PTAB

IPR2025-01379

Amazon.com Services LLC v. VB Assets LLC

Log In
Key Events
Petition
petition

1. Case Identification

2. Patent Overview

  • Title: Generating Adapted Natural Language System Responses
  • Brief Description: The ’699 patent discloses computer-implemented methods and systems for generating natural language responses adapted to a user's specific manner of speaking. The system accumulates and utilizes both short-term knowledge from the current conversation and long-term knowledge from user history to identify the user's manner and tailor its response accordingly.

3. Grounds for Unpatentability

Ground 1: Obviousness over Kennewick, Cooper, and Matsuda - Claims 1-22 are obvious over Kennewick, Cooper, and Matsuda.

  • Prior Art Relied Upon: Kennewick (Application # 2004/0193420), Cooper (Patent 6,757,362), and Matsuda (Japanese Application # 2004/334591).
  • Core Argument for this Ground:

    • Prior Art Mapping: Petitioner asserted that the prior art combination teaches every limitation of the challenged claims. Kennewick was presented as the primary reference, disclosing a speech-based interaction system that generates natural language responses using context and extensive user profiles. Petitioner argued Kennewick teaches most elements of independent claims 1 and 12, including receiving a natural language utterance, recognizing words, identifying context, determining an interpretation, and accumulating long-term knowledge in a user-centric profile.

      However, Petitioner contended Kennewick fails to explicitly teach two key aspects supplied by the other references:

      1. Adapting based on a user's manner of speaking: Cooper was introduced to supply this limitation. Cooper describes a "virtual assistant" that adapts its responses based on the user's input, such as by detecting polite discourse or determining the user's emotional state (e.g., calm or angry) from voice volume, word choice, and speech rate. Cooper's system then generates a matching response, such as a polite or submissive one.
      2. Accumulating short-term knowledge during a "predetermined time period": Matsuda was introduced to supply this limitation. Matsuda discloses a conversational system that tracks the subject of a conversation (e.g., a person's name) and maintains this context for a "prescribed time" (e.g., 10 minutes). If the subject is not mentioned again within that period, the context expires. This was argued to be the claimed accumulation of short-term knowledge related to a single conversation within a defined time period.

      Petitioner mapped how the combination addresses other key limitations. For accumulating "long-term knowledge" from utterances received prior to the predetermined time period, Petitioner combined Kennewick's user profile (storing historical interactions) with Matsuda's teaching of storing name data across multiple conversations. For "identifying a manner" based on both short- and long-term knowledge, Petitioner argued a POSITA would combine Cooper's analysis of a current utterance (short-term) with stored emotional state information from past sessions (long-term) and integrate it into Kennewick's framework. Finally, generating a response based on both interpretation and manner was argued to be a straightforward integration of Cooper's manner-based response generation into Kennewick's agent-based interpretation and response system. Dependent claims were argued to be obvious for similar reasons, as they recite conventional features like varying tone/jargon or using grammatical rules, which were also taught by the combination.

    • Motivation to Combine: Petitioner argued a person of ordinary skill in the art (POSITA) would have been motivated to combine the references to create a more effective and human-like conversational agent. A POSITA would combine Cooper with Kennewick to improve the naturalness and appropriateness of system responses, consistent with Kennewick’s stated goal of creating a "natural environment." Incorporating Cooper's ability to adapt to a user's politeness or emotional state into Kennewick's personality-driven system was presented as a predictable improvement. Similarly, a POSITA would incorporate Matsuda's technique for managing conversational subjects within a time period to improve Kennewick's dialog flow and topic-tracking capabilities, making conversations more coherent and natural. The combinations were described as the application of known techniques to a known system to yield predictable results.

    • Expectation of Success: Petitioner asserted a POSITA would have had a reasonable expectation of success. The proposed combination integrates compatible technologies that all serve the shared goal of enabling more adaptive and natural conversational interfaces. The individual components—dialog management, user profiling, and affect recognition—were well-understood fields, and their integration would have been a predictable step toward improving the user experience.

4. Relief Requested

  • Petitioner requests institution of an inter partes review and cancellation of claims 1-22 of Patent 10,755,699 as unpatentable.