IPR2017-01039 Smyth Declaration U.S. Patent 7,231,379 ......informed that Guada Technologies LLC (“Guada”) is the purported owner of the ’379 Patent, and I note that I have no

IPR2017-01039 Smyth Declaration U.S. Patent 7,231,379

UNITED STATES PATENT AND TRADEMARK OFFICE ____________

BEFORE THE PATENT TRIAL AND APPEAL BOARD ____________

UNIFIED PATENTS INC. Petitioner

v.

GUADA TECHNOLOGIES LLC Patent Owner

____________

IPR2017-01039 Patent 7,231,379 ____________

DECLARATION OF PADHRAIC SMYTH, PH.D.

IPR2017-01039 Unified EX1007 Page 1


Unified Exhibit 1007 2

I, Padhraic Smyth, hereby declare the following:

I. BACKGROUND AND QUALIFICATIONS

1. My name is Padhraic Smyth and I am over 21 years of age and

otherwise competent to make this Declaration. I make this Declaration based on

facts and matters within my own knowledge and on information provided to me by

others, and, if called as a witness, I could and would competently testify to the

matters set forth herein.

2. I am a Professor in the Department of Computer Science at the

University of California, Irvine and Director of the UCI Data Science Initiative. I

have been retained as a technical expert witness in this matter by Counsel for

Petitioner Unified Patents, Inc. to provide my independent opinions on certain

issues requested by Counsel for Petitioner relating to the accompanying petition for

Inter Partes Review of U.S. Patent No. 7,231,379 (“the ’379 Patent”). My

compensation in this matter is not based on the substance of my opinions or the

outcome of this matter. I have no financial interest in Petitioner. I have been

informed that Guada Technologies LLC (“Guada”) is the purported owner of the

’379 Patent, and I note that I have no financial interest in Guada.

3. I have summarized in this section my educational background, career

history, and other qualifications relevant to this matter. I have also included a

current version of my curriculum vitae as Appendix A (Ex. 1009).




4. I received a bachelor’s degree in electronic engineering (B.E., first

class honors) from the National University of Ireland, Galway, in 1984. I received

a master’s degree (M.S.E.E.) and a Ph.D. in electrical engineering from the

California Institute of Technology, Pasadena, CA, in 1985 and 1988, respectively.

My Ph.D. thesis was focused on the use of hierarchical tree structures and rule-

based methods for automated and efficient classification of objects into categories.

5. From 1988 to 1996, I was a technical staff member and technical group

leader (from 1992 onwards) at the Jet Propulsion Laboratory (JPL) in Pasadena,

CA. My role at JPL consisted of research and development in the areas of pattern

recognition, machine learning, data mining, and expert systems, as well as leading

projects involved in the application of these techniques to problems of interest to

JPL and NASA.

6. As part of my work, I published and presented papers during the period

1988-1996 at multiple different conferences in the areas of pattern recognition,

machine learning, and artificial intelligence. One example of my research work

was my involvement in the emerging research area of “knowledge discovery in

databases” (KDD). This began as a small research workshop in 1989 and quickly

evolved into a large annual international conference (with the first conference in

2004 and continuing annually since then). The research area was somewhat unique

in that it involved an interdisciplinary set of researchers working at the intersection




of databases, statistics, and machine learning algorithms. I was involved with the

KDD research field both as a researcher (writing and presenting papers), in the

organization of the conference, and in co-editing the first text on knowledge

discovery from databases (published by MIT Press in 1996, see discussion of

publications below).

7. In 1996, I moved from JPL to the University of California, Irvine, to

take a position as an assistant professor in the Department of Computer Science.

In 1998, I was promoted to associate professor with tenure, and in 2003 I was

promoted to the position of full professor. I also have a joint faculty appointment in

the Department of Statistics at UC Irvine. As a professor at UC Irvine since 1996, I

have conducted research in the areas of pattern recognition, machine learning, and

artificial intelligence, with an emphasis on developing new theories and algorithms

for automatically extracting useful information from very large volumes of data

across a wide variety of applications.

8. In 2007, I was also appointed as the founding director for the Center

for Machine Learning and Intelligent Systems at UC Irvine. This Center has over

30 affiliated faculty members at UC Irvine whom are all involved in research in

areas such as machine learning, database research, and artificial intelligence. In

2014, I was appointed as founding Director of the UC Irvine Data Science

Initiative, a cross-campus research initiative involving computer scientists,




statisticians, engineers, scientists, medical researchers, and more across the

campus.

9. My teaching duties have consisted of teaching both undergraduate and

graduate courses in the Computer Science department, with a focus on courses in

the areas of data mining and machine learning – titles of courses I have taught in

the past few years include Data Mining, Introduction to Artificial Intelligence,

Project in Artificial Intelligence, Applications of Probability for Computer

Scientists, and Probabilistic Learning. These courses include material related to

data structures such as hierarchical trees for automated decision-making and user

navigation, design and evaluation of systems for information retrieval, and

machine learning algorithms that can adapt and learn from data provided via user

input.

10. In addition to my duties at UC Irvine, I also consult with private

industry in the areas of machine learning and pattern recognition. My consulting

work often involves the development of mathematical models, algorithms, and

software for companies who wish to develop and deploy operational systems that

can autonomously interact with a user, such as recommending (on a Web site) the

next item to a user from a large catalog of potential items they may wish to

consider. These systems are typically constructed from large historical databases,

consisting of text data, customer transactions, etc. Over the past 18 years I have




consulted in this manner with companies such as AT&T, Samsung, Nokia, First

Quadrant, SmithKline Beecham, Yahoo!, eBay, and Netflix, as well as with a

number of smaller startup companies. My involvement in consulting projects has

given me the opportunity to develop expertise in the practical application of

machine learning and information retrieval algorithms, and in particular, to

develop an understanding of how these algorithms are deployed within real-world,

large-scale software systems that allow users to interact with databases and

websites.

11. As part of my real-world consulting projects over the past 18 years (at

Samsung, Yahoo!, eBay, and others), I have had direct experience with the

development of systems that use automated algorithms to assist a user who is

interacting with a system and has a specific goal in mind, such as finding a specific

item of information. This includes experience with methods and algorithms that

can extract useful information from large databases of user clickstream and search

query data, as well as from text documents related to customer reviews, customer

emails, product descriptions, Web page content, and search query data. As part of

this work I developed and adapted a variety of mathematical and statistical

frameworks that allow a computer to automatically decide what item of

information to show to a human user who is interacting with a system and who has

a specific goal in mind such as making a purchase or finding information on a Web




page. I developed computer algorithms to apply these mathematical and statistical

frameworks efficiently both to (a) user data (such as log files of clickstreams

recording how users interact with a system) and (b) to text data (such as text from

Web pages and documents). I also wrote and tested software code to implement

those algorithms in software, and I ran and interpreted computational experiments

to evaluate the effectiveness of different approaches.

12. I have published 67 journal papers, 18 book chapters, 9 technical

magazine articles, and 109 peer-reviewed conference papers related to my

research. Several of these publications are among the most highly-cited papers in

the general areas of data mining and artificial intelligence – my papers have

approximately 40,000 citations in total according to Google Scholar. Four of my

conference papers received best paper awards at the Association for Computing

Machinery (ACM) Conference on Knowledge Discovery and Data Mining, the

leading annual international conference on data mining.

13. I co-edited the book Advances in Knowledge Discovery and Data

Mining (AAAI/MIT Press, 1996), which is considered the first book published on

the topic of automated extraction of information from large databases and has over

5000 citations according to Google Scholar. I also co-authored Principles of Data

Mining (MIT Press, 2001), which is widely used as a graduate textbook in data

mining courses, has over 4700 citations, and has been translated into Chinese and




Polish editions. This text contains material describing the general principles of (a)

hierarchical tree structures (chapter 9 on tree-structured clustering, and chapter 10

on tree-structured classification) and (b) information retrieval and text analysis

(chapter 14). I also co-authored the text Modeling the Internet and the Web:

Probabilistic Methods and Algorithms (Wiley, 2003). This text contains material

describing the general principles of text analysis and information retrieval (Chapter

4), tree and graph-based models for representing text information (Chapter 5), and

analysis of how human users interact with information retrieval systems (Chapter

7).

14. I have been elected as a Fellow of the Association of Computing

Machinery (ACM), in 2013, and also elected a Fellow of the Association for

Advancement of Artificial Intelligence (AAAI), in 2010. I received the Innovation

Award from ACM SIGKDD (Special Interest Group in Knowledge Discovery and

Data Mining) in 2009. This award is given annually by ACM to an individual who

has made significant contributions to the research and practice of data mining. The

award is the most prestigious annual award given to a researcher in the data mining

research field. I also received a National Science Foundation CAREER award, a

Google Faculty Research Award, an IBM Faculty Partnership Award, and the Lew

Allen Award for Research at JPL.




15. I have served as an associate editor for the Journal of the American

Statistical Association, the IEEE Transactions on Knowledge and Data

Engineering, and the Machine Learning journal. In addition, I was a founding

editorial board member for the journals Bayesian Analysis, the Journal of Machine

Learning Research, and the Data Mining and Knowledge Discovery Journal. I have

served as a reviewer of papers for all of the major conferences and journals in my

field as well a reviewer of proposals for the National Science Foundation and for

NASA. I served as program chair for the annual ACM SIGKDD Conference on

Knowledge Discovery and Data Mining in 2011, and I was invited to serve as the

keynote speaker at the British International Conference on Databases in July 2015

(http://conferences.inf.ed.ac.uk/BICOD2015/).

16. As part of my work in connection with this proceeding, I have

reviewed the following materials:

• U.S. Patent 8,473,379 (Ex. 1001); • File History for U.S. Patent 8,473,379 (Ex. 1002) • U.S. Patent No. 6,731,724 to Wesemann (Ex. 1004) (“Wesemann”) • U.S. Patent No. 6,366,910 to Rajaraman (Ex. 1005) (“Rajaraman”) • U.S. Patent No. 7,539,656 to Fratkina (Ex. 1006) (“Fratkina”) • Hopcroft, John E., and Jeffrey D. Ullman. Data Structures and

Algorithms. Boston, MA, USA, Addison-Wesley, pp. 75-106, 155-197, 306-346, 1983 (Ex. 1010)

• Donald, B. Crouch, Carolyn J. Crouch, and Glenn Andreas, The use of cluster hierarchies in hypertext information retrieval, Hypertext ’89 Proceedings, ACM Press, pp. 225-237, 1989 (Ex. 1011)




• Yvan Leclerc, Steven W. Zucker, Denis Leclerc, McGill University, A browsing approach to documentation, IEEE Computer, IEEE Press, pp 46-49, 1982 (Ex. 1012)

• Ricky E. Savage, James K. Habinek,Thomas W. Barnhart, The design, simulation, and evaluation of a menu driven user interface, Proceedings of the 1982 Conference on Human Factors in Computing Systems, ACM Press, pp 36-40, 1982 (Ex. 1013)

• Ricardo Baeza-Yates, Berthier Ribiero-Neto, Modern Information Retrieval, pp. 24-40, ACM Press, 1999 (Ex. 1014)

• Daniel Cunliffe, Carl Taylor, and Douglas Tudhope, Query-based navigation in semantically indexed hypermedia, Proceedings of the Eighth ACM Conference on Hypertext, pp. 87-95, ACM Press, 1997 (Ex. 1015)

• Hornstein, Telephone Voice Interfaces on the Cheap at § 2.3, Proceedings of the UBLAB '94 Conference, 1994 (Ex. 1016)

• De Bra, Paul, et al., Information Retrieval in Distributed Hypertexts, in RIAO, pp. 481–493, 1995 (Ex. 1017)

• U.S. Pat. No. 6,198,939 to Holstrom (Ex.1018) • Karen Sparck Jones, A look back and a look forward, Proceedings of the

11th ACM SIGIR International Conference on Research and Development in Information Retrieval, pp. 13-29, ACM Press, 1988 (Ex. 1019)

• Gerard Salton, Anita Wong, and Chung-Shu Yang, A vector space model for automatic indexing, Communications of the ACM, 18(11): 613-620, 1975 (Ex. 1020)

• Jinxi Xu and W. Bruce Croft, Query expansion using local and global document analysis, Proceedings of the 19th ACM SIGIR International Conference on Research and Development in Information Retrieval, pp. 4-11. ACM, 1996 (Ex. 1021)

• Carolyn J. Crouch, A cluster-based approach to thesaurus construction, Proceedings of the 11th ACM SIGIR International Conference on Research and Development in Information Retrieval pp. 309-320. ACM, 1988 (Ex. 1022)

• Hinrich Schütze and Jan O. Pedersen, A cooccurrence-based thesaurus and two applications to information retrieval, 1 Intelligent Multimedia Information Retrieval Systems and Management, pp. 266-274, 1994 (Ex. 1023)




• Güntzer et al., Automatic Thesaurus Construction by Machine Learning from Retrieval Sessions, 25 Information Processing & Management No. 3 pp. 265-273, 1998 (Ex. 1024)

• Mostafa et al., A Multilevel Approach to Intelligent Information Filtering: Model, System, and Evaluation, 15 ACM Transactions on Information Systems No. 4, pp. 368-399, 1997 (Ex.1025)

• U.S. Pat. No. 6,006,225 to Bowman et al. (Ex. 1026)

II. LEGAL FRAMEWORK

A. Obviousness

17. I am a technical expert and do not offer any legal opinions. However,

counsel has informed me as to certain legal principles regarding patentability and

related matters under United States patent law, which I have applied in performing

my analysis and arriving at my technical opinions in this matter.

18. I have been informed that a person cannot obtain a patent on an

invention if the differences between the invention and the prior art are such that the

subject matter as a whole would have been obvious at the time the invention was

made to a person having ordinary skill in the art. I have been informed that a

conclusion of obviousness may be founded upon more than a single item of prior

art. I have been further informed that obviousness is determined by evaluating the

following factors: (1) the scope and content of the prior art, (2) the differences

between the prior art and the claim at issue, (3) the level of ordinary skill in the

pertinent art, and (4) secondary considerations of non-obviousness. In addition, the

obviousness inquiry should not be done in hindsight. Instead, the obviousness




inquiry should be done through the eyes of a PHOSITA at the time of the alleged

invention.

19. In considering whether certain prior art renders a particular patent

claim obvious, counsel has informed me that I can consider the scope and content

of the prior art, including the fact that one of skill in the art would regularly look to

the disclosures in patents, trade publications, journal articles, conference papers,

industry standards, product literature and documentation, texts describing

competitive technologies, requests for comment published by standard setting

organizations, and materials from industry conferences, as examples. I have been

informed that for a prior art reference to be proper for use in an obviousness

analysis, the reference must be “analogous art” to the claimed invention. I have

been informed that a reference is analogous art to the claimed invention if: (1) the

reference is from the same field of endeavor as the claimed invention (even if it

addresses a different problem); or (2) the reference is reasonably pertinent to the

problem faced by the inventor (even if it is not in the same field of endeavor as the

claimed invention). In order for a reference to be “reasonably pertinent” to the

problem, it must logically have commended itself to an inventor's attention in

considering his problem. In determining whether a reference is reasonably

pertinent, one should consider the problem faced by the inventor, as reflected

either explicitly or implicitly, in the specification. I believe that all of the




references I considered in forming my opinions in this IPR are well within the

range of references a PHOSITA would have consulted to address the type of

problems described in the Challenged Claims.

20. I have been informed that, in order to establish that a claimed invention

was obvious based on a combination of prior art elements, a clear articulation of

the reason(s) why a claimed invention would have been obvious must be provided.

Specifically, I am informed that, under the U.S. Supreme Court’s KSR decision, a

combination of multiple items of prior art renders a patent claim obvious when

there was an apparent reason for one of ordinary skill in the art, at the time of the

invention, to combine the prior art, which can include, but is not limited to, any of

the following rationales: (A) combining prior art methods according to known

methods to yield predictable results; (B) substituting one known element for

another to obtain predictable results; (C) using a known technique to improve a

similar device in the same way; (D) applying a known technique to a known device

ready for improvement to yield predictable results; (E) trying a finite number of

identified, predictable potential solutions, with a reasonable expectation of success;

(F) identifying that known work in one field of endeavor may prompt variations of

it for use in either the same field or a different one based on design incentives or

other market forces if the variations are predictable to one of ordinary skill in the

art; or (G) identifying an explicit teaching, suggestion, or motivation in the prior




art that would have led one of ordinary skill to modify the prior art reference or to

combine the prior art references to arrive at the claimed invention.

21. I am informed that the existence of an explicit teaching, suggestion, or

motivation to combine known elements of the prior art is a sufficient, but not a

necessary, condition to a finding of obviousness. This so-called “teaching-

suggestion-motivation” test is not the exclusive test and is not to be applied rigidly

in an obviousness analysis. In determining whether the subject matter of a patent

claim is obvious, neither the particular motivation nor the avowed purpose of the

patentee controls. Instead, the important consideration is the objective reach of the

claim. In other words, if the claim extends to what is obvious, then the claim is

invalid. I am further informed that the obviousness analysis often necessitates

consideration of the interrelated teachings of multiple patents, the effects of

demands known to the technological community or present in the marketplace, and

the background knowledge possessed by a person having ordinary skill in the art.

All of these issues may be considered to determine whether there was an apparent

reason to combine the known elements in the fashion claimed by the patent.

22. I also am informed that in conducting an obviousness analysis, a

precise teaching directed to the specific subject matter of the challenged claim

need not be sought out because it is appropriate to take account of the inferences

and creative steps that a PHOSITA would employ. The prior art considered can be




directed to any need or problem known in the field of endeavor at the time of

invention and can provide a reason for combining the elements of the prior art in

the manner claimed. In other words, the prior art need not be directed towards

solving the same specific problem as the problem addressed by the patent. Further,

the individual prior art references themselves need not all be directed towards

solving the same problem. I am informed that, under the KSR obviousness

standard, common sense is important and should be considered. Common sense

teaches that familiar items may have obvious uses beyond their primary purposes.

23. I also am informed that the fact that a particular combination of prior

art elements was “obvious to try” may indicate that the combination was obvious

even if no one attempted the combination. If the combination was obvious to try

(regardless of whether it was actually tried) or leads to anticipated success, then it

is likely the result of ordinary skill and common sense rather than innovation. I am

further informed that in many fields it may be that there is little discussion of

obvious techniques or combinations, and it often may be the case that market

demand, rather than scientific literature or knowledge, will drive the design of an

invention. I am informed that an invention that is a combination of prior art must

do more than yield predictable results to be non-obvious.

24. I am informed that for a patent claim to be obvious, the claim must be

obvious to a PHOSITA at the time of the alleged invention. I am informed that the




factors to consider in determining the level of ordinary skill in the art include (1)

the educational level and experience of people working in the field at the time the

invention was made, (2) the types of problems faced in the art and the solutions

found to those problems, and (3) the sophistication of the technology in the field.

25. I am informed that it is improper to combine references where the

references teach away from their combination. I am informed that a reference may

be said to teach away when a PHOSITA, upon reading the reference, would be

discouraged from following the path set out in the reference, or would be led in a

direction divergent from the path that was taken by the patent applicant. In general,

a reference will teach away if it suggests that the line of development flowing from

the reference’s disclosure is unlikely to be productive of the result sought by the

patentee. I am informed that a reference teaches away, for example, if (1) the

combination would produce a seemingly inoperative device, or (2) the references

leave the impression that the product would not have the property sought by the

patentee. I also am informed, however, that a reference does not teach away if it

merely expresses a general preference for an alternative invention but does not

criticize, discredit, or otherwise discourage investigation into the invention

claimed.

26. I am informed that even if a prima facie case of obviousness is

established, the final determination of obviousness must also consider “secondary




considerations” if presented. In most instances, the patentee raises these secondary

considerations of non-obviousness. In that context, the patentee argues an

invention would not have been obvious in view of these considerations, which

include: (a) commercial success of a product due to the merits of the claimed

invention; (b) a long-felt, but unsatisfied need for the invention; (c) failure of

others to find the solution provided by the claimed invention; (d) deliberate

copying of the invention by others; (e) unexpected results achieved by the

invention; (f) praise of the invention by others skilled in the art; (g) lack of

independent simultaneous invention within a comparatively short space of time;

(h) teaching away from the invention in the prior art.

27. I am further informed that secondary-considerations evidence is only

relevant if the offering party establishes a connection, or nexus, between the

evidence and the claimed invention. The nexus cannot be based on prior art

features. The establishment of a nexus is a question of fact. While I understand that

the Patent Owner here has not offered any secondary considerations at this time, I

will supplement my opinions in the event that the Patent Owner raises secondary

considerations during the course of this proceeding.

III. OPINION

A. Level of Skill of a Person Having Ordinary Skill in the Art




28. I was asked to provide my opinion as to the level of skill of a person

having ordinary skill in the art (“PHOSITA”) of the ’379 Patent at the time of the

claimed invention, which counsel has informed me to assume is November 19,

2002, the filing date of the ’379 Patent. In determining the characteristics of a

hypothetical person of ordinary skill in the art of the ’379 Patent at the time of the

claimed invention, I was told to consider several factors, including the type of

problems encountered in the art, the solutions to those problems, the rapidity with

which innovations are made in the field, the sophistication of the technology, and

the education level of active workers in the field. I also placed myself back in the

time frame of the claimed invention, and considered the colleagues with whom I

had worked at that time.

29. In my opinion, a person having ordinary skill in the art of the ’379

Patent at the time of its filing would have been a person having the equivalent of a

bachelor’s degree in computer science, electrical engineering, or a similar

discipline, and at least one year of experience working with technology related to

information retrieval and database searching, or an equivalent amount of similar

work experience or education, with additional education substituting for

experience and vice versa. Such a person of ordinary skill in the art would have

been capable of understanding the ’379 patent and the prior art references

discussed herein.




30. Based on my education, training, and professional experience in the

field of the claimed invention, I am familiar with the level and abilities of a person

of ordinary skill in the art at the time of the claimed invention. Additionally, I met

at least these minimum qualifications to be a person having ordinary skill in the art

as of the time of the claimed invention of the ’379 Patent

B. Background of the Technology

31. I was asked to briefly summarize the background of the prior art from

the standpoint of the knowledge of a PHOSITA prior to November 19, 2002, the

filing date of the ’379 Patent.

32. As the ’379 Patent shows in its “Background of the Invention,” the

concept of using keywords to navigate nodes or vertices arranged hierarchically,

such as in a graph structure, in a network was well known long before 2002. The

’379 Patent also recognizes that one “familiar” application of such a hierarchical

system is an automated telephone voice response system. See Ex. 1001 at 1:40-45.

The ’379 Patent also acknowledges that “travers[ing] the network in the most

efficient manner possible” is a desirable feature of hierarchical navigation systems.

Id. at 1:23-26; see also id. at 2:9:18.

33. The representation of interconnected nodes in a hierarchical network,

such as a tree structure like that described in the ’379 Patent, has long been well

known in the art and has long been a subject covered in introductory computer




science courses well before the ’379 Patent’s November 2002 filing date. 1 In

particular, all of the following would have been familiar to a PHOSITA before

2002: Arranging nodes in a network, such as a menu having different categories of

information, in a hierarchical structure,2 allowing users to navigate between nodes

or vertexes using key terms or node descriptors,3 and automatically searching a tree

to direct a user to a node or vertex associated with a key term or descriptor

associated with that node without traversing through other intervening nodes in the

hierarchy.4

34. In such hierarchical tree structures, where the root node contains all

objects, the children of the root node contain disjoint subsets of objects, and so on,

down to leaf nodes. By 2002, a PHOSITA would have known that there are

multiple different ways that an algorithm can search a tree to identify nodes of

interest and perform other operations. The different types of tree search algorithms

1 Hopcroft, John E., and Jeffrey D. Ullman. Data Structures and Algorithms. Boston, MA, USA,

Addison-Wesley, pp. 75-106, 1983. 2 Donald, B. Crouch, Carolyn J. Crouch, and Glenn Andreas, The use of cluster hierarchies in hypertext

information retrieval, Hypertext ’89 Proceedings, ACM Press, pp. 225-237, 1989. 3 Yvan Leclerc, Steven W. Zucker, Denis Leclerc, McGill University, A browsing approach to

documentation, IEEE Computer, IEEE Press, pp. 46-49, 1982. 4 Ricky E. Savage, James K. Habinek,Thomas W. Barnhart, The design, simulation, and

evaluation of a menu driven user interface, Proceedings of the 1982 Conference on Human Factors in Computing Systems, ACM Press, pp 36-40, 1982




are among the basic concepts that have been taught in introductory computer

science courses since the 1980’s and earlier.5

35. For example, one of the more well-known and efficient strategies for

searching a tree is the top-down depth-first approach.6 In the context of a user

searching for information where the information is represented by a tree structure,

the algorithm begins the search by evaluating the similarity of the item to be

matched (such as input from a user in the form of a keyword or query) to a set of

possible matches (represented as branches) at the root node of the tree. If one or

more branches match the input, the best such match is selected, the algorithm

descends the tree to the child node, and the matching process is repeated at the

child node, potentially requesting additional information from the user along the

path. If there is no adequate match at the root node, or at a subsequent child node

along the search path, the search algorithm can halt and return the best result(s)

found to that point or no result. If the algorithm reaches a leaf node (with no

children) the algorithm returns the result at the leaf node to the user, and

potentially additional result(s) found earlier along the search path. This is just one

potential version of a search algorithm to address the problem of a user searching

5 Hopcroft, John E., and Jeffrey D. Ullman. Data Structures and Algorithms. Boston, MA, USA:

Addison-Wesley, pp. 155-197, 306-346; 1983. 6 Donald, B. Crouch, Carolyn J. Crouch, and Glenn Andreas, The use of cluster hierarchies in hypertext

information retrieval, Hypertext ’89 Proceedings, ACM Press, pp. 225-237, 1989.




for information where the information is represented within a tree structure. A

PHOSITA by 2002 would have known that there were multiple different ways that

such algorithms could be constructed.

Navigation of Hierarchical Structures through “Jumping”

36. As the ’379 Patent recognizes, the comparison of input words with

keywords associated with a node to determine if the input matches the node was a

well-known concept before the time of the ’379 Patent. See, e.g., Ex. 1001 at 5:49-

60 (recognizing “fruit” at the “fruit” node). There were many known methods in

the art of performing a comparison between user input and keywords associated

with a node to determine if the input is sufficiently similar (exceeding some set

threshold similarity) to the keywords for the node. Examples of such methods that

were well-known in 2002 included Boolean matching, vector-space matching,

probabilistic matching, and fuzzy set matching, among others.7

37. Further, it was also well known that all of the nodes in a hierarchy

could be searched by comparing user input to keywords associated with all of the

nodes so as to determine a set of sufficiently similar nodes for further exploration

or to determine the most similar node (or, in the alternative, to determine that the

7 Ricardo Baeza-Yates, Berthier Ribiero-Neto, Modern Information Retrieval, pp. 24-40, ACM Press,

1999.




input is not sufficiently similar to any node such that the input is considered not a

proper input).8

38. However, contrary to the ’379 Patent’s disclosure, it was not always

necessary in the prior art for users to navigate through each node/vertex of a

hierarchical menu tree by methodically providing a node selection (e.g., by

answering a question) at each level in the hierarchy so as to reach the goal

node/vertex of interest. Instead, prior art approaches allowed users to simply

“jump” to their goal node of interest. The concept of directly jumping to a node by

mapping user input to keywords associated with nodes was nothing new by 2002.9

By this time, hierarchical navigation systems were becoming increasingly complex

with a large potential number of nodes in a network. Persons of ordinary skill in

the art understood that to save time with such large amounts of data and large trees,

it was conventional for systems to allow users to navigate directly to non-adjacent

nodes upon receiving search terms containing key phrases or descriptions

associated with those nodes.10

8 Daniel Cunliffe, Carl Taylor, and Douglas Tudhope, Query-based navigation in semantically

indexed hypermedia, Proceedings of the Eighth ACM Conference on Hypertext, pp. 87-95, ACM Press, 1997.

9 Hornstein, Telephone Voice Interfaces on the Cheap at § 2.3, UBILAB Rep, Union Bank of Switzerland, Zurich, 1994; De Bra, Paul, et al., Information Retrieval in Distributed Hypertexts, in RIAO, pp. 481–493, 1995

10 Daniel Cunliffe, Carl Taylor, and Douglas Tudhope, Query-based navigation in semantically indexed hypermedia, Proceedings of the Eighth ACM Conference on Hypertext, pp. 87-95, ACM Press, 1997; Yvan Leclerc, Steven W. Zucker, Denis Leclerc, McGill University, A browsing approach to documentation, IEEE Computer, IEEE Press, pp 46-49, 1982




39. Some menu navigation search systems were known in the art that,

upon receiving certain input from a user, would generate a suggested node or

nodes to which users may then choose to jump.11 However, other systems would

not require user input and would, instead, simply jump automatically to the node

best matching the input.12 In my opinion, any distinction between a hierarchical

navigation system that suggests different nodes for a user to choose where to

“jump” and a system which simply chooses a particular node to which it will

“jump” based on input received from the user and, for example a ranking of nodes

related to the input received from the user, would have been obvious to a

PHOSITA. Software to perform such functionality would have been well-

understood by persons having ordinary skill in the art. It would have been a mere

matter of design choice with respect to the particular application and desired user

interface as to whether the system automatically jumps to the node or whether the

system presents potential options that the user then chooses from to jump to the

selected node.

40. Given all of the above, it would have been obvious to a person or

ordinary skill in the art by 2002 to design a hierarchy navigation system that would

automatically “jump” users from one node to another node that is not directly

11 See, e.g., U.S. Pat. No. 6,198,939 to Holstrom. 12 See. e.g., U.S. Pat. No. 6,731,724 to Wesemann and U.S. Pat. No. 7,539,656 to Fratkina.




connected by matching input words with keywords without requiring the user to

traverse through intermediate connecting nodes.

Using and Generating Key-term Thesauruses

41. There is a long history of the use of thesauruses in information

retrieval, going back to the 1960’s. 13 The key idea is that the specific word

provided by a user in their query to an information retrieval system, might not be

present in that specific form in any of the relevant documents or information

known to the system. However, relevant variants of the word provided by the user

might be present in the documents, variants such as synonyms or alternative

morphological forms of the word. Thus, it was known as early as the 1960’s, that

using thesauruses to capture synonyms and alternative variants of words could

significantly enhance the capability of an information retrieval system to match a

user query to a set of relevant documents or information.14

42. It was also well-known by 2002 that there were advantages to updating

or learning a thesaurus via automated analysis of text.15 There were a number of

different types of learning algorithms in 2002 that could be applied to the problem 13 Karen Sparck Jones, A look back and a look forward, Proceedings of the 11th ACM SIGIR

International Conference on Research and Development in Information Retrieval, pp. 13-29, ACM Press, 1988.

14 Gerard Salton, Anita Wong, and Chung-Shu Yang, A vector space model for automatic indexing, Communications of the ACM, 18(11): 613-620, 1975.

15 Jinxi Xu, W. Bruce Croft, Query expansion using local and global document analysis, Proceedings of the 19th ACM SIGIR International Conference on Research and Development in Information Retrieval, pp. 4-11. ACM, 1996.




of learning a thesaurus. These algorithms typically relied on automatically finding

clusters of words (resulting in “concepts” for the learned thesaurus) that tend co-

occur frequently in sentences, paragraphs, and documents, using word co-

occurrences derived from documents known to the system.16

43. Another approach to learning a thesaurus, is to utilize logs of past user

searches in the system.17 The concept of learning new words given a user’s actions

after submitting a previously unrecognized query is a form of “query expansion”

and was already a known computer functionality for different search systems.18

44. Updating a thesaurus based on the inputs or searches of multiple users

would improve the thesaurus’s ability to learn new synonyms and would prevent

systems from returning no results to users searching a previously unknown term

for the first time. Persons of ordinary skill by 2002 would have recognized the

benefits of thesauruses storing new synonyms or keywords based on the learned 16 Carolyn J. Crouch, A cluster-based approach to thesaurus construction, Proceedings of the

11th ACM SIGIR International Conference on Research and Development in Information Retrieval pp. 309-320. ACM, 1988; Hinrich Schütze and Jan O. Pedersen, A cooccurrence-based thesaurus and two applications to information retrieval, 1 Intelligent Multimedia Information Retrieval Systems and Management, pp. 266-274, 1994.

17 Güntzer et al., Automatic Thesaurus Construction by Machine Learning from Retrieval Sessions, 25 Information Processing & Management No. 3 pp. 265-273, 268, 1998 (“User searches allow an unambiguous conclusion to be drawn as regards a thesaurus content and this can accordingly be directly acquired without any query being directed at the user.”); Mostafa et al., A Multilevel Approach to Intelligent Information Filtering: Model, System, and Evaluation, 15 ACM Transactions on Information Systems No. 4, pp. 368-399, 396, 1997; Larry Fitzpatrick, Mei Dent, Automatic feedback using past queries: social searching?, 31 ACM SIGIR Forum, pp. 306-313. ACM, 1997.

18 Jinxi Xu, W. Bruce Croft, Query expansion using local and global document analysis, Proceedings of the 19th ACM SIGIR International Conference on Research and Development in Information Retrieval, pp. 4-11. ACM, 1996




search behaviors of multiple users and would have known how to implement such

a system.19

45. Given all of the above, it would have been obvious to a PHOSITA by

2002 to design a hierarchy navigation system that would incorporate an evolving

thesaurus that would “learn” keywords based on, for example, observing new user

inputs and subsequent navigations and node selections within the hierarchical

network.

C. Obvious to Apply Wesemann to the Claims of the ’379 Patent

46. Wesemann teaches a voice-enabled user interface that enables a user to

navigate systematically through a hierarchy of menu states. See Ex. 1004 at

Abstract, 2:45-65, 3:50-55, Fig. 6.

47. In my opinion, a PHOSITA at the time of the filing of the ’379 Patent

would have concluded that Wesemann is analogous art to the ’379 Patent. For

example, both Wesemann and the ’379 Patent are in the field of navigating in a

transaction processing system and, more specifically, navigating a system having a

hierarchical arrangement of nodes. See Ex. 1001 at Abstract, Field of Invention;

see also Ex. 1004 at 1:13-19, 3:6-14, 3:50-55. Therefore, Wesemann is in the same

field of endeavor as the ’379 Patent. Additionally, much like the ‘379 Patent,

Wesemann attempts to improve the efficiency of existing hierarchical navigation 19 See, e.g., U.S. Pat. No. 6,006,225 to Bowman, et al.




systems by enabling users to jump to any level or menu state without requiring the

user to navigate through intermediate levels or menu states in the hierarchy.

Compare id. at Abstract, 3:10-14, 3:38-56, 10:13-18, 11:47-56, 11:65-12:12,

12:25-42, 12:65-13:2, Figs. 5 and 6; with Ex. 1001 at Abstract, 5:7-12, 12:49-56,

13:53-63, Fig. 6. Therefore, Wesemann is tied directly to a problem purportedly

addressed by the ’379 Patent. In my opinion, Wesemann would have logically

commended itself to the inventors of the ’379 Patent, particularly since Wesemann

addresses a problem the inventors allegedly faced in much the same way as the

inventors did. Therefore, in my opinion, Wesemann is analogous art because it is

reasonably pertinent to a problem faced by the inventors when they allegedly

conceived of the claimed invention.

48. Claim 1 of ’379 Patent recites a method related to navigating nodes

arranged hierarchically, including receiving input at a first node from a user related

to a keyword from among multiple keywords, identifying at least one other node

associated with the keyword, where the second node is not directly connected to

the first node, and jumping to that node.

49. As discussed, Wesemann teaches a voice-enabled navigation system

based on a hierarchy of “levels” or “menu states” that are mapped by a voice-

enabled user interface. See, e.g., Ex. 1004 at 8:35-39. A PHOSITA would

recognize that these levels or menu states are the same as the “vertices” or “nodes”




described in the ’379 Patent, as these levels or menu states each represent choices

or options in the hierarchical network. In the system in Wesemann, a user may

navigate a telephone service system by, for example, speaking into a telephone or

similar device. See, e.g., Ex. 1004 at 3:28-29, 4:58-60, 6:56-58. Speech recognition

software receives and interprets the user’s input, and the input is compared against

acceptable user inputs stored in the template. Id. at 6:56-64, 8:37-63. As the ’379

Patent acknowledges, many such speech recognition techniques were already well

known. Ex. 1001 at 6:60-64. Once this process is complete, the system can

determine what level or menu state the user likely desires and directly “jump” the

user directly to that level or menu state in the hierarchy without traversing through

each “in-between,” or intermediate step in the hierarchy of nodes. See, e.g., id. at

10:13-18, 11:47-56, 11:65-12:12, 12:25-42. A translator is used to convert an

acceptable user input corresponding to the user input into DTMF signals. Id. at

6:56-7:10, 10:65-11:16.20 This translation step in Wesemann is meant to enable

backwards-compatibility with existing DTMF phone service systems, which are

only responsive to commands sent via DTMF signal. See, e.g., id. at 3:6-10, 3:47-

49, 10:10-20.

20 There appears to be a typo in Col. 10, Line 66 of Wesemann. It states, “assuming the input does already comprise a DTMF signal,” but a person of ordinary skill would have appreciated that this is supposed to say, “assuming the input does not already comprise a DTMF signal.”




50. In my opinion, a PHOSITA would understand that the “acceptable

inputs and requests” mapped in the template to all of the menu states (nodes) of

Wesemann’s hierarchy constitute “keywords” within the meaning of the ’379

Patent. See Ex. 1001 at 12:13-24; see also id. at 10:65-11:16. I note that Petitioner

has proposed that the broadest reasonable interpretation of “keyword” in the ’379

Patent is “one or more words or pieces of information, such as a data pattern, that

is associated with at least one node or vertex” based on the patent’s description of a

“keyword.” Id. at 7:5-9 (“Note, there is no requirement for a [] ‘keyword’ to be a

single word, in some implementations, keywords could be single words, phrases of

two or more words, or even some other form of information like a specific data

pattern.”). The inclusion of “data patterns” within this definition of “keyword” is

consistent with the ’379 Patent’s description of embodiments that use an

interactive voice response (“IVR”) system in which a user responds vocally to

prompts and the user’s speech is compared to “keywords.” Id. at 6:63-7:9. A

PHOSITA would understand that the user’s speech is compared to the acceptable

responses mapped in the template by comparing the data patterns present in the

user’s speech with the data patterns representing the acceptable responses. This is

done to determine whether the user’s spoken words are sufficiently similar to any

acceptable response associated with a node (i.e., to “determine[] whether the input

is proper for any state of the telephone service system,” Ex. 1001 at 11:17-21) and,




if so, then to identify the “most similar” node (or node that “most resembles” the

input). Ex. 1004 at 12:12-24, 7:3-14. For instance, using an example from

Wesemann, the user’s spoken input words of “refurbished laptop sales” are

compared to the acceptable responses contained in the template and are determined

to match with “refurbished laptop sales” acceptable input for node 672 in Fig. 6 of

Wesemann. Id. at 11:65-12:6; Fig. 6. If the voice-enabled interface determines that

the input is not sufficiently similar to any acceptable response (keyword)

associated with any node, then the input is deemed “not proper” and the system

may prompt the user for new input. Id. at 11:21-31. In addition, if the voice-

enabled interface cannot determine the most similar node to which the user’s input

corresponds (e.g., if it matches acceptable responses, i.e., keywords, for multiple

nodes), then the interface may prompt the users with a clarifying prompt asking

which node the user would like from among the matching nodes. Id. at 12:4-12.

51. A PHOSITA would have found Wesemann’s “template,” which maps

acceptable responses to each of Wesemann’s menu states (nodes) to be just like the

’379 Patent’s “index,” which maps keywords to each of the nodes in the ’379

Patent’s hierarchy. Ex. 1001 at 5:2-4 (“An index . . . associating these keywords

with the nodes containing them is then created.”); 6:7-11 (“[W]hen a response to a

verbal description is provided by a user, possible keywords are identified in the

response and used to search the index and identify any node to which the response




may be directed … .”). Further, a PHOSITA would have found Wesemann’s

comparison of the user’s spoken words to the “acceptable inputs and requests”

stored in the template for an interactive voice response telephone system to be just

like the ’379 Patent comparison of a user’s spoken words to keywords stored in an

index for the ’379 Patent’s interactive voice response telephone system. The fact

that Wesemann’s interface also additionally translates the accepted user input into

a DTMF signal for purposes of communicating with the underlying telephone

service system, which is only responsive to DTMF signals, does not change this

conclusion, because this translation is merely a change in format that allows the

user’s command to be communicated to the telephone service system (e.g., for

purposes of causing the telephone service system to jump to the desired node).

52. I was also asked to consider a scenario in which I was asked to assume,

in the alternative, that the “acceptable inputs and responses” mapped to the nodes

in Wesemann’s template were not “keywords” per se within the meaning of the

’379 Patent. Even if they were not (which I disagree with as discussed in the

preceding two paragraphs), in my opinion, at a bare minimum, it would have been

exceedingly obvious to a PHOSITA to implement the “acceptable inputs and

responses” as keywords. It would have been obvious, because using keywords to

search a hierarchy of nodes to determine nodes having associated keywords

matching a user’s input had long been well known well before 2002, as discussed




above (see paragraphs 36, 39-40). Indeed, most conventional computer operating

systems at that time (and still today) used file systems allowing users to search for

files or folders of interest in a hierarchy of folders using keywords. Because of

their common usage, using keywords to perform the comparison of the user’s input

with the nodes in the tree would have been natural to a PHOSITA, and a

PHOSITA could have achieved predictable results in implementing this

functionality using keywords, without any undue experimentation. Keywords, at

their base level, can be represented in numerical or binary form, in relation to

which, other keywords (also represented in numerical or binary form) can be

compared using a similarity function defined using the numerical or binary forms,

to determine whether there is sufficient similarity so as to constitute a “match.” See

Ex. 1001 at 3:35-43, 9:13-16, 21:36-48. It would have been a matter of mere

design choice as to the precise algorithm that were used to perform the comparison

between user input and the set of acceptable responses/inputs, and using keywords

would have been a well-known and obvious way to do so. Further, Wesemann’s

express teaching of “acceptable inputs and responses” would have, at a minimum,

at least been highly suggestive to a PHOSITA to use keywords and would have

also motivated a PHOSITA to use keywords.

53. In my opinion, a PHOSITA by November 2002 would have concluded

that the limitations of claim 1 read on, or at a minimum, are at least obvious




variations of, the teachings of Wesemann, both individually and as a whole. For

example, Wesemann teaches receiving user inputs containing words, such as

speech, at a given menu state (node). Ex. 1004 at 3:28-29, 4:58-60, 6:56-58. If the

input is proper (i.e., matches or is sufficiently similar to an acceptable input or

response, or keyword, mapped in the template), a translator provides a DTMF

translation of the closest acceptable user input in the template for transmission to a

legacy telephone service system, which in turn transmits data to a user or performs

a requested process, such as transitioning by “jumping” to a different menu state.

See id. at 10:65-11:16; see also id. at 12:13-19, 3:35-46. The system does this by

having the telephone service system transition between the different menu states

without the knowledge of the user and without requiring any effort on the part of

the user. See id. at 12:53-13:2.

54. Additionally, to the extent that claim 1 requires that multiple keywords

be associated with a single node, Wesemann discloses this concept as well. For

example, Wesemann teaches that “[a]dditional data can also be stored in the

template to associate common correlating inputs with the acceptable inputs and

requests.” Id. at 12:19-21. A PHOSITA would have understood this to mean that

menu states could be mapped to more than one possible acceptable input, or

keyword.




55. The teachings of Wesemann are readily understood through its

exemplary disclosures. In one example, which references the hierarchical menu

tree displayed in Figure 6, Wesemann teaches that a user may input “123” from

main menu state 610 (i.e., receiving a user input at a first node). See id. at 11:52-

56. The input “123” is identifiable with at least one keyword of another node,

“extension 123” among multiple keywords shown in Figure 6. See id. at 11:52-56,

12:30-32, Fig. 6. Although extension 123 is not directly connected to main menu

state 610, upon receiving the input “123,” the system will “identify the user input

as a valid input for another state of the menu hierarchy,” (i.e., identifying a node

other than the first node) and the user will “automatically be transferred” by the

system to extension 123, without forcing the user to traverse through the

intermediate node, “Directory of Personnel” 640 (i.e., jumping the user to the other

node). See id. at 11:33-56, 12:25-32. This example demonstrates “vertical”

jumping.

56. In a second example, Wesemann illustrates each limitation of claim 1

in a “lateral” jumping context. See id. at 12:25-36. Again with reference to Figure

6, a user begins at home laptop sales 652 (i.e. a first node). By saying, “home

computer support,” the system will identify this is a proper input for home

computer support 646 because each menu state is mapped in a template (i.e.

receiving an input at a first node, the input containing at least one word identifiable




with at least one keyword from among multiple keywords). Id.; see also id. at

12:13-24. Upon identifying the desired menu state based on the user input and

associated acceptable input - home computer support - in the template, the system

will “jump” from home laptop sales 652 to home computer support 646 without

requiring the user to traverse intermediate steps (i.e., identifying at least one node

other than the first node associated with the keyword and jumping to that node). Id.

57. Claim 2 depends from claim 1 and adds that the method includes

providing a verbal description associated with the at least one node to the user. The

’379 Patent discloses that “verbal descriptions” may include prompts or questions

designed to elicit further response from a user. See, e.g., Ex. 1001 at 3:35-43, 5:49-

60, 18:34-45. I understand that Petitioner has proposed that the broadest reasonable

interpretation of “verbal description” is “a set of words relating to the subject

matter whether presented audibly or in written form” based on the patent’s

description of this term. Id. at 1:50-52. A PHOSITA would understand that

Wesemann teaches that each of its menu states (nodes) have such associated

“verbal descriptions.” Specifically, Wesemann teaches providing menu prompts

corresponding to each of the menu states—and these prompts may be provided as

pre-recorded messages the user. See, e.g., Ex. 1004 at 5:62-67, 12:13-16, 13:3-22;

see also id. at 11:12-16 (“After transmitting requested data or performing a

requested process, the telephone service system waits for additional user input. The




telephone service system may also prompt the user for additional input.”). For

example, these verbal menu prompts may include “press 1 followed by # for

sales,” “press 2 followed by # for computer support,” “or press 3 followed by # for

a directory of personnel.” Id. at 11:33-38. As other examples, these prompts may

also include prompts for “home computer sales,” “business computer sales,” or

“refurbished computer sales,” for each of those particular nodes. Id. at Fig. 6,

11:65-12:6. Wesemann also teaches a system that provides a “clarifying prompt”

when a user mistakenly enters an incorrect or ambiguous input. See, e.g., id. at

11:65-12:12. Given these disclosures in Wesemann, in my opinion, a PHOSITA

would have understood that Wesemann teaches “providing a verbal description

associated with the at least one node to the user,” as in claim 2. I note that “the at

least one node” in claim 2 refers to the node in claim 1 to which the system has

jumped. A PHOSITA would understand that, as with all of the nodes, after the user

jumps to a node, the user is presented with the verbal description for that node. For

example, if the user jumps to the sales node 630, the user would be presented with

a prompt asking the user whether they would like home computer sales 650,

business computer sales 660, or refurbished computer sales 670.

58. Claim 7 of the ’379 Patent recites a method performed in connection

with an arrangement of nodes representable as a hierarchical graph containing

vertices and edges connecting at least two of the vertices. The method includes




receiving an input from a user as a response to a verbal description associated with

a first vertex, analyzing the input to identify a meaningful term that can be

associated with at least one keyword, and selecting a second vertex not connected

by an edge to the first vertex, based upon an association between the meaningful

term and the at least one keyword and a correlation between the at least one

keyword and the vertex; and jumping to the second vertex.

59. In my opinion, a PHOSITA would have concluded that the limitations

of claim 7 read on, or at least are obvious variations of, the teachings of

Wesemann, both individually and as a whole. For example, similar to the ’379

Patent, Wesemann teaches an arrangement of nodes representable as a hierarchical

graph containing vertices and edges connecting at least two vertices. For

comparison, Figure 6 of each of Wesemann and the ’379 Patent are “menu trees,”

an example of such a graph representation. See, e.g., Ex. 1001 at 3:5-28.




Ex. 1001, Fig. 6 (describing a network of nodes in the airline industry context)

Ex. 1004, Fig. 6 (describing menu hierarchy for computer sales organization)

60. Wesemann teaches receiving an input from a user as a response to a

verbal description, such as menu prompts, associated with a first vertex. See, e.g.,

id. at 5:62-67, 12:13-16, 13:3-22; see also id. at 11:12-16 For example, if a user




wanted to purchase refurbished laptops from a computer sales organization, for

which an exemplary graphical representation of a menu is provided in Figure 6, the

user could “systematically” enter input in response to menu prompts to step

through each of the vertices of sales 630 and refurbished computer sales 670 to

arrive at refurbished laptop sales. See id. at 11:65-12:4. A PHOSITA would

recognize that the user inputs in response to menu prompts at any one of these

nodes in the hierarchical menu of Wesemann would satisfy this limitation of claim

7. For example, at the main menu node 610, the system provides a menu prompt

asking the user if they would like computer support, sales, or a personnel directory.

Id. at 11:33-55, Fig. 6. While Wesemann states that prior art systems would have

only recognized acceptable/responses inputs associated with those three nodes in

response to this prompt at the main menu, Wesemann teaches that its system will

search the entire tree so as to recognize any acceptable response/input, such as the

input “123,” which will cause the system to jump directly to the node associated

with that acceptable response/input (i.e., jumping to extension 123 in the personnel

directory, in this example). Id. at 11:47-55. These verbal prompts provided by the

system to the user over the telephone would be understood by a PHOSITA to be a

verbal description (which Petitioner has construed as “a set of words relating to the

subject matter whether presented audibly or in written form” based on the patent’s

description of this term). A PHOSITA, therefore, would understand that Wesemann




teaches receiving an input from a user as a response to a verbal description

associated with a first vertex.

61. A PHOSITA would understand Wesemann also teaches claim 7’s step

of “analyzing the input to identify a meaningful term that can be associated with at

least one keyword.” First, Wesemann teaches using speech recognition technology

and comparing the user’s spoken words/terms to the set of “acceptable responses

and inputs” that are mapped to all the menu states (nodes) in the Wesemann’s

template. See e.g., id. at 6:65-7:14, 8:37-63, 10:65-11:16, 12:13-24. The system

performs a process of “comparing the user input with the accepted inputs and

requests on file” to determine what the user is attempting to do or say. Id. at 12:13-

19. If an input includes any term that is mapped on the template, the system will

associate this to at least one acceptable response/input (keyword). See, e.g., id. at

11:52-56, 12:25-42, 6:56-7:22. Based on this analysis, the system will identify a

meaningful term which can be associated with an acceptable input/response,

which, as I have discussed above in paragraphs 51-52, would be understood by a

PHOSITA to be a keyword (“one or more words or pieces of information, such as

a data pattern, that is associated with at least one node or vertex”). Second,

Wesemann teaches that the template may also contain additional “common

correlating inputs” for each of the acceptable inputs/responses (keywords) to allow

the system to better determine what a user is attempting to input. Id. at 12:13-24. A




PHOSITA would understand that these additional correlating inputs allows

additional meaningful terms from the user’s spoken input to be associated with an

acceptable response/input (keyword) used to identify a node to jump to. Third, in

examples where the input cannot be used to uniquely determine the most similar

acceptable response/input (keyword), such as if a user mistakenly says

“refurbished notebook sales,” for example, the system will identify at least one

meaningful term, such as the term “refurbished” that is associated with the

acceptable inputs/responses for “refurbished laptop sales” and “refurbished

desktop sales” based on their use of the word “refurbished.” Id. at 11:65-12:12.

The system will then provide the user with a clarifying prompt to suggest either

“refurbished laptop sales” or “refurbished desktop sales.” Id. at 11:65-12:12. A

PHOSITA, therefore, would understand that Wesemann teaches claim 7’s step of

“analyzing the input to identify a meaningful term that can be associated with at

least one keyword.” In addition, for the same reasons I have expressed in

paragraphs 52-53 above, it would have also been obvious to a PHOSITA to use

“keywords” in Wesemann to the extent one were to assume that Wesemann’s

“acceptable inputs and responses” were somehow not “keywords.”

62. Finally, a PHOSITA would conclude that Wesemann teaches the final

limitation of claim 7, which, like the final limitation of claim 1, requires that the

system “jump” from a first vertex to a non-connected vertex associated with the




keyword and meaningful term input by the user at the first vertex. As discussed

above, Wesemann teaches the concept of jumping from a first vertex to a second,

non-directly-connected vertex based on key terms or meaningful terms of a user’s

input. See supra at paragraphs 49-56; see also Ex. 1004 at Abstract, 3:10-14, 3:38-

56, 10:13-18, 11:33-56, 11:65-12:12, 12:25-42, 12:65-13:2, Figs. 5 and 6. To

provide an example, Wesemann teaches comparing the user’s spoken words/terms,

such as “123” or “home computer sales” to the set of acceptable responses/inputs

(keywords) mapped to the vertices of the hierarchy in the template and then

jumping to the vertex having the acceptable response/input (keyword) that matches

the user’s input. Id. at 11:33-56, 12:25-13:2, Fig. 6. Wesemann provides examples

of both a lateral jump (from the “home computer sales” vertex to the “home

computer support” vertex) and a vertical jump (from the “main menu” vertex to the

“extension 123” vertex), as depicted in the annotated version of Fig. 6 below.




A PHOSITA, therefore, would understand that Wesemann teaches claim 7.

63. In my opinion, a PHOSITA would have found that Wesemann teaches

the limitations of claims 1, 2, and 7 of the ’379 Patent. As discussed, any

distinctions whatsoever that could be made between Wesemann and these claims

of the ’379 Patent would have required at most very minor modifications that

would have yielded predictable results to a PHOSITA, and, thus would have been

obvious to such a person.

D. Obvious to Combine Wesemann and Rajaraman

64. Rajaraman teaches a “general purpose search” (GPS) method and

system that generates search results for classifications/nodes that, like the ’379

Patent, are arranged hierarchically. See, e.g., Ex. 1005 at Abstract, 1:5-7, 1:66-67,

2:9-22, 2:57-3:7, Figs. 4, 13:




Ex. 1005 at Fig. 4.

65. As discussed above, the ’379 Patent also relates to methods for

navigating a hierarchical system of nodes. See supra, at ¶ 47. In my opinion,

because Rajaraman and the ’379 Patent are both in the field of navigation in a

transaction processing system, specifically navigating a system having a

hierarchical arrangement of nodes, Rajaraman is analogous art for falling within

the same field of endeavor of the ’379 Patent.

66. Rajaraman also attempts to address at least two problems identified in

the ’379 Patent. First, both the ’379 Patent and Rajaraman purportedly seek to

improve efficiency and prevent user frustration in the navigation of a hierarchy of




nodes (referred to as categories or classifications and sub-classifications in

Rajaraman) to find an item of interest, and both the ’379 Patent and Rajaraman

allow keyword searching across nodes/classifications to address this issue.

Compare Ex. 1001 at 2:9-18 with Ex. 1005 at 1:42-2:22. Second, Rajaraman

teaches the use of a “special terms file” list of synonyms to expand search queries,

like the thesaurus of the ’379 Patent. A PHOSITA would have recognized that the

teachings in Rajaraman are reasonably pertinent to the problem addressed by the

’379 Patent related to improving efficiency of navigational systems by enabling

them to search synonyms of keywords, as well as assisting with the learning of

previously unknown keywords or synonyms. See Ex. 1001 at 3:44-48; see also Ex.

1005 at 7:22-8:25. In my opinion, given the problems faced by the inventors in the

’379 patent, a PHOSITA would have found that Rajaraman logically would have

commended itself to the inventors. Therefore, Rajaraman is also analogous art by

virtue of being reasonably pertinent to problems faced by the inventors of the ’379

patent.

67. Claims 3 through 6 of the ’379 Patent relate to a “thesaurus,” where

the thesaurus is implemented in software to correlate keywords with synonyms. I

note that claim 3 itself recites “a thesaurus correlating keywords with synonyms,”

and the ’379 Patent describes its thesaurus as “equating” synonyms with keywords,

so that input of a synonym by a user can also cause the claimed “jumping” to the




desired node. ’379 Patent at 8:3-32. The patent states that this “equating” can be

done in “any of a myriad of different ways” and gives examples such as correlating

words, correlating synonyms to nodes, etc. Id. at 8:28-41. But the ’379 Patent notes

that “the exact implementation details of which re [sic] irrelevant to the invention.”

Id. at 8:28-32.

68. Rajaraman teaches of a “special terms file” that “lists various words

(i.e., ‘Good Terms’) that are synonymous with classification names.” See Ex. 1005

at 7:22-26, Fig. 7. Rajaraman’s “index builder” accesses the special terms file

(thesaurus) during a search and assigns synonymous terms, which it calls “Good

Terms,” a priority of zero so as to include classifications having synonymous terms

in the search results. Id. at 8:26-30 and Fig. 9; see also id. at 9:7-45 and Fig. 11. In

my opinion, PHOSITA would have concluded that the special terms file taught in

Rajaraman, including the list of “Good Terms,” is the same as the “thesaurus”

identified in claims 3-6 of the ’379 Patent because it serves to correlate synonyms

with classification names (i.e., keywords).

69. Claim 3 is dependent on claim 1, and adds “searching a thesaurus

correlating keywords with synonyms.” Claim 4 depends on claim 3 and adds that

the searching comprises identifying “the at least one word as synonymous with the

at least one keyword.” Rajaraman discloses both of these limitations. With respect

to claims 3 and 4, Rajaraman teaches that its searching system uses an “index




builder” that goes through the “special terms file” (thesaurus) and assigns the

synonymous “Good Terms” a priority of 0 for search purposes (however,

Rajaraman recognizes that a different priority may be used). By giving Good

Terms 0-priority, Rajaraman ensures that search results include classifications

indicated in the special terms file as being synonymous with Good Terms, but that

they do not supersede the original classification terms (i.e., “keywords”), which are

assigned a priority of 1. Id. at 8:26-30 and Fig. 9; see also id. at 9:7-45 and Fig. 11.

As an example, Rajaraman teaches making “blouse” a “Good Term,” or synonym,

for “women’s shirts.” Id. at 7:22-26, Fig. 7. When a user searches for women’s

shirts the search will also produce results that have hits on the term “blouses” as

well, and if a user searches for “sole,” results related to the classification “shoes”

may be produced. See id. at 7:25-26, 7:43-62, 9:7-52, Figs. 7-11. As I discuss in

further detail below in paragraph 76, it would have been obvious to a PHOSITA to

incorporate this searching of synonyms of keywords, as taught by Rajaraman, in

Wesemann’s system.

70. Claim 5 depends from claim 1 and adds that the system would

determine that the user input word is neither a keyword or a synonym and learning

the meaning for the word so that the word will be treated as a learned synonym for

at least one particular keyword. Claim 6 depends from claim 5 and comprises




adding the word to a thesaurus so that when the word is input by a subsequent user,

it will be treated as synonymous.

71. Rajaraman teaches that its system maintains a log of search requests

and search results that may also log which search results are selected by the user,

so as to monitor user search behavior and to allow a determination of whether

synonyms should be added for a search term. Id. at 7:63-8:19. A “log analyzer” is

provided to help determine when to add synonyms by performing a statistical

analysis of search requests, such as those requests that “resulted in no results” or in

very few classifications appearing in the results. Id. Thus, a PHOSITA would

understand that Rajaraman teaches determining that the at least one-word input by

the user is neither a keyword nor a synonym of any keyword.

72. Rajaraman also teaches learning a meaning for the word so that it will

be treated as a learned synonym for at least one particular keyword of the multiple

keywords, as in claim 5, and it teaches learning the meaning by having the search

term added to the special terms list (i.e., the “thesaurus”) as a synonym, as in claim

6. For example, the log analyzer produces statistical results that can, for example,

be reviewed by an analyst to determine whether the search term should be added to

a special terms list. See id. at 7:63-8:25. Similar to the teachings of the ’379 Patent,

subsequent actions by the user after the unknown term is entered, such as

subsequent successful searches (i.e., searches that return results) may be provided




to help determine what classifications or keywords would be synonymous with the

previously unknown word. See id. at 8:16-19; see also, e.g., Ex. 1001 at 9:65-

11:12.

73. Rajaraman teaches that “Good Terms” added to the special terms list

do not always need to be literal synonyms. For example, it may be useful for the

system to treat common misspellings or often-misused homonyms of words as

synonyms. Rajaraman offers the term “aparel” as a potentially common

misspelling of the word, “apparel” and the word “sole” as homonym for “soul” that

may be misused. Ex. 1005 at 7:63-8:25. If these terms were often used but turn up

no-to-few search results, then this would be revealed by the log analyzer’s

statistical analysis.

74. In my opinion, a PHOSITA at the time of the filing of the ’379 patent

would have readily been able to substitute the log analyzer described in Rajaraman

with software algorithms to perform the function of adding synonyms to the

special terms list. As discussed in paragraphs 41-45, automatic thesaurus-updating

was already well known by the time of the ’379 Patent, and persons having

ordinary skill would have been able to automate the process of the “analyst”

described in Rajaraman to use the log analyzer to automatically add learned

synonyms determined from an analysis of user’s search behavior. Such automation

of these steps would have required the application of existing computer technology




using well-known methods of programming that would have been within the level

of ordinary skill in the art. A PHOSITA would have also appreciated the benefit in

allowing automatic addition of synonyms for keywords to avoid the well-known

issues with labor-intensive manual updating of keyword thesauruses, as also

discussed above in paragraphs 42-43.

75. With respect to claim 6’s reference to “a subsequent user” being able

to have the benefit of having the learned word added to the thesaurus for

subsequent searching, I note that the ’379 Patent refers to a subsequent caller/user

being “either the same person or a different person.” Ex. 1001 at 14:32-38.

Nonetheless, counsel has asked me to consider an assumption that “a subsequent

user” in claim 6 must be a different user. Even under this assumption, a PHOSITA

would understand that Rajaraman satisfies such an interpretation. Rajaraman does

not limit its thesaurus to only a specific user, but instead bases it on the behavior of

multiple users and even on homonyms and alternate spellings, as discussed. As

Rajaraman states: “For example, users may enter the search term ‘aparel,’ rather

than ‘apparel.’ Because the term ‘aparel’ is not in the product database and not in

the classification hierarchy, the search result will be empty. Therefore, it would be

useful to add the term ‘aparel’ as a synonym of ‘apparel.’” Rajaraman at 7:67-8:5

(emphasis added); see also id. at 8:22-25 (“For example, if users enter the search

request ‘sole’ and the search results relate only to shoes, the analyst may want to




indicate that ‘sole’ is a synonym for ‘soul,’ as in music.”) (emphasis added).

Further, Rajaraman’s thesaurus (i.e., “the special terms file”) is part of

Rajaraman’s overall system, which is used by all users of the system. A PHOSITA,

therefore, would understand that Rajaraman is not limiting its thesaurus to only a

particular user, and, instead, subsequent users also benefit from the thesaurus of

synonyms, which is developed from the statistical analysis of multiple users’

search behaviors. Id. at 4:55-5:9, Fig. 2, 7:22-42, Fig. 7. It would have also been

obvious to a PHOSITA for the thesaurus to be applied to other users, because it

was well known in the art that thesauruses of synonyms for keyword searching are

improved when they are based on the learned search behaviors of multiple users,

which yields a more robust thesaurus functionality.

76. As mentioned above, claims 3, 4, 5, and 6 all depend at least

indirectly from claim 1. Wesemann teaches each limitation of claim 1. See supra at

paragraphs 48-56. Both Rajaraman and Wesemann relate to the field of navigation

of hierarchical networks. In my opinion, it would have been obvious for a

PHOSITA to incorporate Rajaraman’s use of a “special terms file” (thesaurus) for

searching a hierarchy of classifications, as well as Rajaraman’s log analyzer, in

Wesemann’s system for matching words in a user’s input with keywords

associated with nodes to jump to a node within a hierarchy. As discussed above in

paragraphs 41-45, use of such thesaurus functionality to search synonyms in




addition to keywords was well known by 2002, and a PHOSITA would have

appreciated that implementing the thesaurus functionality taught in Rajaraman

would have improved the system of Wesemann by making it more user-friendly

and increasing the efficiency with which the user could navigate through the

hierarchical network. Further, given that both Wesemann and Rajaraman relate to

searching a hierarchy of nodes and Wesemann already teaches inclusion of

additional data for “common correlating inputs” for the “acceptable responses and

inputs” (keywords) in Wesemann’s template, such modifications to the system in

Wesemann would have required only minor modifications in software and would

have yielded predictable results to a PHOSITA. These conclusions are supported

by the fact that the system in Wesemann expressly teaches adding “common

correlating inputs,” for acceptable user inputs into the template. See Ex. 1004 at

12:19-24. This express teaching of Wesemann would have, at a minimum,

provided a suggestion or motivation to a PHOSITA to use synonymous terms, such

as taught by Rajaraman’s “special terms files,” as the “common correlating inputs”

associated with the acceptable responses and inputs (keywords) of Wesemann,

such that user inputs corresponding to those synonymous terms would be deemed

by the system to be a match for their assigned keyword(s). Further, a PHOSITA

would have also recognized that the application of methods related to searching

keyword synonyms contained in a thesaurus, which were well known at that time,




would have yielded predictable results when applied in the system of Wesemann

and would not render any features of Wesemann, such as the ability to jump

between nodes, inoperable. Therefore, in my opinion, a PHOSITA would have

been motivated to combine the teachings of Wesemann and Rajaraman to render

claims 3-6 obvious.

E. Obvious to Apply Fratkina to the Claims of the ’379 Patent

77. Fratkina discloses a system for providing an automated, multi-step

dialogue with users of the system. Fratkina discloses taxonomies of parent, child,

and sibling nodes through which a user may navigate with the assistance of a

dialogue engine that processes user inputs. See Ex. 1006, 4:42-5:19, 6:65-7:5,

14:42-67. These taxonomies may be arranged hierarchically. See id.; see also id. at

14:47-67, 22:19-29, 26:26-27:27, Figs. 4, 5, 10-12. Like the ’379 Patent, Fratkina

allows the user’s input words to be spoken words that are recognized using voice

recognition software in an interactive voice response (IVR) system. Id. at 13:15-

29.

78. In my opinion, Fratkina is analogous art to the ’379 Patent. Both

Fratkina and the ’379 Patent relate to the navigation of transaction processing

systems, specifically in a hierarchical fashion. See id. at 14:47-67, 22:19-29, 26:26-

27:27, Figs. 4, 5, 10-12; see also Ex. 1001 at Abstract, Field of Invention.

Additionally, both the ’379 Patent and Fratkina disclose the use of “interactive




voice response” (“IVR”) systems. See Ex. 1006 at Abstract, 2:56-62, 13:15-17; see

also Ex. 1001 at 6:66-67. Therefore, Fratkina is within the field of endeavor of the

’379 Patent.

79. Fratkina is also reasonably pertinent to at least one problem addressed

in the ’379 Patent – improving the efficiency of navigating menu systems. Id. at

2:12-14, 40:30-33. Additionally, like the ’379 Patent, Fratkina teaches the use of

“jumping” between nodes to accomplish efficient navigation. See id. at 34:32-53;

see also id. at 27:25-43; see also id. at 37:54-63, Fig. 15. In my opinion, Fratkina

would have logically commended itself to the problems purportedly addressed of

the ’379 Patent. Therefore, Fratkina is also analogous art because it is reasonably

pertinent to the problems faced by the inventors of the ’379 Patent.

80. Claim 1 of ’379 Patent recites a method related to navigating nodes

arranged hierarchically, including receiving input at a first node from a user related

to a keyword from among multiple keywords, identifying at least one other node

associated with the keyword, where the second node is not directly connected to

the first node, and jumping to that node. As discussed above, Fratkina teaches a

system in which navigable nodes are arranged hierarchically. See Ex. 1006 at Figs.

4, 5, 10-12, 20.

81. Fratkina teaches limitation [1(a)]; specifically, Fratkina discloses

receiving input from a user at a node, where the input contains at least one word




that, when processed by a dialog engine, may be identifiable with at least one “tag”

(i.e., keyword), such as the name of the node. Id. at 13:15-29 (regarding user

input); 5:13-28 (“keywords”), 14:27-31 (regarding tags); see also id. at 26:26-

27:27. Multiple tags may be used for different nodes on the hierarchy. Id at 5:13-

28, 14:27-3, 26:26-27:27.

82. An example of Fratkina’s teachings is provided in the form of a meal

menu. See, e.g., id. at Figs. 10-12, 20, 21. Figure 11 of Fratkina is provided for

illustration:




83. In navigating this menu, the system may ask the user about what they

are seeking, and the user may answer using “keywords or natural language,” such

as by speaking, which can be recognized by voice recognition software in

Fratkina’s “interactive voice response (IVR)” implementation. See id. at 26:50-57,

22:19-20; see also id. at 5:24-29, 13:15-29. For example, a user may input “eggs”

at the breakfast node to move to the eggs node. Id. at 26:50-57; see also id. at Fig.

10, 22:19-29. A PHOSITA would understand that Fratkina teaches a system




where, at a first node, a user may input a command/response containing one or

more words identifiable with at least one or more keywords for navigating the

taxonomies.

84. Fratkina also teaches limitation [1(b)], i.e., “identifying at least one

node, other than the first node, that is not directly connected to the first node but is

associated with the at least one keyword, and jumping to the at least one node.” For

example, Fratkina states that while a user would “typically” advance through the

menu along the edges of the taxonomy graph, the system may allow users to

“jump” to a node “more than one edge away from the previous focus” (i.e., the first

node the user was occupying). Id. at 27:25-43; see also id. at 34:32-53, 37:54-63.

The system is able to identify the non-adjacent node to which the user would like

to go through “autocontexualization.” Id. at 34:32-53 (“Autocontextualization can

be used to jump to a specific place in the taxonomy and the dialog designer can

explicitly specify a place to jump to.”). A PHOSITA would recognize these

teachings to render obvious the identifying and jumping limitation of claim 1. A

PHOSITA would have, for example, recognized immediately how the jumping

functionality provided by Fratkina’s autocontextualization process could be

applied to the breakfast menu example. If a user input “scrambled eggs,” for

example, while at the “breakfast node,” a PHOSITA would have understood that

the system would be able to recognize where the user desired to go and transfer the




user to the scrambled eggs node, without traversing through the “eggs” node

between the “breakfast” and “egg” nodes. Fratkina’s express teachings regarding

use of autocontextualization to “jump” to non-connected nodes would have

motivated a PHOSITA to allow such jumping within the menu hierarchy to

provide for a more natural and conversational dialogue, and, ultimately, a more

efficient flow in Fratkina’s system. Therefore, in my opinion, claim 1 is obvious

over Fratkina.

85. Claim 2 depends from claim 1 and adds that the method includes

providing a verbal description associated with the at least one node to the user. The

’379 Patent discloses that verbal descriptions may include prompts or questions

designed to elicit further response from a user. See, e.g., Ex. 1001 at 3:35-43, 4:32-

45, 5:49-60, 18:34-45. Fratkina teaches providing users with menu prompts in

response to user inputs associated with the at least one node (i.e., the second node

identified in claim 1). For example, Fratkina discloses prompting users for

information regarding whether they would like eggs or pancakes at a breakfast

node, and then whether the user would like poached or scrambled eggs upon the

user selecting eggs. Id. at 26:46-60, Fig. 11, Fig. 21; see also id. at 13:15-24 (“The

dialog engine 232 response is passed to a text-to-speech system that turns it into a

vocal response to the user.”), 26:34-45. Thus, in my opinion, a PHOSITA would

understand that Fratkina teaches the limitation added by claim 2 of the ’379 Patent.




86. Claim 7 of the ’379 Patent recites a method performed in connection

with an arrangement of nodes representable as a hierarchical graph containing

vertices and edges connecting at least two of the vertices. The method includes

receiving an input from a user as a response to a verbal description associated with

a first vertex, analyzing the input to identify a meaningful term that can be

associated with at least one keyword, and selecting a second vertex not connected

by an edge to the first vertex, based upon an association between the meaningful

term and the at least one keyword and a correlation between the at least one

keyword and the vertex; and jumping to the second vertex.

87. Fratkina discloses an arrangement of nodes that can be represented in a

graph containing vertices and edges connecting at least two of the vertices:




See, e.g., id. at Figs. 10, 20; see also id. at Figs. 11, 12.




88. Fratkina also discloses the second limitation of claim 7 related to

receiving an input from a user as a response to a verbal description. As just one

example, with reference to the menu example, Fratkina discloses that a user may

input “scrambled” as a response to the question at the breakfast node, “How would

you like your eggs prepared?” See id. at 26:34-60. Such questions align with the

teachings of the ’379 Patent regarding a “verbal description.” Thus, a PHOSITA

would understand that Fratkina teaches this limitation.

89. As discussed with respect to claim 1, Fratkina also discloses the

“jumping” concepts of the final limitation of claim 7. Using autocontextualization,

the system in Fratkina can select a vertex in a graph that is not connected to a first

vertex and, based on the user’s input and a keyword or “tag” associated with the

second vertex, jump from the first vertex to the second vertex. 34:32-53; see also

id. at 27:25-43; see also id. at 37:54-6. Although Fratkina does not apply its

teachings of jumping to a concrete example, one having skill in the art would

understand from Fratkina’s disclosure how to apply the teachings of Fratkina to

claim 7. For example, as with the breakfast menu example provided above for

claim 1 (see paragraph 84, limitation [1(b)]), a PHOSITA would understand that

the system would be able to jump from the “breakfast node” to the “scrambled”

node, skipping the “eggs node” altogether, upon receiving a user input such as

“scrambled eggs.” Thus, Fratkina taches all of the limitations of claim 7. Further, it




would have been obvious to a PHOSITA that Fratkina would be able to operate in

this way based on Fratkina’s express teachings with regard to how

autocontextualization can cause the system to jump to non-connected nodes. That

is, as discussed above for claim 1 (see paragraph 84, limitation [1(b)]), Fratkina’s

express teachings regarding use of autocontextualization to “jump” to non-

connected nodes would have motivated a PHOSITA to allow such jumping within

the menu hierarchy to provide for a more natural and conversational dialogue, and,

ultimately, a more efficient flow in Fratkina’s system.

90. In my opinion and based on the foregoing, claims 1, 2, and 7 would

have been obvious to persons having ordinary skill in the art over the teachings and

disclosures in Fratkina.

F. Obvious to Combine Fratkina and Rajaraman

91. As discussed in Section E, supra, claim 1 is obvious over Fratkina.

Additionally, as discussed in Section D, supra, Rajaraman teaches and/or renders

obvious the limitations of claims 3, 4, 5, and 6.

92. In my opinion, it would have been obvious for one of ordinary skill in

art to incorporate Rajaraman’s use of a “special terms file” (thesaurus) for

searching a hierarchy of classifications, as well as Rajaraman’s log analyzer, into

Fratkina’s system. As discussed above in paragraphs 41-45, use of such thesaurus

functionality to search synonyms in addition to keywords was well known by that




time, and a PHOSITA would have appreciated that implementing the thesaurus

functionality taught in Rajaraman would have improved the system of Fratkina by

making it more user-friendly and increasing the efficiency with which the user

could navigate through the hierarchical network.

93. Further, I note that Fratkina already expressly teaches that user inputs

can be processed using a “thesaurus” functionality: “Text disambiguation queries

(DAQs): Identified by characteristics of the user’s text, such as misspellings or

ambiguity of words in the text (as defined by an external machine-readable

dictionary, thesaurus, or lexicon, e.g., WordNet), or by a lack of information in the

autocontextualization engine about words in the text.” Fratkina at 7:47-52. And I

note that Fratkina expressly teaches that its dialog engine “is designed to work with

any prior art search/retrieval engines to produce a search space.” Id. at 8:51-60.

These express teachings would have motivated a PHOSITA to combine Fratkina

with the thesaurus functionality of Rajaraman’s search engine. Given that both

Fratkina and Rajaraman relate to searching a hierarchy of nodes and Fratkina

already teaches inclusion of thesaurus functionality, such modifications to the

system in Fratkina would have required only minor modifications in software and

would have yielded predictable results to a PHOSITA. This express “thesaurus”

teaching of Fratkina would have, at a minimum, provided a suggestion or

motivation to a PHOSITA to use synonymous terms, such as taught by



Rajaraman's "special terms file" for Fratkina's "thesaurus" function, such that user

inputs co1Tesponding to synonymous terms would be deemed by the system to be a

match for their assigned keyword(s). Further, a PHOSITA would have also

recognized that the application of methods related to searching keyword synonyms

contained in a thesaurus, which were well known at that time, would have yielded

predictable results when applied in the system of Fratkina and would not render

any features of Fratkina, such as the ability to jump between nodes, inoperable.

Therefore, in my opinion, a person having skill in the art would have been

motivated to combine the teachings of Fratkina and Rajaraman to render claims 3-

6 obvious.

IV. CONCLUSION

94. I declare that all statements made herein of my knowledge are true, and

that all statements made on information and belief are believed to be true, and that

these statements were made with the knowledge that willful false statements and

the like so made are punishable by fine or imprisonment, or both, under Section

1001 of Title 18 of the United States Code.

Date: MtA.Wl 21 2017

By: �� Pad�Smyth, Ph.D.

65 Unified Exhibit 1007 IPR2017-01039

Unified EX1007 Page 65

Documents

IPR2017-01039 Smyth Declaration U.S. Patent 7,231,379 ......informed that Guada Technologies LLC (“Guada”) is the purported owner of the ’379 Patent, and I note that I have no