DCT

5:25-cv-07063

Valtrus Innovations Ltd v. Google LLC

Log In
Key Events
Complaint

I. Executive Summary and Procedural Information

  • Parties & Counsel:
  • Case Identification: 4:22-cv-00020, N.D. Tex., 01/10/2022
  • Venue Allegations: Venue is based on Google's regular and established places of business within the Northern District of Texas, including a 260,000 square foot data center in Midlothian, a content distribution network node, and other offices in the Dallas-Fort Worth area where infringing activities are alleged to occur.
  • Core Dispute: Plaintiff alleges that Defendant’s Google Search and Google Cloud platforms infringe six patents related to search result merging and ranking, hardware utilization metering, and server resource allocation.
  • Technical Context: The patents-in-suit relate to foundational technologies for operating large-scale internet search engines and cloud computing infrastructure, addressing challenges of efficiency, relevance, and resource management.
  • Key Procedural History: The complaint alleges Google's pre-suit knowledge of several asserted patents based on citations in Google’s own patent portfolio, Google's employment of the patents’ inventors, and a pre-suit notice letter sent in April 2021 regarding two of the patents. Subsequent to the filing of this complaint, all asserted claims (1-19) of U.S. Patent No. 6,738,764 were cancelled in an Inter Partes Review proceeding (IPR2022-01497) requested by Google, with a certificate issued on September 6, 2024. All claims of U.S. Patent No. 7,523,454 and U.S. Patent No. 7,748,005 were also cancelled in separate IPRs.

Case Timeline

Date Event
1999-10-15 Priority Date for U.S. Patent No. 7,346,604
2001-03-06 Priority Date for U.S. Patent No. 7,523,454
2001-05-08 Priority Date for U.S. Patent No. 6,738,764
2001-08-27 Priority Date for U.S. Patent No. 6,728,704
2002-07-23 Priority Date for U.S. Patent No. 6,816,809
2004-04-27 U.S. Patent No. 6,728,704 Issued
2004-05-18 U.S. Patent No. 6,738,764 Issued
2004-09-10 Priority Date for U.S. Patent No. 7,748,005
2004-11-09 U.S. Patent No. 6,816,809 Issued
2007-05-22 Alleged Knowledge Date for ’764 Patent (via citation)
2008-03-18 U.S. Patent No. 7,346,604 Issued
2009-04-21 U.S. Patent No. 7,523,454 Issued
2010-06-29 U.S. Patent No. 7,748,005 Issued
2012-06-26 Alleged Knowledge Date for ’604 Patent (via citation)
2013-01-01 Alleged Knowledge Date for ’704 Patent (via citation)
2020-05-27 Alleged Knowledge Date for ’454 Patent (via inventor)
2021-04-14 Pre-suit Notice Letter Sent to Google re: ’809 & ’454 Patents
2022-01-10 Complaint Filed

II. Technology and Patent(s)-in-Suit Analysis

U.S. Patent No. 6,728,704 - Method and apparatus for merging result lists from multiple search engines

Issued April 27, 2004

The Invention Explained

  • Problem Addressed: The patent’s background section describes that merging multiple result lists into a single ranked list is computationally intensive because it typically requires examining and ranking every single entry, which can "nullify any advantage gained by operating multiple search engines at the same time" (’704 Patent, col. 2:48-56; Compl. ¶15).
  • The Patented Solution: The invention proposes reducing this computational overhead by analyzing only a subset of entries from each result list to assign a "representative value" (e.g., an average score) to the entire list (’704 Patent, col. 7:12-14). The system then merges the lists based on these representative values, rather than performing a full, item-by-item re-ranking of all entries from all lists combined (’704 Patent, col. 3:20-24; Compl. ¶16).
  • Technical Importance: This approach aimed to enable search systems to gather, rank, and present results from numerous distinct sources more quickly without sacrificing response time (Compl. ¶17).

Key Claims at a Glance

  • The complaint asserts at least independent claim 1 (Compl. ¶18).
  • Claim 1 requires:
    • transmitting a query to a set of search engines;
    • receiving a result list from each search engine;
    • selecting a subset of entries from each result list to form a set of selected entries;
    • assigning a scoring value to each selected entry;
    • assigning to each subset a representative value according to the scoring values assigned to its entries; and
    • producing a merged list of entries in a predetermined manner based on the representative value assigned to each result list, wherein the representative value varies in accordance with the predetermined manner.

U.S. Patent No. 6,738,764 - Apparatus and method for adaptively ranking search results

Issued May 18, 2004

The Invention Explained

  • Problem Addressed: The patent states that it is "impossible to predict a priori" for any given query whether a traditional "static method" of ranking or an "adaptive method" based on user behavior will produce better results, making it desirable to have a technique that can "selectively emphasize" one or the other (’764 Patent, col. 1:51-59; Compl. ¶46-47).
  • The Patented Solution: The invention provides a method to improve search results by creating a "feature vector" for documents that have been viewed by users in the past; this vector "characterizes attributes and query words associated" with that document from prior user interactions (’764 Patent, col. 4:4-10). For a new query, a "similarity score" is calculated between the query and this feature vector. This similarity score is then combined with a standard "relevance score" to produce a final, adaptively corrected rank value (’764 Patent, col. 2:30-35; Compl. ¶47-48).
  • Technical Importance: The method provides a system for blending static relevance signals with dynamic user behavior signals to improve the ranking of search results (Compl. ¶47).

Key Claims at a Glance

  • The complaint asserts at least independent claim 1 (Compl. ¶49).
  • Claim 1 requires:
    • producing a relevance score for a document in view of a query;
    • calculating a similarity score for said query utilizing a feature vector that characterizes attributes and query words of a different query associated with said document; and
    • assigning a rank value for said document based upon said relevance score and said similarity score.
  • The complaint reserves the right to assert additional claims (Compl. ¶49).

U.S. Patent No. 6,816,809 - Hardware based utilization metering

Issued November 9, 2004

  • Technology Synopsis: The patent addresses the challenge of gathering accurate processor utilization data in partitioned computer systems where isolated operating systems cannot easily communicate with each other (Compl. ¶70). The proposed solution is a hardware-based metering method that uses inputs from a "system clock" and an "idle indicator" to directly measure the CPU cycles during which the processor is in a non-idle state, thereby avoiding the need for inter-OS communication software (Compl. ¶71).
  • Asserted Claims: At least independent claim 13 is asserted (Compl. ¶72).
  • Accused Features: The complaint accuses Google Cloud of infringement based on its use of modern CPUs (e.g., AMD EPYC) that include hardware performance counters to determine when a processor core is in a "busy" state (the C0 state), which is allegedly used for measuring processor utilization (Compl. ¶74-77).

U.S. Patent No. 7,346,604 - Method for ranking hypertext search results by analysis of hyperlinks from expert documents and keyword scope

Issued March 18, 2008

  • Technology Synopsis: The patent describes the difficulty of assessing the "authoritativeness" of web pages for broad search queries based on content alone (Compl. ¶96). The patented solution is a method that first identifies a set of "expert documents" and then, in response to a query, ranks "target documents" based on the hyperlinks they receive from those ranked experts, thereby leveraging the collective judgment of the expert community (Compl. ¶97-98).
  • Asserted Claims: At least independent claim 1 is asserted (Compl. ¶99).
  • Accused Features: Google Search is accused of infringement through its alleged use of algorithms like "Hilltop," which reportedly builds a "special index of expert documents" and uses link analysis from those documents to determine authority and rank search results (Compl. ¶104-105, ¶112).

U.S. Patent No. 7,523,454 - Apparatus and method for routing a transaction to a partitioned server

Issued April 21, 2009

  • Technology Synopsis: The patent addresses a deficiency in "current approaches to load balancing" that do not "recognize the partitions of a server and the various configurations thereof" (Compl. ¶123). The invention is a method for routing a transaction to a specific partition on a server based not only on general load, but also on the specific configuration of the partition and characteristics of the transaction itself (Compl. ¶124-125).
  • Asserted Claims: At least independent claim 17 is asserted (Compl. ¶126).
  • Accused Features: The complaint accuses Google Cloud's load balancing feature, which allegedly "redirect[s] traffic away from busy or unavailable" virtual machines (partitions) based on their "health" (a configuration) and on characteristics of the incoming data packet, such as its source IP address (Compl. ¶129, ¶133, ¶135).

U.S. Patent No. 7,748,005 - System and method for allocating a plurality of resources between a plurality of computing domains

Issued June 29, 2010

  • Technology Synopsis: The patent addresses the challenge of allocating limited resources, particularly CPU resources, among different applications in a virtualized environment composed of multiple "computing domains" (Compl. ¶148-149). The patented method improves computer performance by dynamically reallocating these resources between domains "in response to received requests for additional resources according to service level parameters" (Compl. ¶150).
  • Asserted Claims: At least independent claim 8 is asserted (Compl. ¶151).
  • Accused Features: Google's Cloud Run product is accused of infringement because it creates "computing domains" (containers), allocates resources like CPU and memory to them, and "automatically scale[s]" the number of instances in response to performance data (e.g., "rate of incoming requests") and in view of service level parameters (e.g., minimum/maximum instances) (Compl. ¶155, ¶167-169).

III. The Accused Instrumentality

Product Identification

  • The complaint identifies Google Search and Google Cloud as the accused instrumentalities, with specific features of Google Cloud including its load balancing and the Cloud Run service also being implicated (Compl. ¶19, ¶73, ¶127, ¶152).

Functionality and Market Context

  • Google Search: A web search engine that accepts user queries and returns a ranked and merged list of results from various internal sources, such as web pages, maps, and news (Compl. ¶20-21). A screenshot provided in the complaint shows results from distinct categories such as web links, local business listings, and news stories merged into a single page (Compl. ¶21). The complaint alleges Google Search is the "dominant search engine" with approximately 90% market share (Compl. ¶5).
  • Google Cloud: A suite of cloud computing services that provides customers with access to scalable computing resources, including virtual machines powered by a plurality of processors (Compl. ¶73-75). The complaint alleges these services include load balancing features that route traffic to virtual machines based on their status and on characteristics of the transaction (Compl. ¶129, ¶133-135).
  • Cloud Run: A specific Google Cloud product described as a "managed compute platform that enables you to run containers" (Compl. ¶155). It allocates resources such as CPU and memory to these containers and automatically scales the number of active instances "in response to incoming requests" and user-configured service level parameters (Compl. ¶157, ¶168-169).

IV. Analysis of Infringement Allegations

U.S. Patent No. 6,728,704 Infringement Allegations

Claim Element (from Independent Claim 1) Alleged Infringing Functionality Complaint Citation Patent Citation
transmitting a query to a set of search engines; A query to Google Search is transmitted to multiple internal search engines (e.g., for web results, map results, news results). A screenshot shows these distinct result types on a single page (Compl. ¶21). ¶22-23 col. 7:1-3
receiving in response to said query a result list from each search engine... Google Search receives a result list from each of its internal search engines to display the grouped results. ¶24-25 col. 7:4-7
selecting a subset of entries from each result list to form a set of selected entries; Google Search may return millions of results but selects only a small subset for display on the first page. A screenshot highlights "About 675,000,000 results" being found, from which a limited number are shown (Compl. ¶27). ¶26-27 col. 7:8-11
assigning to each selected entry of said set of selected entries a scoring value... Entries are displayed in order of relevance, which the complaint alleges is based on a scoring function. For example, Maps results are sorted based on metrics like user distance (Compl. ¶29). ¶28-29 col. 7:12-14
assigning to each subset a representative value according to the scoring values assigned to its entries Result subsets (e.g., Maps, News) are presented in different orders for different queries, which the complaint alleges is because each subset is assigned a representative value based on the scores of its entries. ¶30-31 col. 7:15-18
producing a merged list of entries in a predetermined manner based on the representative value...wherein the representative value varies... The final merged list displays results in groups, and the ordering of these groups changes based on the query. A screenshot comparing searches for "hospital" and "target" shows different group orderings, which allegedly demonstrates that the representative value varies (Compl. ¶35, p. 14). ¶32-35 col. 8:14-20
  • Identified Points of Contention:
    • Scope Questions: A central question may be whether Google’s internal architecture, which provides different types of content (web, maps, news) that are grouped for display, constitutes a "set of search engines" each returning a "result list" as understood by the patent.
    • Technical Questions: The analysis may focus on whether the complaint provides evidence that Google assigns a single "representative value" to an entire group of results (e.g., all map results) and merges based on that group-level value, as opposed to an alternative process where all individual results are ranked in a unified system and then grouped for presentation.

U.S. Patent No. 6,738,764 Infringement Allegations

Claim Element (from Independent Claim 1) Alleged Infringing Functionality Complaint Citation Patent Citation
producing a relevance score for a document in view of a query; Google's search algorithms "look at many factors," including the "relevance and usability of pages," to rank results. The complaint alleges this process involves assigning a relevance score to each result. ¶53-54 col. 5:5-12
calculating a similarity score for said query utilizing a feature vector that characterizes attributes and query words of a different query associated with said document; Google's BERT technology, a "neural network-based technique," is alleged to extract feature vectors that "consider the full context of a word by looking at the words that come before and after it" to understand user intent, thereby determining similarity based on other queries (Compl. ¶56). ¶55-57 col. 5:13-19
assigning a rank value for said document based upon said relevance score and said similarity score. Google's final ranking is based on "a whole series of algorithms" and "many factors." The complaint alleges that this final rank value is a function of both the relevance score and the similarity score produced by algorithms like BERT. ¶58-59 col. 5:19-22
  • Identified Points of Contention:
    • Scope Questions: The dispute may center on whether the ’764 patent's "feature vector," which is explicitly tied to "query words of a different query associated with said document," can be construed to read on the functionality of a general-purpose natural language processing model like BERT.
    • Technical Questions: A key question will be whether Google's ranking system calculates two distinct scores (a "relevance score" and a "similarity score") and then combines them as required by the claim's structure, or if it employs a single, integrated machine learning model where such signals are not separately calculated and combined.

V. Key Claim Terms for Construction

The Term: "representative value"

(from ’704 Patent, claim 1)

  • Context and Importance: This term is central to the infringement theory of the ’704 patent. The case may turn on whether Google is found to assign a single, aggregate value to an entire list of results (e.g., all Maps results) before merging, or if it ranks all individual results from all sources in a unified manner. Practitioners may focus on this term because it distinguishes the patented method from a simple, brute-force re-ranking of all items.
  • Intrinsic Evidence for Interpretation:
    • Evidence for a Broader Interpretation: The specification states the representative score "may be the arithmetic average or a value proportional to the average for a set of scoring values" (’704 Patent, col. 6:1-3), suggesting flexibility beyond a simple average.
    • Evidence for a Narrower Interpretation: The primary embodiment and figures show a distinct step of calculating an "Avg. Scoring Value" for each list's subset and then using that single value to determine the final merged order, implying the value represents the list as a whole in the merging process (’704 Patent, Fig. 3, items 130, 132, 134).

The Term: "feature vector that characterizes attributes and query words of a different query associated with said document"

(from ’764 Patent, claim 1)

  • Context and Importance: This term defines the adaptive component of the invention. The infringement analysis depends on whether Google's modern NLP systems are found to create vectors that are "associated with" a specific document based on prior, different queries made by users who viewed that document.
  • Intrinsic Evidence for Interpretation:
    • Evidence for a Broader Interpretation: The specification describes the feature vector as capturing "document signatures" and "query words that might capture information about user's behavior and interest," which could be argued to encompass generalized user behavior signals not tied to specific historical queries for a document (’764 Patent, col. 4:40-44).
    • Evidence for a Narrower Interpretation: The detailed description specifies a process of creating a "document-query database, which stores a list of queries associated with a viewed document" (’764 Patent, col. 4:29-31). This language may support a narrower construction requiring a direct, logged association between a document and the specific past queries that led users to view it.

VI. Other Allegations

  • Indirect Infringement: The complaint alleges that Google induces infringement of all asserted patents by offering Google Search and Google Cloud for public use with the intent to encourage and facilitate infringing uses of the patented methods (Compl. ¶37, ¶61, ¶87, ¶114, ¶139).
  • Willful Infringement: The complaint alleges willful infringement for all asserted patents. For the ’704, ’764, and ’604 patents, willfulness is based on alleged pre-suit knowledge dating back to 2007-2013, evidenced by numerous citations to the patents-in-suit in Google's own patent portfolio and Google's employment of several of the named inventors (Compl. ¶36, ¶60, ¶113). For the ’809 and ’454 patents, willfulness is based on alleged knowledge from at least an April 14, 2021 pre-suit notice letter, and for the ’454 patent, also on a Google-assigned European patent naming a co-inventor (Compl. ¶86, ¶138). The complaint does not provide a specific factual basis for pre-suit knowledge of the ’005 patent beyond the general willfulness allegation in the prayer for relief.

VII. Analyst’s Conclusion: Key Questions for the Case

  • Technical Equivalence: A core issue will be one of technical mapping: can the specific, sequential methods described in patents from the early 2000s—such as calculating a single "representative value" for a list (’704 patent) or combining two distinct "relevance" and "similarity" scores (’764 patent)—be shown to be practiced by Google's modern, highly integrated, and AI-driven systems? This raises a fundamental question of whether the accused products operate in a manner that is functionally equivalent to the claimed inventions.
  • Definitional Scope: The case will likely involve significant disputes over claim construction. A key question of definitional scope is whether terms rooted in the context of early search architecture, like "search engines" and "result lists," can be construed to cover the internal content verticals and presentation layers of Google's contemporary search product.
  • Procedural Viability: Given that all asserted claims of the ’764, ’454, and '005 patents were cancelled in post-filing Inter Partes Review proceedings, a dispositive question for a significant portion of the case will be the dismissal of these infringement counts and the resulting impact on the scope of discovery, overall case strategy, and potential damages.