I/P Engine, Inc. v. AOL, Inc. et al
Filing
427
Opposition to 237 MOTION for Summary Judgment Defendants AOL Inc., Google Inc., IAC Search & Media, Inc., Gannett Company, Inc., and Target Corporation's Motion for Summary Judgment filed by I/P Engine, Inc.. (Attachments: # 1 Exhibit 1, # 2 Exhibit 2-8 (Filed Under Seal), # 3 Exhibit 9, # 4 Exhibit 10, # 5 Exhibit 11-19 (Filed Under Seal), # 6 Exhibit 20, # 7 Exhibit 21, # 8 Exhibit 22-24 (Filed Under Seal), # 9 Exhibit 25, # 10 Exhibit 26, # 11 Exhibit 27-32 (Filed Under Seal), # 12 Exhibit 33, # 13 Exhibit 34-39 (Filed Under Seal), # 14 Exhibit 40, # 15 Exhibit 41, # 16 Exhibit 42, # 17 Exhibit 43, # 18 Exhibit 44, # 19 Exhibit 45 (Filed Under Seal), # 20 Exhibit 46, # 21 Exhibit 47, # 22 Exhibit 48, # 23 Exhibit 49, # 24 Exhibit 50, # 25 Exhibit 51, # 26 Exhibit 52, # 27 Exhibit 53-54 (Filed Under Seal), # 28 Exhibit 55, # 29 Exhibit 56 (Filed Under Seal))(Sherwood, Jeffrey)
Exhibit 52
IN THE UNITED STATES DISTRICT COURT
FOR THE EASTERN DISTRICT OF VIRGINIA
I/P ENGINE, INC.,
)
)
Plaintiff,
)
)
v.
)
)
AOL, INC., GOOGLE INC., IAC SEARCH & )
MEDIA, INC., TARGET CORP., and
)
GANNETT CO., INC.
)
)
Defendants.
C.A. No. 2:11-cv-512-RAJ
JURY TRIAL DEMANDED
REPORT OF DEFENDANTS’ EXPERT
LYLE H. UNGAR, PH.D., CONCERNING INVALIDITY OF
CLAIMS 10, 14, 15, 25, 27, AND 28
OF U.S. PATENT NO. 6,314,420
AND
CLAIMS 1, 5, 6, 21, 22, 26, 28, AND 38
OF U.S. PATENT NO. 6,775,664
01980.51928/4874260.1
UNGAR EXPERT REPORT ON INVALIDITY
CASE 2:11-cv-512-RAJ
66.
Culliss discloses that the articles are presented to the user in the order dictated by
their combined key term scores. (Id. at 5:2-10). For example, if Article 1 had a key term score
of 5 for “Paris,” 3 for “museum,” and 2 for “vacations,” its aggregate score for the query “Paris
museum vacations” would be 10 (5 + 3 +2). If Article 2 had a key term score of 4 for “Paris,” 2
for museum,” and 3 for “vacations,” its aggregate score for the query “Paris museum vacations”
would be 9 (4 + 2 +3). Thus, Article 1 would be presented above Article 2 because it had a
higher aggregate score.
67.
When a user selects an article whose squib is presented to him, the key term
scores for that article which correspond to the terms in the user’s query are increased. (Id. at
4:37-49). This is because the user, by selecting the article in response to his query, has implicitly
endorsed the idea that these key terms from the query are appropriately matched to the article.
(See id.)
68.
For example, if our hypothetical first user who queried “Paris museum vacations”
selected Article 2, then Article 2’s key term scores for “Paris,” “museum,” and “vacations”
might each rise by +1. (See id. at 4:43-45 (“To alter the key term scores, a positive score such as
(+1) can be added to the key term scores, for example . . .”) The next user who enters the same
query would thus see a different rank of articles, based on the new key term scores that reflect
the input of the prior user. (See id. at 4:66-5:1). Sticking with our example, Article 2 would
have a new aggregate score of 12 (instead of 9) after the first user selected it, because its key
term scores for “Paris,” “museum,” and “vacations” each increased by +1 when the first user
selected it. Thus, a later user who queries “Paris museum vacations” would see Article 2 (which
has a new aggregate score of 12) presented above Article 1 (which still has its old aggregate
score of 10).
01980.51928/4874260.1
UNGAR EXPERT REPORT ON INVALIDITY
25
CASE 2:11-cv-512-RAJ
item’s content, feature extraction would be a predictable element of a content-based filter
system.
246.
Delivering filtered information to users: As noted above, even the most
sophisticated information-filtering system has little utility if the filtered information cannot be
somehow delivered to the users who have a need for it. Accordingly, it would be predictable for
any information-filtering method to include the element of delivering filtered information to
users.
247.
Filtering advertisements: There are no technical or conceptual difficulties in
filtering advertisements as opposed to filtering other types of digital media. Thus, it would be
predictable for an information filtering system to filter advertisements. Indeed, this would allow
such a system to be used in the lucrative market for computerized advertising services – a market
that was well-established by the asserted patents’ priority date of December 1998. Accordingly,
it is unsurprising that many of the prior art references discussed in this Report specifically
disclose filtering advertisements. (See Culliss at 9:56-62; Bowman at 5:4, 9:2-3; Ryan at 4:5767).
D.
Claims 10, 14, 15, 25, 27, and 28 of the ‘420 Patent and claim 5 of the ‘664
Patent are obvious over Rose in view of Bowman
1.
248.
Claim 10 of the ‘420 Patent is obvious over Rose in view of Bowman
As noted above, I/P Engine’s infringement contentions assert that the preamble of
claim 10 is not a limitation. To the extent that the preamble is considered a limitation here, Rose
recites a “search engine system,” as recited by the preamble to claim 10. Specifically, Rose
discloses filtering “search results obtained through an online text retrieval service.” (Id. at 2:5455) (emphasis added). Similarly, Rose discloses that its information access system may
comprise “an electronic search and retrieval system.” (Claim 26).
01980.51928/4874260.1
UNGAR EXPERT REPORT ON INVALIDITY
91
CASE 2:11-cv-512-RAJ
(a)
249.
a system for scanning a network to make a demand search for
informons relevant to a query from an individual user
As noted above, I/P Engine takes the position that ‘420 claim 10(a) is satisfied if a
system conducts a search for information in response to a user query. Rose meets this element,
because Rose’s system accepts a search query from a user and returns a set of search results in
response. (See id. at 2:54-55; claim 26).
(b)
250.
a content-based filter system for receiving the informons from
the scanning system and for filtering the informons on the
basis of applicable content profile data for relevance to the
query
Rose discloses a content-based filter system that filters informons on the basis of
applicable content profile data. Specifically, the search results in Rose’s system are filtered by
comparing a vector representing a document’s content to a vector representing the user’s
preferences. (Id. at 6:11-58). The closer the vectors are to each other, the more relevant the
document is judged to be for the user. (Id. at 6:56-58).
251.
While Rose’s comparison of document vector to user vector filters for relevance
to the user rather than relevance to the query, it would be obvious modify Rose so that Rose
filtered for relevance to the query. As Rose discloses, any sort of information content can be
described by a vector. (Id. at 6:26-35). Thus, while Rose compares a search result vector against
a vector representing the user profile, Rose’s method could be just as easily used to compare a
search result vector against a vector representing the user’s query. Moreover, one of skill the art
would be motivated to modify Rose in this manner. As discussed above, there are a limited
number of ways to filter content-based information: chiefly, one could filter for relevance to a
user or to a user’s query. Furthermore, other references such as Bowman already teach the utility
of filtering search results for relevance to a query. See Section VI.A, supra (explaining how
01980.51928/4874260.1
UNGAR EXPERT REPORT ON INVALIDITY
92
CASE 2:11-cv-512-RAJ
Bowman combines content-based and feedback-based methods to filter search results for
relevance to a query).
252.
One of skill in the art would be motivated to combine Bowman with Rose, such
that Rose’s system would filter information for relevance to the query. Rose and Bowman are
both directed to the problem of efficiently filtering large amounts of information. Moreover,
they both solve this problem in similar ways – namely, by combining content data with feedback
data to create a hybrid content/feedback filtering method. Because Rose and Bowman propose
similar approaches to solving the same problem, one of skill in the art would be strongly
motivated to combine their teachings.
(c)
253.
a feedback system for receiving collaborative feedback data
from system users relative to informons considered by such
users
Rose discloses a feedback system for receiving collaborative feedback data from
system users relative to informons considered by such users. Specifically, Rose receives
feedback from system users about how highly they rated search results presented to them. (See
Rose at 5:31-46). For each user, Rose uses the feedback from users most similar to that user to
determine how relevant a given search result will be for that user. (See id. at 6:59-7:10). In
other words, Rose judges that a search result will be deemed more relevant to a given user if
other users who are similar to that user had rated that search result highly.
(d)
254.
The filter system combining pertaining feedback data from the
feedback system with the content profile data in filtering each
informon for relevance to the query
Rose combines the feedback data with the content profile data to filter each
document for relevance. See Rose at Abstract (“Items of information to be presented to a user
are ranked according to their likely degree of relevance to that user and displayed in order of
ranking. The prediction of relevance is carried out by combining data pertaining to the content
01980.51928/4874260.1
UNGAR EXPERT REPORT ON INVALIDITY
93
CASE 2:11-cv-512-RAJ
of each item of information with other data regarding correlations of interests between users.”)
(emphasis added); see also id. at 7:35-50.
255.
While Rose uses this hybrid filtering method to filter documents for relevance to
the user, it would be obvious to modify Rose so that it filtered for relevance to the query. This
could be done simply by comparing each document vector to a query vector instead of to a user
vector, and by recording feedback from the subset of other users who had entered the same
search query (instead of recording feedback from all users). There would be no technical
difficulties to modifying Rose in this manner. Moreover, one of skill the art would be motivated
to modify Rose in this manner, given the limited number of ways to do relevance filtering and
the fact that other references such as Bowman already teach the utility of filtering search results
for relevance to a query. See also Section VII.D.1(b), supra (explaining why one of skill in the
art would be motivated to combine Rose’s and Bowman’s teachings).
2.
256.
Claims 14 and 15 of the ‘420 Patent are obvious over Rose in view of
Bowman
Claim 14 depends from claim 10 and further requires “wherein the collaborative
feedback data comprises passive feedback data.” Claim 15 depends from claim 14 and further
requires “wherein the passive feedback data is obtained by passively monitoring the actual
response to a proposed informon.” Rose discloses a feedback system where users actively state
their level of interest in each document that they view. (See Rose at 5:8-30). However, Bowman
teaches passive feedback. Bowman’s feedback data is derived from passively monitoring users’
actual response to search results – namely, monitoring how frequently users who had entered the
same query selected each of those search results. (See Bowman at 2:31-35).
257.
It would have been obvious to modify Rose so that it utilized passive feedback in
the manner that Bowman does. There are only two basic types of user feedback that can be
collected – active feedback and passive feedback. Moreover, one of skill in the art would
01980.51928/4874260.1
UNGAR EXPERT REPORT ON INVALIDITY
94
CASE 2:11-cv-512-RAJ
Disclaimer: Justia Dockets & Filings provides public litigation records from the federal appellate and district courts. These filings and docket sheets should not be considered findings of fact or liability, nor do they necessarily reflect the view of Justia.
Why Is My Information Online?