Ebates Performance Marketing Inc v. IBM Corp

1. Case Identification

• Case #: IPR2022-00438
• Patent #: 6,785,676
• Filed: February 22, 2022
• Petitioner(s): Ebates Performance Marketing, Inc. DBA Rakuten Rewards
• Patent Owner(s): International Business Machines Corporation
• Challenged Claims: 14-28

2. Patent Overview

• Title: Method for Annotating Resource Results Obtained in a Customer Self Service System
• Brief Description: The ’676 patent describes a customer self-service system for resource search and selection. The system applies an "ordering and annotation function" that maps a "user context vector" with a query response set to generate and present an "annotated response set" to the user.

3. Grounds for Unpatentability

Ground 1: Claims 14, 21, and 22 are obvious over Uchiyama alone or in combination with Wasfi.

• Prior Art Relied Upon: Uchiyama (Application # 2002/0065802) and Wasfi (a 1998 conference proceeding on collaborative filtering).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner argued that Uchiyama, which discloses a personalized search engine that re-orders results based on a user’s profile, taught every limitation of independent claim 14 except for representing the user profile as a "user context vector." However, Uchiyama disclosed using "user preference vectors" to match users with similar interests. A POSITA would have found it obvious to represent Uchiyama's entire user profile as a vector to facilitate comparisons with large data sets from the internet. The combination with Wasfi, which explicitly taught representing user profiles and web pages as vectors and comparing them using a function to find relevant results, rendered this limitation obvious. Uchiyama further disclosed re-ordering search results based on a user's profile, which met the "applying an ordering and annotation function" limitation.
  • Motivation to Combine: A POSITA would combine Wasfi's vector-based comparison technique with Uchiyama's system to achieve Uchiyama's stated goal of providing personalized recommendations from a vast array of HTML pages. Wasfi provided a well-known and efficient method (vector comparison) to implement the personalization described in Uchiyama.
  • Expectation of Success: A POSITA would have a reasonable expectation of success because using vectors for information retrieval was conventional. Implementing Wasfi's established vector comparison function within Uchiyama's system would require only routine programming skills.

Ground 2: Claims 15 and 23 are obvious over the Ground 1 references in view of Lee.

• Prior Art Relied Upon: Uchiyama (Application # 2002/0065802), Wasfi (a 1998 conference proceeding), and Lee (Patent 6,463,428).
• Core Argument for this Ground:
  • Prior Art Mapping: This ground addressed dependent claims 15 and 23, which added the step of "bolding one or more resource results." While Uchiyama disclosed annotating results with a "compatibility gauge," it did not explicitly teach bolding. Lee disclosed a user interface for a database search where more relevant records were depicted with bold lines and bold text compared to less relevant records.
  • Motivation to Combine: A POSITA would be motivated to apply Lee's bolding technique to Uchiyama's system as a simple, common, and space-efficient way to visually indicate relevance to the user. Bolding would be a desirable alternative to Uchiyama's "compatibility gauge," which required additional screen space and could be distracting.
  • Expectation of Success: Modifying Uchiyama's system to present certain results in bold would be a straightforward and predictable adjustment to the program's display code.

Ground 3: Claims 16 and 24 are obvious over the Ground 1 references in view of Littlefield.

• Prior Art Relied Upon: Uchiyama (Application # 2002/0065802), Wasfi (a 1998 conference proceeding), and Littlefield (Patent 6,564,208).
• Core Argument for this Ground:
  • Prior Art Mapping: This ground addressed dependent claims 16 and 24, which added determining primary results for a first display screen and secondary results for a secondary screen. Littlefield taught a common search engine technique of displaying only the top N ranking hits on an initial results page and providing a control (e.g., a button) for the user to access a secondary page with the next N hits.
  • Motivation to Combine: A POSITA would combine Littlefield's display method with Uchiyama's system to solve the known problem of overwhelming a user with too many search results at once. This approach would increase search efficiency by presenting a small, digestible number of the most relevant hits first.
  • Expectation of Success: Implementing a paginated display of results as taught by Littlefield was a conventional technique for web search engines and would have been a routine modification to Uchiyama's system.

Ground 4: Claims 17-20 and 25-28 are obvious over the Ground 1 references in view of Boyan.

• Prior Art Relied Upon: Uchiyama (Application # 2002/0065802), Wasfi (a 1998 conference proceeding), and Boyan (a 1996 workshop paper on machine learning).
• Core Argument for this Ground:
  • Prior Art Mapping: This ground addressed claims requiring a database of user interactions, receiving an "annotation scoring metric," generating an adaptable ordering function, performing generation off-line, and implementing a supervised learning algorithm. Uchiyama disclosed its system could improve over time but did not specify how. Boyan disclosed a machine learning architecture ("LASER") that optimized a search retrieval function by evaluating its performance using a metric based on user feedback (e.g., the rank of documents a user selects). Boyan used a supervised learning algorithm ("simulated annealing") to adjust function parameters off-line to improve performance.
  • Motivation to Combine: A POSITA would be motivated to incorporate Boyan's machine learning methods to implement Uchiyama's stated but undeveloped goal of improving search results over time. Boyan provided the specific back-end operational details for optimizing a search function based on collected user interaction data, which Uchiyama's system already collected.
  • Expectation of Success: A POSITA would have a reasonable expectation of success as Uchiyama's system already disclosed the necessary infrastructure (client-side monitoring, central server database). Implementing Boyan's optimization algorithm would involve routine adjustments to the programming on Uchiyama's central server.

4. Arguments Regarding Discretionary Denial

• Petitioner argued that discretionary denial under 35 U.S.C. §314(a) based on Fintiv factors would be inappropriate. The co-pending district court litigation was in its earliest stages, with trial scheduled nearly two years in the future, ensuring the Patent Trial and Appeal Board’s Final Written Decision (FWD) would issue long before trial. Furthermore, Petitioner stipulated not to assert the unpatentability grounds raised in the inter partes review (IPR) petition in the district court litigation.

5. Relief Requested

• Petitioner requested institution of an IPR and cancellation of claims 14-28 of the ’676 patent as unpatentable.