PTAB

IPR2025-01365

Amazon.com Services LLC v. VB Assets LLC

Log In
Key Events
Petition
petition

1. Case Identification

2. Patent Overview

  • Title: Facilitating Natural Language System Responses Using Short-Term Knowledge
  • Brief Description: The ’249 patent describes a computer-implemented method for a device to understand and respond to human utterances. The system uses multi-modal inputs, such as a microphone and a touch-screen, and generates responses based on "short-term knowledge" gathered during a user interaction to determine context.

3. Grounds for Unpatentability

Ground 1: Obviousness over Kennewick, Oviatt, and Fujimoto - Claims 1, 6, 9, 12-16, and 21-28 are obvious over Kennewick in view of Oviatt and Fujimoto.

  • Prior Art Relied Upon: Kennewick (Application # 2004/0193420), Oviatt (a 2000 Human-Computer Interaction journal article), and Fujimoto (Japanese Application # 2004145025A).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner argued that the combination of these references teaches every limitation of independent claim 1. Kennewick was asserted to teach a speech-based system that uses "short-term knowledge" (dialog history) from a conversation to facilitate natural language responses. It also discloses a multi-modal device with a microphone array and a touch-screen display. Oviatt was cited for its teaching on interpreting simultaneous "pen-voice" inputs (a non-voice and voice input) and using them to form a description of a user's request, which Petitioner argued is a form of generating short-term knowledge from multi-modal interactions. For limitations requiring the comparison and filtering of sound from two voice inputs, Petitioner relied on Fujimoto, which discloses a voice-matching device using a main microphone and a reference microphone, comparing the signals via a "cross correlation determiner" to remove noise. Petitioner contended that Kennewick’s microphone array provides the two separate voice inputs, which could be processed using Fujimoto's noise-canceling technique. Dependent claims were argued to be obvious extrapolations, such as using recognized words for interpretation (claim 6, taught by Kennewick) and combining short- and long-term knowledge (claim 9, taught by Kennewick's use of user profiles).
    • Motivation to Combine: Petitioner asserted a POSITA would combine Kennewick and Oviatt to improve Kennewick’s system by incorporating Oviatt’s more effective methods for processing multi-modal inputs, which are particularly suited for mobile tasks like navigation—a feature disclosed in Kennewick. The motivation to add Fujimoto’s teachings was to improve speech recognition in Kennewick's multi-microphone system by implementing Fujimoto's known noise-reduction technique to achieve a better signal-to-noise ratio, a stated goal of Kennewick.
    • Expectation of Success: Petitioner argued for a reasonable expectation of success because the combination involved applying known techniques (Oviatt's multi-modal integration, Fujimoto's cross-correlation for noise filtering) to a known, analogous system (Kennewick's speech interface) to achieve the predictable result of improved functionality and performance.

Ground 2: Obviousness over Kennewick, Oviatt, Fujimoto, and Shimomura - Claims 2-5 and 17-20 are obvious over the combination of Ground 1 in view of Shimomura.

  • Prior Art Relied Upon: Kennewick (Application # 2004/0193420), Oviatt (a 2000 Human-Computer Interaction journal article), Fujimoto (Japanese Application # 2004145025A), and Shimomura (Application # 2001/0021909).

  • Core Argument for this Ground:

    • Prior Art Mapping: This ground adds Shimomura to address claims related to processing consecutive utterances and managing context changes. Petitioner argued that Shimomura discloses a robot designed to hold a conversation with a user and explicitly teaches detecting topic changes. It determines whether a second user utterance corresponds to the topic of the first by using a scoring system. This was mapped to the claim limitations of determining whether a second utterance corresponds to a first context, and if not, establishing a second context. Claims 3-5, which introduce a "context stack" for tracking multiple consecutive utterances, were allegedly taught by Kennewick's disclosure of using a context stack combined with Shimomura's teaching of tracking conversation history to manage topic changes.
    • Motivation to Combine: A POSITA would have been motivated to incorporate Shimomura's techniques into the Kennewick system to better "hold natural conversations." Because Kennewick's system relies heavily on context, adding Shimomura's method for detecting and managing context or topic changes between utterances would have been a logical improvement to enhance the system's conversational ability.
    • Expectation of Success: Success would have been expected because the combination applies a known technique for topic management (from Shimomura) to a similar conversational system (Kennewick) to yield the predictable result of more robust dialogue capabilities.

  • Additional Grounds: Petitioner asserted additional obviousness challenges, including:

    • Claims 7, 8, and 22 are obvious over the Ground 1 combination in view of Matsuda (Japanese Application # JP2004334591A), which allegedly teaches expiring short-term knowledge (a "subject person number") after a prescribed time to maintain conversational relevance.
    • Claims 10, 11, 23, and 24 are obvious over the Ground 1 combination in view of Cooper (Patent 6,757,362), which allegedly teaches determining the manner (e.g., tone, pace, politeness) in which an utterance is spoken and adapting the system’s response accordingly.

4. Relief Requested

  • Petitioner requests institution of an inter partes review and cancellation of claims 1-28 of the ’249 patent as unpatentable.