Google LLC v. Valtrus Innovations Limited

1. Case Identification

• Case #: IPR2022-01497
• Patent #: 6,738,764
• Filed: September 2, 2022
• Petitioner(s): Google LLC
• Patent Owner(s): Valtrus Innovations Limited
• Challenged Claims: 1-19

2. Patent Overview

• Title: Ranking Search Results
• Brief Description: The ’764 patent describes a method for ranking search results by combining two types of scores. It calculates a "relevance score" based on relatively static, time-invariant document characteristics and a separate "similarity score" from an adaptive method that learns from user behavior over time, such as which documents are selected for viewing in response to queries. The two scores are then combined, often using weighting variables, to produce a final rank for a document.

3. Grounds for Unpatentability

Ground 1: Claims 1-19 are obvious over Li

• Prior Art Relied Upon: Li (Patent 7,231,381).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner argued that Li teaches all elements of the challenged claims, but in the context of an image search engine rather than a document search engine. Li discloses a system that generates two scores for images: a "low-level" score based on intrinsic image features (e.g., color, size) and a "high-level" score based on surrounding text on the web page. These scores are linearly combined to rank the images. Li’s system is also adaptive; it uses relevance feedback from users to update its "high-level" feature vectors, incorporating information from prior queries. Petitioner contended that Li’s “low-level score” corresponds to the claimed "relevance score" and its adaptive "high-level score" corresponds to the claimed "similarity score."
  • Motivation to Combine (for §103 grounds): The sole modification needed was to have Li's system present ranked links to the web pages containing the images, rather than just the images themselves. Petitioner asserted a person of ordinary skill in the art (POSITA) would be motivated to do this because search engines at the time conventionally returned lists of hypertext links to web pages, and users would naturally be interested in the source and context of a relevant image.
  • Expectation of Success: A POSITA would have a reasonable expectation of success in modifying Li's user interface to display links to the source web pages, as this was a standard, well-understood technique with predictable results.

Ground 2: Claims 1-19 are obvious over Edlund in view of Li

• Prior Art Relied Upon: Edlund (Patent 6,546,388) and Li (Patent 7,231,381).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner asserted that Edlund discloses nearly all limitations of the claims. Edlund teaches a search engine that combines a "content relevance" score with a "popularity" score derived from user behavior (i.e., which documents users select for a given query). Edlund’s popularity score functions as the claimed "similarity score," as it is based on user interaction with documents in response to different queries. Edlund combines these scores using a weighted linear function to determine a final rank. Petitioner argued that the claimed "feature vector" is disclosed by Edlund's HASH and QamHash data structures, which store lists of prior queries associated with document URLs.
  • Motivation to Combine (for §103 grounds): While Edlund discloses the core combination, Petitioner argued a POSITA would combine Edlund with Li to implement the adaptive feature vector limitations more explicitly. Li teaches a known method for creating and adaptively updating feature vectors based on user feedback to improve similarity scoring. Since Edlund explicitly contemplates using multiple relevance metrics, a POSITA would be motivated to incorporate Li's known technique to provide a more accurate ranking by combining more measures of relevance, including Li's adaptive, keyword-based feature vector approach.
  • Expectation of Success: Combining Li's adaptive scoring technique with Edlund’s ranking framework would be a combination of known methods in the predictable field of search technology, with a high expectation of success.

4. Arguments Regarding Discretionary Denial

• Arguments against §325(d) Denial: Petitioner argued denial under §325(d) is unwarranted. The primary reference for Ground 1, Li, was never presented to or considered by the USPTO during prosecution. The primary reference for Ground 2, Edlund, was listed on an Information Disclosure Statement (IDS) but was never substantively discussed by the Examiner. Petitioner asserted the Examiner erred by allowing the claims based on a finding that the prior art of record failed to teach combining a relevance score and a similarity score, a feature Petitioner claims Edlund teaches explicitly.
• Arguments against §314(a) Denial (Fintiv Factors): Petitioner argued the Fintiv factors weigh in favor of institution. The parallel district court litigation was in its early stages with no trial date set, minimal investment by the parties, and no substantive orders issued. The petition also challenges all 19 claims of the ’764 patent, whereas the patent owner had asserted a smaller subset in the litigation, meaning institution would promote patent system integrity by resolving invalidity for all claims. Finally, Petitioner contended the merits of the petition are compelling, which under USPTO guidance weighs strongly against discretionary denial.

5. Relief Requested

• Petitioner requests the institution of an inter partes review and the cancellation of claims 1-19 of the ’764 patent as unpatentable under 35 U.S.C. §103.