 2005).

4

The Federal Circuit has held “patentees to the scope of what they ultimately claim, and [has] not

5

allowed them to assert that claims should be interpreted as if they had surrendered only what they

6

had to.” Id. at 1361-62.
Significantly for the purposes of the bulk of the claim construction disputes at issue here, a

7
8

claim term that lacks a “plain or established meaning to one of ordinary skill in the art . . .

9

ordinarily cannot be construed broader than the disclosure in the specification.” Indacon, Inc. v.
Facebook, Inc., No. 2015-1129, 2016 WL 3162043, at *3 (Fed. Cir. June 6, 2016); see also Irdeto

11
United States District Court
Northern District of California

10

Access, Inc. v. Echostar Satellite Corp., 383 F.3d 1295, 1300 (Fed. Cir. 2004) (holding that there

12

is no presumption of ordinary and customary meaning “where a disputed term lacks an accepted

13

meaning in the art,” and that “absent such an accepted meaning, we construe a claim term only as

14

broadly as provided for by the patent itself”). In such circumstances, “[t]he duty . . . falls on the

15

patent applicant to provide a precise definition for the disputed term,” Irdeto, 383 F.3d at 1300;

16

see also J.T. Eaton & Co. v. Atl. Paste & Glue Co., 106 F.3d 1563, 1570 (Fed. Cir. 1997), and “a

17

court must resort to the remaining intrinsic evidence – the written description and the prosecution

18

history – to obtain the meaning of th[e] term,” Goldenberg v. Cytogen, Inc., 373 F.3d 1158, 1164

19

(Fed. Cir. 2004); see also Honeywell Int’l Inc. v. Universal Avionics Sys. Corp., 488 F.3d 982, 991

20

(Fed. Cir. 2007) (construing the term “terrain floor boundary,” which had “no ordinary meaning to

21

a skilled artisan,” according to the particular description of the term in the specification); Network

22

Commerce, Inc. v. Microsoft Corp., 422 F.3d 1353, 1359-61 (Fed. Cir. 2005) (construing the term

23

“download component,” which had “no commonly understood meaning reflected in general

24

dictionaries or similar sources” and “[no] specialized meaning in the relevant art,” to include the

25

particular attributes described in the specification).
DISCUSSION

26
27
28

I.

UNDISPUTED TERMS
The parties agree that “plausibility” means “[a] measurement of whether an interpretation
5

1

is consistent in the context of the conversation with respect to grammar, semantics, experience,

2

and knowledge.” Jnt. Claim Constr. Stmt. at 2.

3

II.

DISPUTED TERMS NOT INVOLVING 35 U.S.C. § 112 ¶ 6
A.

4

“word sense number”
WTI’s Construction

5

Claims

6

’087 patent, An identifier for a word
claim 1, 4, meaning, the identifier
17, 23
including an
identification number.
’091 patent,
claim 1, 12

7
8
9
10

United States District Court
Northern District of California

11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

’840 patent,
claim 1, 2,
3, 5, 15, 16
’509 patent,
claim 9, 10,
16
’603 patent,
claim 9, 10,
13, 14
’436 patent,
claim 1, 7,
11, 14
’468 patent,
claim 1, 8,
12, 16, 24,
31, 33

Defendants’
Construction
An address to the
meaning of a word, which
contains attributes
regarding the meaning
and has meaning data that
is (1) utilized to
determine the intended
meaning of a word usage,
and (2) organized into
relations to other word
sense numbers that can be
used for selecting a word
sense number which has
the intended meaning of a
word contained in natural
language.

Court’s Construction
Defendants’ construction,
slightly modified:

An address to the
meaning of a word, which
contains attributes
regarding the meaning
and has meaning data that
is (1) utilized to
determine the intended
meaning of a word usage,
and (2) organized into
relations to other word
sense numbers that can be
used for selecting a word
sense number which has
the intended meaning of a
For the following types of word contained in natural
word sense numbers, the
language.
structure is as follows:
For the following types of
An adjective word sense
word sense numbers, the
number is composed of
structure is as follows:
an identification number,
a state value or value
An adjective word sense
range, and an owner word number is composed of
sense number.
an identification number,
a state value or value
The word sense number
range, and an owner word
of a concrete noun
sense number.
contains a word sense
identifying number, a
The word sense number
type number, a specificity of a concrete noun
number, and an
contains a word sense
experience number.
identifying number, a
type number, a specificity
The word sense number
number, and an
of a state abstract noun
experience number.
contains an identification
number, a type number, a The word sense number
specificity number, and
of a state abstract noun
an experience number.
contains an identification
number, a type number, a
A verb word sense
specificity number, and
number contains an
an experience number.
6

identification number
which defines the verb.
word sense number, and
includes partial to
complete word sense
identification numbers of
main sentence roles.

1
2
3
4
5
6
7

’603 patent,
claim 9

Example Claims
A method of processing natural language in an apparatus, which comprises steps
utilizing a natural language processor to provide a data base of natural language
with associated word sense numbers and/or function codes in memory associated
with said apparatus,

8
9

associating additional data with said word sense numbers and/or function codes
associated with said data base of natural language in memory associated with said
apparatus,

10
11
United States District Court
Northern District of California

A verb word sense
number contains an
identification number
which defines the verb.
word sense number, and
includes partial to
complete word sense
identification numbers of
main sentence roles.

12

indexing said data base of natural language with respect to word sense numbers
and/or function codes with said apparatus,

13

storing said index in memory associated with said apparatus,

14

providing electronically encoded data which is representative of natural language
in memory associated with said apparatus,

15

providing a dictionary data base in memory associated with said apparatus
wherein said dictionary data base contains a plurality of entries which are
comprised of one or more of syntax usage data, associated word sense numbers,
and/or function codes,

16
17
18

lexically processing said electronically encoded data to access said dictionary
data base with said apparatus,

19

providing a grammar specification in memory associated with said apparatus,

20

utilizing said natural language words which are associated with said electronically
encoded data and said associated data which are from said dictionary data base
entries with reference to said grammar specification to select word sense numbers
and/or function codes with said apparatus,

21
22
23
24
25
26
27
28

’436 patent,
claim 1

utilizing said index to said data base of natural language to access word sense
numbers and/or function codes with said associated additional data to match or
partially match said selected word sense numbers and/or function codes
associated with said natural language which is associated with said electronically
encoded data with said apparatus.
A method of processing natural language in an apparatus, which comprises steps:
providing natural language which is processed by said apparatus to provide
electronically encoded data which is representative of said natural language,
providing a dictionary data base in memory associated with said apparatus
7

wherein said dictionary data base contains a plurality of entries which are
comprised of one or more of syntax usage data, associated word sense numbers
with associated state representation data, and/or function codes,

1
2

lexically processing said electronically encoded data to access said dictionary
data base with said apparatus,

3
4

providing a natural language plausibility and expectedness processor in said
apparatus,

5
6
7
8

’436 patent,
claim 7

providing a context data base wherein said context data base contains a plurality
of entries which are comprised of one or more of word sense numbers having
associated state representation data, and/or function codes,

9
10

United States District Court
Northern District of California

11

utilizing said natural language plausibility and expectedness processor to initiate
accessing entries of said dictionary data base which are associated with words of
said natural language.
A method of processing as defined in claim 1 which comprises steps:

utilizing said context data base for processing natural language.
The parties’ dispute over the meaning of “word sense number” focuses on three issues,

12

corresponding to the three parts of defendants’ construction: (1) whether the term should be

13

defined as “[a]n address to the meaning of a word;” (2) whether the term should be defined to

14

require that the “meaning data” be “utilized” and “organized” in certain ways; and (3) whether the

15

term should be defined to require certain structures for adjective, concrete noun, state abstract

16

noun, and verb word sense numbers. See Opening Br. at 5-9; Defs. Br. at 4-9; Reply Br. at 2-7

17

(Dkt. No. 105).

18

1.

“address to the meaning of a word”

19

Defendants argue that “word sense number” should be construed to mean “address to the

20

meaning of a word” because the term is explicitly defined in this way in the abstracts for each of

21

the patents-in-suit except the ’468 patent. Defs. Br. at 4-6; see also, e.g., ’087 patent at abstract

22

(“A word sense number is an address to the meaning of a word.”); ’603 patent at abstract (same);

23

’436 patent at abstract (same). WTI counters that the intrinsic record as a whole supports using

24

“identifier for a word meaning” instead of “address to the meaning of a word.” Opening Br. at 5-

25

6. WTI also argues that use of the word “address” “invites confusion” because of its “well-known

26

meaning . . . [i]n the field of computer technology” as “a specific location in computer memory.”

27

Id. at 6-7.

28

I agree with defendants. The parties do not dispute that “word sense number” has no
8

1

ordinary meaning in the art. See Defs. Br. at 4 (noting that “WTI does not attempt to show that

2

‘word sense number’ has an accepted ordinary meaning in the field”); see also Opening Br. at 5-9;

3

Reply Br. at 2-7. Accordingly, it was up to Budzinski to “provide a precise definition for the . . .

4

term,” Irdeto, 383 F.3d at 1300, which is exactly what he appears to have attempted to do in

5

explicitly stating in the abstracts for all but one of patents-in-suit what a word sense number “is.”

6

Cf. Sinorgchem Co., Shandong v. Int’l Trade Comm’n, 511 F.3d 1132, 1136 (Fed. Cir. 2007)

7

(“Moreover, the word ‘is,’ again a term used here in the specification, may signify that a patentee

8

is serving as its own lexicographer.”); accord Abbott Labs. v. Andrx Pharm., Inc., 473 F.3d 1196,

9

1210 (Fed. Cir. 2007). Indeed, Budzinski specifically amended the original abstract to include this
sentence following the examiner’s initial rejection of the ’468 patent as unpatentable over

11
United States District Court
Northern District of California

10

Katayama.

12

WTI contends that a word sense number is better understood as an “identifier for a word

13

meaning,” but WTI fails to cite anything in the specification or prosecution history that provides

14

meaningful support for this position. See Opening Br. at 5-6; Reply Br. at 3. Although the

15

specification describes certain types of word sense numbers as containing “identification

16

numbers,” see, e.g., ’468 patent at 8:63-65 (“The word sense number of a state abstract noun

17

contains an identification number.”); id. at 9:67-10:02 (“A verb word sense number contains an

18

identification number.”), that is not the same as defining “word sense number” as “identifier for a

19

word meaning.” And, although the specification uses the word “identifier” in various contexts,

20

WTI fails to identify a single instance when the specification uses that word in connection with a

21

description of word sense numbers.

22

Given the clear definitional language in the abstracts, and the absence of intrinsic evidence

23

to support WTI’s construction of “word sense number” as “identifier for a word meaning,” WTI’s

24

concern that using the word “address” “invites confusion” is not persuasive. “Under [Federal

25

Circuit] precedent, the patentee’s lexicography must govern the claim construction analysis.”

26

Braintree Labs., Inc. v. Novel Labs., Inc., 749 F.3d 1349, 1356 (Fed. Cir. 2014). A word sense

27
28
9

1

number is an “address to the meaning of a word.”3
2.

2
3
4

“meaning data that is (1) utilized to determine the intended meaning of
a word usage, and (2) organized into relations to other word sense
numbers that can be used for selecting a word sense number which has
the intended meaning of a word contained in natural language”

Defendants also seek to define the term “word sense number” to require “meaning data that

5

is (1) utilized to determine the intended meaning of a word usage, and (2) organized into relations

6

to other word sense numbers that can be used for selecting a word sense number which has the

7

intended meaning of a word contained in natural language.” Defs. Br. at 6-7.

8
9
10

In support of this portion of their construction, defendants rely on statements made by
Budzinski during prosecution. In distinguishing Vanderwende during prosecution of the ’509 and
’603 patents, Budzinski explained that

United States District Court
Northern District of California

11

[w]ord sense numbers are a novel, nonobvious invention not taught
in Vanderwende. A word sense number, which is described below,
has an associated state representation which makes it possible to
have new capabilities compared to the organized text words . . . in
Vanderwende . . . First, I will describe the features of word sense
numbers and some of the new capabilities of the present invention
that are made possible with word sense numbers, and I will compare
word sense numbers and these capabilities with Vanderwende . . . A
word sense number has meaning data which is utilized to
determine the intended meaning of a word usage. The meaning
data of a word sense number is organized into relations to other
word sense numbers, and these relations can be used for selecting
a word sense number which has the intended meaning of a word
contained in natural language.

12
13
14
15
16
17
18

’509 file at 5-6 (emphasis added); ’603 file at 14-15 (emphasis added).

19

Similarly, during prosecution of the ’468 patent, Budzinski distinguished Katayama by

20
stating,
21

Word sense numbers are a novel, nonobvious invention not taught in
Katayama . . . The meaning data of a word sense number is
organized into relations to other word sense numbers, and these
relations can be used for selecting word sense numbers which
have the intended definition sense of a word for a usage in text.

22
23
24

’468 file at 38-39 (emphasis added). When the examiner rejected the claims of the ’468 patent

25

because of the indefiniteness of terms like “word sense number,” Budzinski relied on this

26
27
3

28

At the hearing, WTI stated that it agreed with this portion of the Court’s construction. Hearing
Tr. at 6 (Dkt. No. 114).
10

1

language in responding to the rejection. The examiner found that the claims were
indefinite for failing to particularly point out and distinctly claim the
subject matter which applicant regards as the invention. The claims
are replete with alternate expressions such as “natural language text
word or said non-textual natural language equivalent” and “word
sense numbers or function word codes” which renders the claims
vague and indefinite.

2
3
4
5

Mead Decl. Ex. C at WTI00003542-43 (Dkt. No. 103-4). Budzinski responded, in relevant part,

6

by explaining that “[w]ord sense numbers are described in the applicant’s June 25, 1996 response

7

on page 38, line 17 to page 40, line 5.” Mead Decl. Ex. D at WTI00004215 (Dkt. No 103-5). That

8

portion of Budzinski’s June 25, 1996 response includes the language quoted above.

9

Budzinski’s prosecution statements support defining the term “word sense number” to
require “meaning data that is (1) utilized to determine the intended meaning of a word usage, and

11
United States District Court
Northern District of California

10

(2) organized into relations to other word sense numbers.”4 WTI contends that the statements

12

describe only “advantageous features” of the claimed invention as a whole as opposed to

13

“definitional limitations” of word sense numbers in particular. Reply Br. at 4-5. Similarly, WTI

14

argues that the statements are better understood as a description of “how the combination of word

15

sense numbers, state representation data, and other features of the invention are used.” Id.

16

(emphasis omitted). These arguments ignore Budzinski’s repeated use of definitional language

17

aimed specifically at word sense numbers. See, e.g., ’468 file at 39 (“The meaning data of a word

18

sense number is . . .”); id. at 45 (“[W]ord sense numbers have . . .”); ’509 file at 5 (“Word sense

19

numbers are a novel, nonobvious invention . . .”); id. at 6 (“I will describe the features of word

20

sense numbers . . .”); id. (“A word sense number has . . .”). They also ignore Budzinski’s explicit

21

use of his prosecution statements to define word sense numbers in responding to the examiner’s

22

indefiniteness rejection. In light of these circumstances, and in particular given the absence of any

23

evidence that the term “word sense number” has a customary and ordinary meaning in the art,

24

Budzinski’s prosecution statements are not reasonably understood except as explanations of the

25

meaning of the term.

26

On the other hand, I am not convinced that the phrase, “that can be used for selecting a

27
28

4

At the hearing, WTI also agreed with this portion of the Court’s construction. Hearing Tr. at 6.
11

1

word sense number which has the intended meaning of a word contained in natural language,” is

2

properly added to the construction of “word sense number.” Although Budzinski’s prosecution

3

statements plainly link word sense numbers to meaning data that is (1) “utilized to determine the

4

intended meaning of a word usage,” and (2) “organized into relations to other word sense

5

numbers,” his statements regarding how those relations “can be used” are less clear. Defendants

6

offer no other basis for this portion of their construction. Absent a more clear basis for the

7

portion, I do not find it appropriate at this time.

8
9

3.

Format Limitations

Defendants further argue that Budzinski defined certain types of word sense numbers to
require particular formats. Defs. Br. at 7-9. Specifically, defendants contend that Budzinski

11
United States District Court
Northern District of California

10

provided the following formats for “adjective,” “concrete noun,” “state abstract noun,” and “verb”

12

word sense numbers: (1) “An adjective word sense number is composed of an identification

13

number, a state value or value range, and an owner word sense number.” (2) “The word sense

14

number of a concrete noun contains a word sense identifying number, a type number, a specificity

15

number, and an experience number.” (3) “The word sense number of a state abstract noun

16

contains an identification number, a type number, a specificity number, and an experience

17

number.” (4) “A verb word sense number contains an identification number which defines the

18

verb word sense number, and includes partial to complete word sense identification numbers of

19

main sentence roles.” Id.

20

These format limitations are properly incorporated into the construction of “word sense

21

number.” Defendants point out that the formats are taken verbatim from the “Summary of the

22

Invention” section of the specification. See ’468 patent at 6:15-19 (adjective word sense

23

numbers); id. at 6:37-39 (concrete noun word sense numbers); id. at 8:63-67 (state abstract noun

24

word sense numbers); id. at 9:67-10:03 (verb word sense numbers). During prosecution,

25

Budzinski repeatedly cited to these portions of the specification in describing word sense numbers.

26

See ’468 file at 38-39 (noting where in the specification adjective, concrete noun, state abstract

27

noun, and verb word sense numbers are “summarized” and “described in detail”); ’509 file at 6-8

28

(same); ’603 file at 15-17 (same). Further, in responding to the indefiniteness rejection during
12

1

prosecution of the ’468 patent and explaining word sense numbers to the examiner, Budzinski

2

directed the examiner to the pages of his June 25, 1996 response that cite to these same portions of

3

the specification. See Mead Decl. Ex. C at WTI00003542-43 (“Word sense numbers are described

4

in the applicant’s June 25, 1996 response on page 38, line 17 to page 40, line 5.”).
WTI asserts that incorporation of the format limitations would limit word sense numbers to

5
6

particular implementations of the invention, but WTI cites nothing in the specification or

7

prosecution history that describes or implies the existence of alternative implementations of word

8

sense numbers or otherwise supports a broader understanding of the term. See Opening Br. at 7-9;

9

Reply Br. at 5-7. WTI does point out that claim 3 of the ’509 patent specifically recites
“experience and knowledge” and “directed graphs,” and that during prosecution Budzinski also

11
United States District Court
Northern District of California

10

described verb word sense numbers as “hav[ing] associated data including experience and

12

knowledge which is stored in terms of directed graph paths of word sense numbers.” ’509 file at

13

9. WTI contends that this shows that when Budzinski wanted to incorporate specific data fields

14

into the meaning of “word sense number,” he did not just reference them in the specification or

15

during prosecution – he specifically recited them in the claims. Opening Br. at 9. However, in

16

contrast with the language highlighted by defendants and discussed above, the single prosecution

17

statement highlighted by WTI concerns data that is “associated” with verb word sense numbers,

18

not the data fields that certain word sense numbers are “composed of” or “contain.” See ’509 file

19

at 9. Further, neither claim 3 of the ’509 patent, nor claim 1 on which it depends, uses the term

20

“word sense number.” Both claims use only the term “clause implying word sense number,”

21

discussed below.

22

B.

“clause implying word sense number”
WTI’s Construction

23

Claims

24

’087 patent, A word sense number for
claim 23
a verb, adjective, or
abstract noun that
’509 patent, represents a clause.
claim 9, 10,
16

25
26
27
28

’603 patent,
claim 14

Defendants’
Construction
A word sense number for
a verb, adjective, or
abstract noun that
represents a clause, and
has (1) an associated state
representation which
includes conditions and
requirements which are
met for the clause
13

Court’s Construction
WTI’s construction.

1
2
3
4

’468 patent,
claim 12,
31
’840 patent,
claim 15,
16

5
6
7
8
9
10

’840 patent,
claim 15

Example Claims
A method of processing natural language, which comprises steps
providing electronically encoded data which are representative of said natural
language,

11
United States District Court
Northern District of California

implying word sense to
have stored relations, and
thus be valid, (2)
associated purpose
relations which are
organized by the concept,
i.e. function, of the
purpose relation and
stored in entries of a
purpose node, and
(3) word sense numbers
which have been selected
utilizing the state
representation data to
have all relations of the
clause to be stored.

12

providing a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax usage data,
associated clause implying word sense numbers having associated state
representation data,

13
14
15

lexically processing said electronically encoded data to access said dictionary
data base,

16

providing a grammar specification,

17

providing a context data base wherein said context data base contains a plurality
of entries which are comprised of one or more of clause implying word sense
numbers having associated state representation data,

18
19
20
21
22
23
24
25
26
27
28

’509 patent,
claim 9

utilizing one or more of said syntax usage data and said clause implying word
sense numbers which are from entries of said dictionary data base and which are
associated with words of said natural language with reference to said grammar
specification and with reference to said context data base to select clause
implying word sense numbers associated with said natural language words.
A method of processing natural language in an apparatus, which comprises steps
providing in memory associated with said apparatus an experience and
knowledge data base which is comprised of directed graphs comprised of nodes
with associated clause implying word sense numbers organized into paths of said
nodes such that said nodes have access conditions which determine zero or more
next said nodes on zero or more said paths that are accessible,
utilizing a natural language processor to provide natural language with associated
clause implying word sense numbers in memory associated with said apparatus,
purpose relation path identification processing with said apparatus to find zero or
more said paths from said nodes associated with said clause implying word sense
numbers associated with said natural language with reference to said experience
14

and knowledge data base such that said access conditions of said nodes on said
found paths are met,

1
2

providing criteria for selecting said found experience and knowledge paths in
memory associated with said apparatus,

3
4
5
6
7
8
9
10

United States District Court
Northern District of California

11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

’509 patent,
claim 10

utilizing said criteria to select one or more of said found paths with said
apparatus.
A method of processing as defined in claim 9, which comprises steps
providing a context data base wherein said context data base contains a plurality
of entries which are comprised of one or more of clause implying word sense
numbers having associated state representation data including associated
experience and knowledge paths,
selecting experience and knowledge paths associated with said clause implying
word sense numbers associated with said natural language such that said
experience and knowledge paths associated with said clause implying word sense
numbers associated with said natural language have accessable [sic] paths to said
experience and knowledge paths associated with said clause implying word sense
numbers from said context data base entries.

The parties do not dispute that “clause implying word sense number” lacks a customary
and ordinary meaning in the art, or that it requires at least “a word sense number for a verb,
adjective, or abstract noun that represents a clause.” See Opening Br. at 10-11; Defs. Br. at 9-10;
Reply Br. at 7. However, defendants contend that the term should be defined to also include three
additional elements: “(1) an associated state representation which includes conditions and
requirements which are met for the clause implying word sense to have stored relations, and thus
be valid, (2) associated purpose relations which are organized by the concept, i.e. function, of the
purpose relation and stored in entries of a purpose node, and (3) word sense numbers which have
been selected utilizing the state representation data to have all relations of the clause to be stored.”
Defs. Br. at 9-10.
In support of their construction, defendants again rely on statements made by Budzinski
during prosecution of the ’509 patent. Defs. Br. at 9-10. They cite the following passage to
support the first element of their construction:
The clause implying word sense numbers associated with natural
language of the present invention have an associated state
representation which includes conditions and requirements which
are met for the clause implying word sense to have stored
relations, and thus be valid. The present invention has a process
that selects clause implying word sense numbers to have such stored
relations, and thus be valid. In contrast for Vanderwende, a sense
number is a numbered definition of an online dictionary . . . which is
15

1

appended to a text word.
’509 file at 16 (emphasis added). They cite the following passage to support the second element:

2

In Vanderwende, purpose is only used in the sense of goal or
intention . . . A clause implying word sense [number] has
associated purpose relations which are organized by the concept,
i.e. function, of the purpose relation, and these purpose relations
are stored in entries of a purpose node.

3
4
5

Id. at 12 (emphasis added). And they cite the following passage to support the third element:

6

The clause implying word sense numbers associated with natural
language have word sense numbers which have been selected
utilizing the state representation data to have all relations of the
clause to be stored . . . In contrast for Vanderwende, a sense number
is a numbered definition of an online dictionary . . . which is
appended to a text word

7
8
9

Id. at 39-40 (emphasis added).
10
For its construction, WTI relies on the following portion of the specification, which
11
United States District Court
Northern District of California

Budzinski identified during prosecution as describing cause implying word sense numbers:
12

Clauses are represented by word sense numbers of verbs, adjectives,
or abstract nouns in Experience and Knowledge Memory 150. Verb
word sense numbers are directly convertible into natural language
clauses. A characterizing clause associated with a clausal abstract
noun is also directly convertible into a natural language clause.
Clauses of adjectives are realized with the owner of the adjective as
a subject with the adjective as a subject complement such as: “John
is sick.” State abstract nouns are expressed as the owner of the state
abstract noun, a form of “to have” with a “to possess” word sense
and the state abstract noun such as: “John has good health.” Nouns
and relations between nouns can have associated clauses which
belong to purpose paths in Memory 150. Thus, all types of state
representation words can have related experience and knowledge in
Memory 150.

13
14
15
16
17
18
19
20
21
22
23
24

’468 patent at 13:43-57; see also ’468 file at 37 (“I do not mean that a clause implies a word sense
number in these claims. Rather, I mean that a word sense number implies a clause, i.e., a clause
can be formed with the data associated with the word sense number. Clause implying word sense
numbers are described in the specification on page 23, lines 4-165); ’603 file at 18 (noting where
in the specification clause implying word sense numbers are summarized).

25
Before the hearing, I was tentatively inclined to adopt defendants’ construction. See Dkt.

26
27
5

28

The citation to “page 23, lines 4-16” corresponds to the ’468 patent at 13:43-57. See, e.g.,
Opening Br. at 10 (noting same).
16

No. 111 (Tentative Rulings re June 17, 2016 Claim Construction Hearing). On further

2

consideration, however, I find that Budzinski’s prosecution statements do not provide a sufficient

3

basis for inserting the additional elements proposed by defendants. Those statements lack the

4

unambiguously definitional nature of those regarding, e.g., the meaning of “word sense number,”

5

and defendants identify nothing else in the prosecution history or specification that supports their

6

construction. Meanwhile, WTI’s construction is supported by both the specification and

7

Budzinski’s citation to the specification during prosecution. See ’468 patent at 13:43-57; ’468 file

8

at 37. Indeed, defendants do not dispute that WTI’s construction accurately reflects the

9

specification passage cited by Budzinski – defendants’ only complaint is that WTI’s construction

10

is incomplete. Because I find that defendants’ construction would unduly limit and complicate the

11
United States District Court
Northern District of California

1

meaning of “clause implying word sense number,” I adopt WTI’s construction as the Court’s.
C.

12
13
14
15
16
17
18
19
20
21
22
23
24
25
26

“state representation data”

Defendants’
Court’s Construction
Construction
’087 patent, Meaning data composed
Meaning data composed
WTI’s construction.
claims 1, 4, of states, their values, and of states, their values, and
17
their relations, associated their relations, associated
with word sense numbers. with word sense numbers
’091 patent,
that includes
claims 1,
requirements and access
12
conditions which are
utilized for selecting a
’840 patent,
word sense number which
claims 1, 3,
has the intended meaning
15, 16
of a word contained in the
natural language.
’509 patent,
claims 10,
16
Claims

WTI’s Construction

’436 patent,
claims 1, 7,
11, 14
’468 patent,
claims 1,
33
Example Claims

27
28
17

1

’087 patent,
claim 1

providing electronically encoded data which is representative of said natural
language,

2
3

providing a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax usage data,
associated word sense numbers having associated state representation data,

4
5

lexically processing said electronically encoded data to access said dictionary
data base,

6
7
8
9

’087 patent,
claim 4

accessing said state representation data of said word sense number having said
relation.

United States District Court
Northern District of California

11

13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

utilizing said syntax usage data and said word sense numbers which are from
entries of said dictionary data base and which are associated with words of said
natural language to access said state representation data.
A method of processing as defined in claim 1, which comprises steps
providing a relation between two or more said word sense numbers,

10

12

A method of processing natural language, which comprises steps

’509 patent,
claim 10

10. A method of processing as defined in claim 9, which comprises steps
providing a context data base wherein said context data base contains a plurality
of entries which are comprised of one or more of clause implying word sense
numbers having associated state representation data including associated
experience and knowledge paths,
selecting experience and knowledge paths associated with said clause implying
word sense numbers associated with said natural language such that said
experience and knowledge paths associated with said clause implying word sense
numbers associated with said natural language have accessable paths to said
experience and knowledge paths associated with said clause implying word sense
numbers from said context data base entries
9. A method of processing natural language in an apparatus, which comprises
steps
providing in memory associated with said apparatus an experience and
knowledge data base which is comprised of directed graphs comprised of nodes
with associated clause implying word sense numbers organized into paths of said
nodes such that said nodes have access conditions which determine zero or more
next said nodes on zero or more said paths that are accessible,
utilizing a natural language processor to provide natural language with associated
clause implying word sense numbers in memory associated with said apparatus,
purpose relation path identification processing with said apparatus to find zero or
more said paths from said nodes associated with said clause implying word sense
numbers associated with said natural language with reference to said experience
and knowledge data base such that said access conditions of said nodes on said
found paths are met,
providing criteria for selecting said found experience and knowledge paths in
memory associated with said apparatus,
18

1
2
3

utilizing said criteria to select one or more of said found paths with said
apparatus.
The parties do not dispute that “state representation data” has a meaning unique to the
patents-in-suit and that the meaning of the term should be limited at least to “meaning data

4
composed of states, their values, and their relations, associated with word sense numbers.” See
5
Opening Br. at 11-12; Defs. Br. at 11; Reply Br. at 8. However, defendants contend that the
6

“meaning data” in this construction must also include “requirements and access conditions which

7
are utilized for selecting a word sense number which has the intended meaning of a word
8
9
10

United States District Court
Northern District of California

11
12

contained in the natural language.” Defs. Br. at 11. In support of this construction, defendants
rely on Budzinski’s repeated explanation during prosecution that
The state representation data associated with word sense
numbers includes requirements and access conditions
which are utilized for selecting a word sense number which
has the intended meaning of a word contained in natural
language.

13

’509 file at 30, 34, 50, 54, 73 (emphasis added). According to defendants, WTI is “bound by these

14

statements.” Defs. Br. at 11.

15
16
17

WTI responds that its construction aligns with the specification’s description of the
invention as “storing all that is known for the definition and all that is known to be related to the
definition by realizing the definition with a state representation which is in terms of states, their

18

values, and/or their relations.” ’468 patent at 3:60-63; see also Opening Br. at 11-12. WTI also

19

argues that defendants’ construction “adds complexity without adding clarity” and amounts to “a

20
21
22
23
24

distinction without a difference, as defendants have not shown how their [construction] materially
alters the scope of the claims.” Reply Br. at 8; see also Opening Br. at 12 (arguing same).
I agree with WTI’s construction. Defendants do not dispute that this construction is
supported by the specification, and as WTI points out, defendants offer no explanation of how,
practically speaking, their construction would materially alter the scope of the asserted claims.

25

Given the support for WTI’s construction in the specification, and the absence of evidence that

26

Budzinski’s prosecution statements could be read as a disavowal of actual claim scope with

27

respect to the term “state representation data,” I cannot say at this juncture that those statements

28
19

1

are an appropriate basis for further limiting the meaning of the term.
D.

2

“grammar specification”
WTI’s Construction

3

Claims

4

’091 patent, Specification of rules
claims 1,
defining a language’s
12
syntactic structure.

5
6
7
8
9
10

United States District Court
Northern District of California

11

’840 patent,
claims 1,
15
’509 patent,
claim 16
’603 patent,
claim 9

Court’s Construction
WTI’s construction.

’468 patent,
claim 1, 33

12
13

Defendants’
Construction
Specification of the rules
defining a natural
language’s syntactic
structure that must be
used to parse incoming
natural language and
form natural language
output, represent natural
language, and select word
senses for natural
language words.

’091 patent,
claim 1

14

Example Claims
A method of processing natural language, which comprises steps
providing electronically encoded data which is representative of said natural
language,
providing a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax usage data,
associated word sense numbers having associated state representation data, and/or
function codes,

15
16
17
18

lexically processing said electronically encoded data to access said dictionary
data base,

19

providing a grammar specification,

20

providing a data base of requirements such that said requirements must be met by
said associated state representation data of said word sense numbers for said word
sense numbers to be selected,

21
22
23
24
25
26

’840 patent,
claim 15

utilizing said syntax usage data which are from entries of said dictionary data
base and which are associated with words of said natural language with reference
to said grammar specification to select word sense numbers associated with said
natural language words such that said associated state representation data of said
associated word sense numbers meet said requirements for selecting said
associated word sense numbers.
A method of processing natural language, which comprises steps
providing electronically encoded data which are representative of said natural
language,

27
28

providing a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax usage data,
20

associated clause implying word sense numbers having associated state
representation data,

1
2

lexically processing said electronically encoded data to access said dictionary
data base,

3

providing a grammar specification,

4

providing a context data base wherein said context data base contains a plurality
of entries which are comprised of one or more of clause implying word sense
numbers having associated state representation data,

5
6
7
8
9
10

United States District Court
Northern District of California

11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

’468 patent,
claim 1

utilizing one or more of said syntax usage data and said clause implying word
sense numbers which are from entries of said dictionary data base and which are
associated with words of said natural language with reference to said grammar
specification and with reference to said context data base to select clause
implying word sense numbers associated with said natural language words.
A method of processing natural language, which comprises steps
providing electronically encoded data which is representative of said natural
language,
providing a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax usage data,
associated word sense numbers having associated state representation data and/or
function codes,
lexically processing said electronically encoded data to access said dictionary
data base,
providing a grammar specification,
utilizing said syntax usage data which are from entries of said dictionary data
base and which are associated with words of said natural language with reference
to said grammar specification to produce output data representative of a
grammatical parse of said natural language, said output data including selected
syntax usage.

The parties agree that a grammar specification is a “[s]pecification of rules defining a
language’s syntactic structure.” See Opening Br. at 12-13; Defs. Br. at 12-14; Reply Br. at 8-9.
Defendants’ construction includes two additional requirements. First, it specifies that the
grammar specification is limited to natural languages. Defs. Br. at 12-13. Second, it specifies that
the grammar specification is “used to parse incoming natural language and form natural language
output, represent natural language, and select word senses for natural language words.” Id.
I agree with WTI’s construction. With respect to specifying natural languages, defendants
do not offer any evidence to dispute WTI’s assertion that a person of ordinary skill in the art
“would understand that every language – natural or artificial – has a grammar which defines the
21

1

language’s structure.” Opening Br. at 13. Defendants contend that the meaning of “grammar

2

specification” is nevertheless appropriately limited to natural languages because claim 1 of the

3

’468 patent repeatedly uses the term “natural language.” See Defs. Br. at 12 (“The ‘grammar

4

specification’ in this claim must be rules defining natural language because that is what is being

5

processed as well as being used to produce a ‘grammatical parse.’”). But this point merely

6

highlights that incorporating a natural language limitation into the meaning of “grammar

7

specification” would be redundant; claim 1 of the ’468 patent, as well as all of the other asserted

8

claims in which the term “grammar specification” appears, already recite “natural language” and,

9

presumably, are thus limited in this way. See Reply Br. at 9 (noting that an instrumentality that
“does not involve natural language as recited in th[e] limitations is already excluded from the

11
United States District Court
Northern District of California

10

scope of the invention”). The various portions of the specification cited by defendants also fail to

12

establish that the scope of grammar specification should be limited to natural languages. See Defs.

13

Br. at 12-13. For example, although the specification uses the phrase “natural language” in

14

distinguishing a prior art reference, the specification plainly distinguishes that reference on

15

grounds other than the use of a natural language grammar – according to the specification, the

16

reference itself discloses use of a natural language grammar. See ’468 patent at 2:51-60 (noting

17

that the reference “discloses a thorough description of English grammar”).

18

With respect to the functional limitation that defendants seek to insert, defendants cite to

19

the specification’s distinction of a prior art reference on the ground that the reference’s “grammar

20

description does not include a method for representing natural language nor does it include a

21

method for selecting word senses of natural language words.” ’468 patent at 2:57-60. Defendants

22

also cite to similar statements elsewhere in the specification and in the prosecution history. Defs.

23

Br. at 13-14. However, the statements are largely directed at how a grammar specification can be

24

used as opposed to what a grammar specification is. In line with this distinction, several of the

25

asserted claims already recite the functional limitations that defendants seek to squeeze into the

26

term “grammar specification.” For example, the function of parsing incoming natural language is

27

recited in claim 1 of the ’468 patent (“utilizing said syntax usage data . . . to produce output data

28

representative of a grammatical parse of said natural language”); the function of forming natural
22

1

language output is recited in claim 33 of the ’468 patent (“utilizing said syntax usage data and said

2

natural language words . . . to generate outgoing natural language”); and the function of selecting

3

word sense numbers for natural language words is recited in claim 1 of the ’091 patent (“utilizing

4

said syntax usage data . . . to select word sense numbers associated with natural language words”).
E.

5

“lexically processing”
WTI’s Construction

6

Claims

7

’087 patent, Accessing syntax usage
claim 1, 17 data, associated word
sense numbers, and/or
’091 patent, function codes of a word
claim 1, 12 in a dictionary data base.

8
9
10

United States District Court
Northern District of California

11
12
13
14
15

’840 patent,
claim 1, 15
’603 patent,
claim 9
’436 patent,
claim 1, 11
’468 patent,
claim 1

16
17
18
19

’468 patent,
claim 1

Defendants’
Construction
(1) Processing each word
by accessing syntax usage
data, associated word
sense numbers, and/or
function codes of the
word in a dictionary data
base.
OR

Court’s Construction
Defendants’ first
construction:
Processing each word by
accessing syntax usage
data, associated word
sense numbers, and/or
function codes of the
word in a dictionary data
base.

(2) Processing each word
by accessing in a
dictionary data base the
syntax usage data for
each word, the associated
word sense numbers for
each word that is a
meaning word, and the
function codes for each
word that is a function
word.
Example Claims
A method of processing natural language, which comprises steps
providing electronically encoded data which is representative of said natural
language,

20
providing a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax usage data,
associated word sense numbers having associated state representation data and/or
function codes,

21
22
23

lexically processing said electronically encoded data to access said dictionary
data base,

24
providing a grammar specification,
25
26
27
28

’087 patent,

utilizing said syntax usage data which are from entries of said dictionary data
base and which are associated with words of said natural language with reference
to said grammar specification to produce output data representative of a
grammatical parse of said natural language, said output data including selected
syntax usage.
A method of processing natural language, which comprises steps
23

1

claim 1
providing electronically encoded data which is representative of said natural
language,

2

providing a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax usage data,
associated word sense numbers having associated state representation data,
lexically processing said electronically encoded data to access said dictionary
data base,

3
4
5
6
7
8
9
10

United States District Court
Northern District of California

11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

’087 patent,
claim 17

utilizing said syntax usage data and said word sense numbers which are from
entries of said dictionary data base and which are associated with words of said
natural language to access said state representation data.
A method of processing natural language, which comprises steps
providing electronically encoded data which is representative of said natural
language,
providing a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax usage data,
associated word sense numbers having associated state representation data, and/or
function codes,
lexically processing said electronically encoded data to access said dictionary
data base,
utilizing said syntax usage data and said word sense numbers and/or said function
codes which are from entries of said dictionary data base and which are
associated with words of said natural language with reference to said associated
state representation data to select word sense numbers and/or function codes for
words of said natural language.

The parties’ dispute over the meaning of “lexically processing” centers on whether
lexically processing requires the processing of each word in a natural language phrase (or, as WTI
puts it, whether “every data entity” in the lexically processing step “must correspond to exactly
one textual word of incoming natural language”). See Opening Br. at 16-17; Defs. Br. at 17-19;
Reply Br. at 11-12.
WTI defends its construction – which it contends does not require the processing of each
word – on the grounds that it is taken verbatim from Budzinksi’s statements during prosecution,
and that the specification indicates that the claimed invention may lexically process whole phrases
at a time. Opening Br. at 16-17; see also ’603 file at 33, 62, 82 (“In the present invention, lexical
processing accesses syntax usage data, associated word sense numbers, and/or function codes of a
word in [a] dictionary data base of the present invention.”).
Defendants respond that “[t]he literature in the field of natural language processing shows
24

1

that lexical processing operates on the individual word level,” and that “[t]he explanations

2

throughout the specification always contemplate the processing of each word.” Defs. Br. at 17.

3

Defendants submit a 1992 article, “Progress in the Application of Natural Language Processing to

4

Information Retrieval Tasks,” stating that “[f]or [natural language processing], lexical processing

5

operates at the single word level and involves identifying words and determining their

6

grammatical classes or parts of speech so that higher levels language analysis can take place.”

7

Mead Decl. Ex. H (Dkt. No. 103-9). Defendants also point to the specification’s figure 5A, which

8

appears to depict processing on a word-by-word basis, and the following language from the

9

“Summary of the Invention” portion of the specification: “It is an object of this invention to
provide a new and improved natural language syntax processing method for separating incoming

11
United States District Court
Northern District of California

10

natural language into each word’s sentence role.” ’468 patent at 20:12-15.

12

I agree with defendants that “lexically processing” as used in the claims is properly read to

13

require processing on a word-by-word basis. WTI does not dispute that the 1992 article accurately

14

reflects the customary and ordinary meaning of “lexically processing” and offers no competing

15

evidence on the customary and ordinary meaning of the term. See Opening Br. at 16-17; Reply

16

Br. at 11-12. WTI contends that the prosecution history supports a broader reading of lexically

17

processing than the customary and ordinary meaning. But the single prosecution statement WTI

18

relies on does not squarely address whether lexically processing must operate on a word-by-word

19

basis. See ’603 file at 33, 62, 82. If anything, the prosecution statement appears to support

20

defendants’ position, in that Budzinksi stated that “lexical processing accesses syntax usage data,

21

associated word sense numbers, and/or function codes of a word in [a] dictionary data base of the

22

present invention.” ’603 file at 33, 62, 82 (emphasis added).

23

The specification also fails to support a broader reading of “lexically processing.” WTI

24

does not dispute that figure 5A and the “Summary of the Invention” language cited by defendants

25

supports defendants’ narrow construction. WTI cites its own portions of the specification, but I

26

agree with defendants that these portions do not establish a broader reading of the term. First,

27

while the specification describes storing “idioms” in the “Dictionary,” ’468 patent at 5:06-09,

28

immediately thereafter the specification states that “Dictionary Look-Up Step looks up the syntax
25

wordsets which each input word belongs to,” id. at 5:09-12 (emphasis added). WTI does not

2

dispute that this language indicates that the lexically processing step “would still need to process

3

each word in order to determine that the natural language contained an idiom.” Defs. Br. at 17-18.

4

Second, while the specification describes how “[w]ords formed with a verb base (e.g.,

5

‘surprisingly’) often imply a clause for complete interpretation,” ’468 patent at 5:61-63, WTI does

6

not explain how this shows that lexically processing operates other than on a word-by-word basis.

7

That one word can “imply a clause for complete interpretation” does not necessarily mean that the

8

invention can lexically process a complete clause without lexically processing each word in the

9

clause. See Defs. Br. at 18 (making same point). Finally, while the specification describes

10

“perform[ing] ellipsis processing to replace ellipted words, i.e., left out words, and then . . .

11
United States District Court
Northern District of California

1

perform[ing] processing which determines if the replaced words are consistent with the context of

12

the conversation and stored experience and knowledge,” ’468 patent at 3:44-48, WTI again fails to

13

explain how the processing of additional, ellipted words means that lexically processing operates

14

other than on a word-by-word basis.
F.

15

“syntax usage data”
WTI’s Construction

16

Claims

17

’087 patent, Information indicating
claim 1, 17 how a word or words can
be used in relation to
’091 patent, other words.
claim 1, 12

18
19
20
21
22
23
24
25
26
27
28

’840 patent,
claim 1, 15

Defendants’
Construction
Data comprised of sets of
words which can
syntactically be used
interchangeably in a
natural language
construction.

Court’s Construction
Defendants’ construction.

’509 patent,
claim 16
’603 patent,
claim 9
’436 patent,
claim 1
’468 patent,
claim 1, 33
Example Claims
’468 patent, A method of processing natural language, which comprises steps
26

1
2
3
4
5
6
7
8
9
10

United States District Court
Northern District of California

11
12
13
14

claim 1
providing electronically encoded data which is representative of said natural
language,
providing a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax usage data,
associated word sense numbers having associated state representation data and/or
function codes,
lexically processing said electronically encoded data to access said dictionary
data base,
providing a grammar specification,
utilizing said syntax usage data which are from entries of said dictionary data
base and which are associated with words of said natural language with reference
to said grammar specification to produce output data representative of a
grammatical parse of said natural language, said output data including selected
syntax usage.
’603 patent, A method of processing natural language in an apparatus, which comprises steps
claim 1
utilizing a natural language processor to provide a data base of natural language
with associated word sense numbers and/or function codes in memory associated
with said apparatus,
associating additional data with said word sense numbers and/or function codes
associated with said data base of natural language in memory associated with said
apparatus,

15

indexing said data base of natural language with respect to word sense numbers
and/or function codes with said apparatus,

16

storing said index in memory associated with said apparatus,

17

providing electronically encoded data which is representative of natural language
in memory associated with said apparatus,

18
19
20
21
22
23
24
25
26
27
28

providing a dictionary data base in memory associated with said apparatus
wherein said dictionary data base contains a plurality of entries which are
comprised of one or more of syntax usage data, associated word sense numbers,
and/or function codes,
lexically processing said electronically encoded data to access said dictionary
data base with said apparatus,
providing a grammar specification in memory associated with said apparatus,
utilizing said natural language words which are associated with said electronically
encoded data and said associated data which are from said dictionary data base
entries with reference to said grammar specification to select word sense numbers
and/or function codes with said apparatus,
utilizing said index to said data base of natural language to access word sense
numbers and/or function codes with said associated additional data to match or
partially match said selected word sense numbers and/or function codes
associated with said natural language which is associated with said electronically
encoded data with said apparatus.
27

1
2
3
4
5
6
7
8
9
10

United States District Court
Northern District of California

11

’436 patent, A method of processing natural language in an apparatus, which comprises steps:
claim 1
providing natural language which is processed by said apparatus to provide
electronically encoded data which is representative of said natural language,
providing a dictionary data base in memory associated with said apparatus
wherein said dictionary data base contains a plurality of entries which are
comprised of one or more of syntax usage data, associated word sense numbers
with associated state representation data, and/or function codes,
lexically processing said electronically encoded data to access said dictionary
data base with said apparatus,
providing a natural language plausibility and expectedness processor in said
apparatus,
utilizing said natural language plausibility and expectedness processor to initiate
accessing entries of said dictionary data base which are associated with words of
said natural language.
The parties do not dispute that syntax means “the way in which words are put together to
form phrases, clauses, or sentences,” see Defs. Br. at 20 n.5; Reply Br. at 12-13, but disagree on

12
the exact form of the syntax usage data recited in the asserted claims. Defendants contend that
13
both the prosecution history and the specification support their construction. Defs. Br. at 19-22.
14

During prosecution, Budzinski explained how his claimed “syntax usage data” is different from

15
the data disclosed in Vanderwende:
16
17
18
19
20
21
22
23
24
25
26
27

The next paragraph of claim 7 of the present invention is: “providing
a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax
usage data, associated word sense numbers having associated state
representation data, and/or function codes.” The dictionary data base
entry of the present invention is described at page 49, line 33 to page
51, line 7, and illustrated in Fig. 3a. The syntax usage data is
comprised of syntax wordsets. A syntax wordset is a set of words
which can syntactically be used interchangeably in a natural
language construction. A word’s syntax wordset can be used by the
parser to determine the phrase the word belongs to, and the relation
of the word to other words in the phrase of incoming natural
language to an embodiment of the present invention. A word’s
syntax wordset can also be used to generate outgoing natural
language from an embodiment of the present invention. The syntax
data can be advantageously utilized to generate correct outgoing
natural language. For example, certain words have preferred
orderings, e.g. “sweet enough” and “sufficiently sweet,” but not
“enough sweet” and “sweet sufficiently.” Vanderwende utilizes
syntactic rules for parsing incoming text, but not for outgoing text,
and the Vanderwende knowledge base of semantic relations does
not contain syntax usage data.

28
28

1

’509 file at 29 (emphasis added). Defendants state that their construction simply combines

2

Budzinski’s definition of “syntax usage data” as “comprised of syntax wordsets” with his

3

definition of “syntax wordset.” Defs. Br. at 19. That definition of syntax wordset – i.e., “a set of

4

words which can syntactically be used interchangeably in a natural language construction” –

5

appears both in the prosecution history and also in the “Preferred Embodiment” section of the

6

specification. See ’468 patent at 28:20-22. Defendants also cite to a number of other portions of

7

the specification that refer to “wordsets” or “syntax wordsets.” See Defs. Br. at 20 (citing ’468

8

patent at 5:10-12; 28:04-07, 25-27; 32:33-46).

9

WTI counters that the specification supports its broader construction of “syntax usage
data,” and that the specification’s description of wordsets and syntax wordsets is a “classic

11
United States District Court
Northern District of California

10

example of a preferred embodiment.” Opening Br. at 19. With respect to the prosecution history,

12

WTI contends that Budzinski distinguished Vanderwende not on the ground that his invention

13

employs syntax usage data in the form of wordsets, but rather because Vanderwende does not

14

disclose syntax usage information at all. Id. at 19-20; see also Reply Br. at 12-13. WTI asserts

15

that a person of ordinary skill in the art would appreciate that Budzinski described the preferred

16

embodiment “merely as helpful background, not as a limitation of the claims.” Opening Br. at 20.

17

I agree with defendants’ construction. Budzinski’s description of the claimed “syntax

18

usage data” during prosecution is not reasonably characterized as mere “helpful background,” as

19

WTI contends. Id. Budzinski gave the description in the context of distinguishing Vanderwende

20

and used language strongly indicating that his purpose was to define the term, i.e., “[t]he syntax

21

usage data is comprised of . . .” and “[a] syntax wordset is a . . .” Further, WTI fails to identify

22

any portion of the specification that supports a construction of “syntax usage data” that does not

23

require wordsets. To the contrary, nearly every portion of the specification that WTI cites refers to

24

“wordsets” or “syntax wordsets.” See ’468 patent at 28:22-25 (“A word’s syntax wordset is used

25

by the parser to determine the parser to determine the phrase the word belongs to, and the relation

26

of the word to other words in the phrase.”) (emphasis added); id. at 5:07-19 (“The Dictionary 20

27

stores words, and idioms which each have associated syntax wordsets . . . Dictionary Look-Up

28

Step 14 looks up the syntax wordsets which each input word belongs to and passes this
29

1

information to the Parsing Step 16.”) (emphasis added). WTI does not explain how these portions

2

support its construction, except to state that the portions show that the asserted patents use

3

“syntax” in its “ordinary sense.” Opening Br. at 18. But how the asserted patents use the term

4

“syntax” is a different issue from how they use the term “syntax usage data.” With respect to the

5

latter issue, WTI identifies nothing in the asserted patents or elsewhere that provides meaningful

6

support for its position.
G.

7

“access conditions”
WTI’s Construction

8

Claims

9

’509 patent, Conditions that determine
claim 9
the nodes on a path that
are accessible.
’603 patent,
claim 14

10

United States District Court
Northern District of California

11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

’468 patent,
claim 12,
31

Defendants’
Court’s Construction
Construction
Values or data on paths in None.
a directed graph of
experience and
knowledge which are met
in order for purpose
relations common to
clause implying word
sense numbers to be
valid, and which
determine a feasible
purpose path when the
access conditions of
nodes on a path are met.

Claims
’468 patent, 12. A method of processing as defined in claim 1, which comprises steps
claim 12
providing an experience and knowledge data base wherein said experience and
knowledge data base is comprised of directed graphs comprised of nodes with
associated clause implying word sense numbers organized into paths of said
nodes such that said nodes have access conditions which determine the zero or
more next said nodes on the zero or more said paths that are accessible, said
nodes having optional related data,
purpose path identification processing to find zero or more said paths between
said nodes with reference to said experience and knowledge data base such that
said access conditions of said nodes on said found paths are met.
1. A method of processing natural language, which comprises steps
providing electronically encoded data which is representative of said natural
language,
providing a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax usage data,
associated word sense numbers having associated state representation data and/or
function codes,
lexically processing said electronically encoded data to access said dictionary
data base,
30

1
2
3
4
5
6
7
8

providing a grammar specification,
utilizing said syntax usage data which are from entries of said dictionary data
base and which are associated with words of said natural language with reference
to said grammar specification to produce output data representative of a
grammatical parse of said natural language, said output data including selected
syntax usage.
’468 patent, A method of processing experience and knowledge, which comprises steps
claim 31
providing said experience and knowledge data base which is comprised of
directed graphs comprised of nodes with associated clause implying word sense
numbers organized into paths of said nodes such that said nodes have access
conditions which determine the zero or more next said nodes on the zero or more
said paths that are accessible,

9
10

United States District Court
Northern District of California

11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

’509 patent
claim 9

purpose relation path identification processing to find zero or more said paths
between said nodes with reference to said experience and knowledge data base
such that said access conditions of said nodes on said found paths are met.
A method of processing natural language in an apparatus, which comprises steps
providing in memory associated with said apparatus an experience and
knowledge data base which is comprised of directed graphs comprised of nodes
with associated clause implying word sense numbers organized into paths of said
nodes such that said nodes have access conditions which determine zero or more
next said nodes on zero or more said paths that are accessible,
utilizing a natural language processor to provide natural language with associated
clause implying word sense numbers in memory associated with said apparatus,
purpose relation path identification processing with said apparatus to find zero or
more said paths from said nodes associated with said clause implying word sense
numbers associated with said natural language with reference to said experience
and knowledge data base such that said access conditions of said nodes on said
found paths are met,
providing criteria for selecting said found experience and knowledge paths in
memory associated with said apparatus,
utilizing said criteria to select one or more of said found paths with said
apparatus.

In support of their construction of “access conditions,” defendants rely on two statements
made by Budzinski during prosecution of the ’509 and ’603 patents. Defs. Br. at 22. During
prosecution of the ’509 patent, Budzinski stated that
[i]n the present invention, a purpose relation includes any subject or
action common to the related clause implying word sense numbers
in the purpose relation, and these clause implying word sense
numbers have associated nodes with access conditions on paths in a
directed graph of experience and knowledge. These access
conditions on a path are met for such purpose relations to be valid.
’509 file at 38-39 (emphasis added). Similarly, during prosecution of the ’603 patent, Budizski
31

1

described access conditions as follows:
The directed graph storing experience and knowledge has nodes
which have access conditions which determine a feasible purpose
path when the access conditions of nodes on a path are met.

2
3
4
5

’603 file at 26 (emphasis added). Defendants contend that their construction reflects these
descriptions. Defs. Br. at 22.
WTI agrees that access conditions “determine accessible or feasible paths in a directed

6
7
8
9
10

United States District Court
Northern District of California

11
12
13
14

graph,” Reply Br. at 13, and that “a directed graph is made up of nodes . . . and paths between
nodes representing relationships . . . between clauses,” Opening Br. at 21. WTI asserts not that
defendants’ construction is inaccurate, but that it is overcomplicated and redundant, in that it
“repeats terms such as ‘directed graph’ and ‘experience and knowledge’ which are already recited
in claims 12 and 31 [of the ’468 patent].” Id. In support of its construction, WTI points to the
statement in the abstract of ’087, ’091, ’840, ’509, and ’603 patents that “[n]odes in the directed
graph have access conditions which determine if a node is accessible on a path.” Opening Br. at
21. WTI asserts that its construction captures this meaning. Id.
I find that neither party’s construction is appropriate at this juncture. The parties do not

15
16
17
18
19
20
21
22
23
24

appear to dispute how the meaning of access conditions impacts the scope of the relevant asserted
claims; their only dispute appears to be how best to convey that meaning through claim
construction. That is, the parties appear to agree that access conditions determine the accessibility
of nodes on paths in directed graphs, thereby determining feasible paths between nodes. But this
is essentially how the claims themselves describe access conditions; each of the relevant asserted
claims discloses “directed graphs” comprised in part of “access conditions which determine [the]
zero or more next said nodes on [the] zero or more said paths that are accessible.”6 Given the
absence of a coherent dispute over claim scope, and the clarifying language already present in the
relevant asserted claims, I decline to adopt either party’s construction at this time.
H.

25
26
27
28

Claims

“dictionary data base”
WTI’s Construction

Defendants’

6

Court’s Construction

The relevant asserted claims from the ’468 patent include the bracketed “the”s; the relevant
asserted claims from the ’509 and ’603 patents omit them.
32

1
2
3
4
5
6
7
8
9
10

United States District Court
Northern District of California

11

’087 patent, None.
claim 1, 17
OR
’091 patent,
claim 1, 12 A collection of data
entities representing
’840 patent, words or word meanings.
claim 1, 5,
15, 16
’509 patent,
claim 16

Construction
A collection of data
entities representing
words or word meanings
which, at a minimum,
contains (1) the state
representation data
associated with a word
sense number, and (2) the
function selection process
of a function code for a
function word.7

WTI’s alternative
construction:
A collection of data
entities representing
words or word meanings.

’603 patent,
claim 9
’436 patent,
claim 1, 11
’468 patent,
claim 1, 33

12
13
14

’468 patent,
claim 1

15

providing a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax usage data,
associated word sense numbers having associated state representation data and/or
function codes,

16
17

lexically processing said electronically encoded data to access said dictionary
data base, providing a grammar specification,

18
19
20
21
22

Example Claims
A method of processing natural language, which comprises steps
providing electronically encoded data which is representative of said natural
language,

’087 patent,
claim 1

utilizing said syntax usage data which are from entries of said dictionary data
base and which are associated with words of said natural language with reference
to said grammar specification to produce output data representative of a
grammatical parse of said natural language, said output data including selected
syntax usage.
A method of processing natural language, which comprises steps
providing electronically encoded data which is representative of said natural
language,

23
24

providing a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax usage data,
associated word sense numbers having associated state representation data,

25
26
27
28

7

Defendants first proposed this construction at the hearing. Hearing Tr. at 40-44. After the
hearing, I granted WTI’s request to submit supplemental briefing regarding the construction, and
both parties submitted supplemental briefs. Dkt. Nos. 118, 119, 120.
33

lexically processing said electronically encoded data to access said dictionary
data base,

1
2
3
4
5
6
7
8
9
10

United States District Court
Northern District of California

11
12
13
14

’091 patent,
claim 1

utilizing said syntax usage data and said word sense numbers which are from
entries of said dictionary data base and which are associated with words of said
natural language to access said state representation data.
A method of processing natural language, which comprises steps
providing electronically encoded data which is representative of said natural
language,
providing a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax usage data,
associated word sense numbers having associated state representation data, and/or
function codes,
lexically processing said electronically encoded data to access said dictionary
data base, providing a grammar specification,
providing a data base of requirements such that said requirements must be met by
said associated state representation data of said word sense numbers for said word
sense numbers to be selected,
utilizing said syntax usage data which are from entries of said dictionary data
base and which are associated with words of said natural language with reference
to said grammar specification to select word sense numbers associated with said
natural language words such that said associated state representation data of said
associated word sense numbers meet said requirements for selecting said
associated word sense numbers.

15

In support of their construction of “dictionary data base,” defendants again rely on

16

prosecution statements by Budzinski. Defs. Br. at 22-24; see also Dkt. No. 120. They highlight

17

that Budzinski distinguished Vanderwende by stating that

18
19
20
21
22
23
24
25
26
27
28

Vandervende utilizes syntactic rules for parsing incoming text, but
not outgoing text, and the Vandervende knowledge base does not
contain syntax usage data of the present invention. The dictionary
data base of the present invention is utilized to locate the state
representation data associated with a word sense number. The
state representation data associated with word sense numbers
includes requirements and access conditions which are utilized for
selecting a word sense number which has the intended meaning of a
word contained in natural language. Vanderwende . . . does teach
how to construct a lexical data base of semantic relations of text
associated with text headwords with optionally appended dictionary
definition numbers. A semantic relation of Vanderwende is an
arrangement of text, and the Vanderwende lexical data base is
composed of organized units of text wherein these units are
semantic relations of arranged text. The semantic relations of the
Vanderwende knowledge base do not contain requirements and
access conditions. Thus, the Vanderwende knowledge base cannot
be utilized to determine valid semantic relations for natural
language, but rather the knowledge base can be used to find possible
semantic relations which may or may not be true for the current
natural language because even if the semantic relations are sense
34

1
2
3
4
5
6
7
8
9
10

United States District Court
Northern District of California

11
12
13

disambiguated in the data base, the semantic relations are only
possibly true. Thus, Vanderwende does not teach how to select valid
meanings for natural language. The dictionary data base of the
present invention is utilized to locate the function selection process
of a function code for a function word. The function selection
process is needed to find the intended meaning of natural language.
For example, the valid referent of a usage of ‘it’ in natural language
is needed to determine the intended meaning of the clause
containing ‘it,’ and Vanderwende does not teach how to select valid
meanings for natural language . . . Vanderwende does not teach how
to perform function word processing.
’509 file at 33-34 (emphasis added). Defendants contend that their construction “reflects verbatim
these two distinguishing features of the dictionary data base,” i.e., that it is “utilized to locate”
both (1) “the state representation data associated with a word sense number,” and (2) “the function
selection process of a function code for a function word.” Defs. Br. at 23.
According to WTI, the term “dictionary data base” is “well known in the art and does not
require construction.” Opening Br. at 22. Alternatively, it contends that a “simple elucidation of
the ordinary meaning is sufficient,” which is what it contends its construction provides. Id. WTI
criticizes defendants’ construction on the ground that the prosecution statements highlighted by

14
defendants are aimed at how the entries of a dictionary data base are used, not what a dictionary
15
16

data base is. Reply Br. at 14-15. WTI also argues that defendants’ construction would effectively
rewrite the claims from requiring a dictionary data based comprised of “syntax usage data,

17
associated word sense numbers having associated state representation data, and/or function
18

codes,” to requiring a dictionary database comprised of both associated state representation data

19
and function codes. Dkt. No. 119 at 2.
20

I agree with WTI’s criticisms of defendants’ construction and adopt WTI’s alternative

21
construction of dictionary data base. As WTI points out, the prosecution statements highlighted
22
by defendants appear to be directed at how the entries of a dictionary data base are used,
23

not what a dictionary data base is, and defendants’ reading of the statements is at odds with the

24
actual language of the claims. Apart from those prosecution statements, defendants offer no basis
25
for their construction, and identify nothing in the claims or specification that supports it. See Defs.
26

Br. at 22-24; Dkt. No. 120. Absent more support for defendants’ construction, I adopt WTI’s.

27
28
35

I.

1

“context data base”
WTI’s Construction

2

Claims

3

’840 patent, A collection of data
claims 1, 2, entities containing
3, 5, 15, 16 context information.

4
5
6
7
8

’509 patent,
claims 10,
16
’436 patent,
claims 7,
14

Defendants’
Construction
A database separate from
the other knowledge
bases that stores semantic
relations of text which
have recently been
extracted from a
conversation or other
natural language.

Court’s Construction
Defendants’ construction,
slightly modified:
A database separate from
the other knowledge
bases that stores semantic
relations of text which
have recently been
extracted from a
conversation or other
natural language.

9
10

United States District Court
Northern District of California

11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27

’509 patent,
claim 10

Example Claims
10. A method of processing as defined in claim 9, which comprises steps
providing a context data base wherein said context data base contains a plurality
of entries which are comprised of one or more of clause implying word sense
numbers having associated state representation data including associated
experience and knowledge paths,
selecting experience and knowledge paths associated with said clause implying
word sense numbers associated with said natural language such that said
experience and knowledge paths associated with said clause implying word sense
numbers associated with said natural language have accessable paths to said
experience and knowledge paths associated with said clause implying word sense
numbers from said context data base entries.
9. A method of processing natural language in an apparatus, which comprises
steps
providing in memory associated with said apparatus an experience and
knowledge data base which is comprised of directed graphs comprised of nodes
with associated clause implying word sense numbers organized into paths of said
nodes such that said nodes have access conditions which determine zero or more
next said nodes on zero or more said paths that are accessible,
utilizing a natural language processor to provide natural language with associated
clause implying word sense numbers in memory associated with said apparatus,
purpose relation path identification processing with said apparatus to find zero or
more said paths from said nodes associated with said clause implying word sense
numbers associated with said natural language with reference to said experience
and knowledge data base such that said access conditions of said nodes on said
found paths are met,
providing criteria for selecting said found experience and knowledge paths in
memory associated with said apparatus,

28
36

1
2

’840 patent,
claim 15

utilizing said criteria to select one or more of said found paths with said
apparatus.
A method of processing natural language, which comprises steps
providing electronically encoded data which are representative of said natural
language,

3
4

providing a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax usage data,
associated clause implying word sense numbers having associated state
representation data,

5
6
7

lexically processing said electronically encoded data to access said dictionary
data base,

8

providing a grammar specification,

9

providing a context data base wherein said context data base contains a plurality
of entries which are comprised of one or more of clause implying word sense
numbers having associated state representation data,

10

United States District Court
Northern District of California

11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

’840 patent,
claim 16

utilizing one or more of said syntax usage data and said clause implying word
sense numbers which are from entries of said dictionary data base and which are
associated with words of said natural language with reference to said grammar
specification and with reference to said context data base to select clause
implying word sense numbers associated with said natural language words.
A method of processing as defined in claim 15, which comprises steps
providing an experience and knowledge data base which is comprised of directed
graphs comprised of nodes having associated clause implying word sense
numbers organized into paths of said nodes wherein said paths have associated
identifiers,
providing a context data base wherein said context data base contains a plurality
of entries which are comprised of one or more of clause implying word sense
numbers having associated state representation data including associated
experience and knowledge data base path identifiers,
selecting experience and knowledge data base paths for said clause implying
word sense numbers which are from said dictionary data base and which are
associated with said natural language words such that said experience and
knowledge data base path identifiers match or partially match said experience and
knowledge data base path identifiers which are associated with said clause
implying word sense numbers from said context data base entries and said clause
implying word sense numbers having said associated natural language words
match or partially match one or more of said clause implying word sense numbers
which are on said experience and knowledge data base paths having said
experience and knowledge data base path identifiers which match or partially
match said experience and knowledge data base path identifiers which are
associated with said clause implying word sense numbers from said context data
base entries.

Defendants again rely on Budzinki’s statements during prosecution to support their
construction. Defs. Br. at 24-25. They point to the same introductory remark discussed above
37

1

(i.e., “the terms of the present invention have significantly different descriptions in the

2

specification”) as well as the following portions of the ’509 file:

3
4
5
6
7
8
9
10

United States District Court
Northern District of California

11
12
13
14
15
16
17

The context data base of the present invention is stored in Context
Memory 120. Context Memory 120 is updated after each clause
implying word sense has been selected including function word
processing. The state representation of the clause is stored including
the state representation of nouns, the word sense numbers of clauses,
and the purpose relations of the current natural language clause to
the other clauses in the conversation or other natural language
including the purpose paths.
[...]
Context Memory 120 contains information from a conversation or
other natural language and is separate from the other knowledge
data bases of the present invention including the state
representation memories and the experience and knowledge
memory. Context Memory 120 contains pointers to existing state
representations, and experience and knowledge in the corresponding
memory. New state representations and new experience and
knowledge are also stored in Context Memory 120. Also, the stated
and implied relations among words in the clause are stored in
Context Memory 120. Having a context data base for natural
language not utilized to build the knowledge bases of the present
invention is advantageous because the state representation, including
word sense numbers and purpose relations of clause implying word
senses, of the current natural language is often related to previous
natural language of the conversation or other natural language.
[...]

24

Vanderwende uses “relational context” which humans use to
understand natural language . . . , and Vanderwende utilizes
“contexts” of definitions . . . which are definitions which contain a
particular word like “flower” . . . and which are used to find
semantic relations of the particular word. In contrast to the present
invention, Vanderwende only has semantic relations stored in the
Vanderwende knowledge base, and this knowledge base is the
result of all text which has been processed to extract semantic
relations. Vanderwende does not teach how to build and utilize a
separate knowledge base to store semantic relations of text which
have recently been extracted. Vanderwende does not teach how to
build a context data base of purpose relations of the present
invention between clause implying word sense numbers which are
valid.

25

[...]

26

Also, Vanderwende does not teach how to build a context data base
which includes word sense numbers having associated state
representation data, and/or function codes. A context data base
separate from other data bases can be utilized to generate outgoing
natural language which contains the words and phrases expressed

18
19
20
21
22
23

27
28

38

in the conversation or other natural language. Vanderwende does
not teach how to create such a context data base.

1
2

’509 file at 22-23, 37. Defendants contend that their construction of context data base – which

3

requires both that the context data base be “separate,” and that it store information “recently

4

extracted from a conversation or other natural language” –“capture[s] the meaning [Budzinski]

5

explained when acting as his own lexicographer” during prosecution. Defs. Br. at 24-25.
WTI appears to dispute only the “separate” aspect of defendants’ construction. See

6
7

Opening Br. at 23-24; Reply Br. at 15 (“The essential dispute is whether the context database

8

must be separate from the dictionary data base of the claims.”). According to WTI, the key

9

distinction between Vanderwende and the claimed invention is not that context information in the
claimed invention is “separately stored,” but that the “so-called context information [in

11
United States District Court
Northern District of California

10

Vanderwende] is not really contextual at all.” Opening Br. at 24. That is, “instead of having

12

‘recently been extracted’ from related expressions, [the context information in Vanderwende] is

13

‘the result of all text which has been processed to extract semantic relations.’” Id. (quoting ’509

14

file at 23). WTI contends that, “[t]aken as a whole, [Budzinki’s] remarks would not lead a person

15

of ordinary skill to conclude the context information must necessarily be stored separately.”

16

Reply Br. at 15.
I agree with WTI that the prosecution history does not support a construction of “context

17
18

data base” that requires a separate structure for this element. It is not all clear from the

19

prosecution history that Budzinski was using the word “separate” to mean separate in a structural

20

sense, as defendants’ construction appears to require. The rest of defendants’ construction,

21

however, is both supported by the prosecution history and not meaningfully disputed by WTI,

22

and I adopt it as the Court’s.

23

III.

24
25
26
27
28

DISPUTED TERMS INVOLVING 35 U.S.C. § 112 ¶ 6
The parties dispute whether the final paragraph of claim 1 of the ’468 patent is a step-

plus-function limitation that is governed by, and indefinite under, 35 U.S.C. § 112 ¶ 6. That
claim states in whole, with emphasis added to the disputed paragraph:
’468 patent,
claim 1

A method of processing natural language, which comprises steps
providing electronically encoded data which is representative of said natural
language,
39

1
2
3

providing a dictionary data base wherein said dictionary data base contains a
plurality of entries which are comprised of one or more of syntax usage data,
associated word sense numbers having associated state representation data and/or
function codes,

4

lexically processing said electronically encoded data to access said dictionary
data base,

5

providing a grammar specification,

6

utilizing said syntax usage data which are from entries of said dictionary data
base and which are associated with words of said natural language with reference
to said grammar specification to produce output data representative of a
grammatical parse of said natural language, said output data including selected
syntax usage.

7
8

Section 112 ¶ 6 provides that “[a]n element in a claim for a combination may be expressed

10

as a means or step for performing a specified function without the recital of structure, material, or

11
United States District Court
Northern District of California

9

acts in support thereof, and such claim shall be construed to cover the corresponding structure,

12

material, or acts described in the specification and equivalents thereof.” 35 U.S.C. § 112 ¶ 6. In

13

other words, the provision applies where a claim recites “a means or step for performing a

14

function,” but no corresponding “structure, material, or acts.” Id. The terms “structure” and

15

“material” are generally associated with “means,” while the term “acts” is generally associated

16

with “step for.” Seal-Flex, Inc. v. Athletic Track & Court Const., 172 F.3d 836, 848 (Fed. Cir.

17

1999). Thus, “a claim element deserves step-plus-function treatment when expressed as a step for

18

performing a specified function without the recital of acts in support thereof.” Id. (internal

19

quotation marks and emphasis omitted).

20

Where a patentee uses “step for” in a claim limitation, there is a rebuttable presumption

21

that the limitation is a step-plus-function limitation governed by section 112 ¶ 6. See Masco Corp.

22

v. United States, 303 F.3d 1316, 1326 (Fed. Cir. 2002); Seal-Flex, 172 F.3d at 849. Conversely,

23

where the patentee does not use “step for,” the limitation may only be construed as a step-plus-

24

function limitation upon a “showing that the limitation contains nothing that can be construed as

25

an act.” Masco, 303 F.3d at 1327. Whether language in a method claim represents an act or a

26

function can be hard to decipher; “method claim elements often recite phrases susceptible to

27

interpretation as either a function or as an act for performing a function,” and “[b]oth acts and

28

functions are often stated using verbs ending in ‘ing.’” Seal-Flex, 172 F.3d at 849. In general,
40

1

however, the function of a method claim limitation “corresponds to what that element ultimately

2

accomplishes in relationship to what the other elements of the claim and the claim as a whole

3

accomplish,” whereas “acts correspond to how the function is accomplished.” Id. at 849-50

4

(emphasis in original).

5

Here, the “utilizing said syntax usage data” limitation in claim 1 of the ’468 patent does

6

not use “step for.” Accordingly, the burden is on defendants to show that “the limitation contains

7

nothing that can be construed as an act.” Masco, 303 F.3d at 1327.

8
9

Defendants have not made this showing. They contend that the latter part of the limitation
describes a function – i.e., “to produce output data representative of a grammatical parse of said
natural language, said output data including selected syntax usage” – but that there is no recitation

11
United States District Court
Northern District of California

10

of acts to accomplish this function. Defendants assert that the first part of the limitation – i.e.,

12

“utilizing said syntax usage data which are from entries of said dictionary data base and which are

13

associated with words of said natural language with reference to said grammar specification” – “is

14

not an act, and to extent that it could be considered one, it is not sufficient to show how the

15

function is accomplished.” Defs. Br. at 16. According to defendants, the first part of the

16

limitation is “not sufficient” to serve as an act because “[t]here is no statement of how to use the

17

syntax usage data or how to reference the grammar specification.” Id.

18

The problem with this argument is that defendants do not explain how the first part of the

19

limitation could be reasonably construed as anything but an act, and it is not clear to me how it

20

could be. While “[b]oth acts and functions are often stated using verbs ending in ‘ing,’” Seal-

21

Flex, 172 F.3d at 849, the phrase “utilizing said syntax usage data which are from entries of said

22

dictionary data base and which are associated with words of said natural language with reference

23

to said grammar specification” reads to me like an act. That is, the phrase, explains “how [a]

24

function is accomplished,” id. at 849-50, the function in this case being “produc[ing] output data

25

representative of a grammatical parse of said natural language, said output data including selected

26

syntax usage.” Defendants make no argument to the contrary; they simply assert without

27

explanation that the limitation “is not an act.” See Defs. Br. at 16.

28
41

1

Defendants’ assertion that the first part of the limitation lacks sufficient detail to serve as

2

an act is also unconvincing. See id. Defendants cite no authority indicating that this sort of detail

3

is required. In the one case cited by either party on this issue, Neurografix v. Regents of Univ. of

4

California, No. 11-cv-07591-MRP, 2012 WL 8281409 (C.D. Cal. June 13, 2012), the accused

5

infringer argued that that the term “processing” could not serve as an act “because the mere word

6

‘processing’ provides no information regarding how to accomplish the claimed function.” Id. at

7

*6. The district court disagreed, explaining that

8
9
10

United States District Court
Northern District of California

11
12
13
14
15
16
17

[t]here is a subtle but critical difference between how a function is
accomplished and how an act accomplishes a function. The former
is an identify-the-act question, appropriate for deciding if [section
112] ¶ 6 applies in the first instance. The latter is an analyze-the-act
question, appropriate for determining if a claim element is valid
under the enablement, written description, and definiteness inquiries
under [section] 112 ¶¶ 1-2 . . . Identifying an act to see if [section
112] ¶ 6 applies (step-plus-function identification) is an inherently
less searching inquiry than analyzing the same act under [section
112] ¶¶ 1-2 (enablement, written description, definiteness) . . . . The
Federal Circuit has not, to this Court's knowledge, analyzed an act
past the point of identification to determine [section 112] ¶ 6
applicability.
Id. at *6-7. In the absence of any authority to the contrary, I find Nuerografix persuasive and
follow it here. Defendants’ attack on the clarity of the act recited in the first part of the “utilizing
said syntax usage data” limitation does not justify applying section 112 ¶ 6.
CONCLUSION

18
19
20
21

For the foregoing reasons, the claim terms are construed as stated above.
IT IS SO ORDERED.
Dated: July 12, 2016
_____________________________________
WILLIAM H. ORRICK
United States District Judge

22
23
24
25
26
27
28
42