PTAB

IPR2025-01351

Microsoft Corp v. Dialect LLC

Log In
Key Events
Petition
petition

1. Case Identification

2. Patent Overview

  • Title: Systems and Methods for Responding to Natural Language Speech Utterance
  • Brief Description: The ’209 patent relates to a system for processing natural language speech commands. The system recognizes words and phrases from a user's utterance using dictionary and phrase tables, parses them to determine a context, selects a corresponding domain agent to process the request, and provides results to the user.

3. Grounds for Unpatentability

Ground 1: Obviousness over Khan - Claims 1-2, 4-9, and 12-14 are obvious over Khan.

  • Prior Art Relied Upon: Khan (International Publication No. WO 2001/080096A1).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner argued that Khan discloses a complete, integrated service engine that fulfills user requests submitted in natural language, mirroring the entire process of the ’209 patent. Khan’s system was asserted to perform speech recognition, parse the recognized text to determine meaning and domain (context), select an "intelligent service agent" (the claimed "domain agent") for that domain, format the request for the agent using a domain-specific template (the claimed "grammar"), and present results. Critically, Petitioner contended that Khan explicitly discloses using "domain specific keyword and phrase databases" and "grammars and dictionaries" for recognition, which directly corresponds to the "dictionary and phrase tables" limitation that was key to the ’209 patent's allowance. Petitioner also asserted Khan teaches maintaining a set of probabilities at each processing stage that are dynamically updated to improve performance over time, mapping to the ’209 patent's "maintaining a dynamic set of prior probabilities" limitation.

Ground 2: Obviousness over Khan and Riccardi - Claims 1-2, 4-9, and 12-14 are obvious over Khan in view of Riccardi.

  • Prior Art Relied Upon: Khan (WO 2001/080096A1) and Riccardi (Patent 8,392,188).

  • Core Argument for this Ground:

    • Prior Art Mapping: This ground incorporated all arguments regarding Khan from Ground 1. Petitioner added Riccardi to expressly teach an "adaptive spoken dialog system" that improves speech recognition by transforming a prior probability distribution from one domain to another without transcription ("on-line adaptation"). This was presented as a specific, well-known method for implementing the dynamic probability features Petitioner argued were inherent or obvious in Khan.
    • Motivation to Combine: A POSITA would combine Riccardi's dynamic adaptation with Khan's service engine to improve the accuracy and flexibility of Khan's system. This combination would allow the system to "adapt to unseen speech and language events," a goal that complemented the adaptive learning of user requests already taught by Khan.
    • Expectation of Success: Petitioner argued success was expected because Khan's system was designed to use "generic speech recognition software," making it readily compatible with advanced modules like Riccardi's. Furthermore, POSITAs were generally familiar with methods for updating probabilistic models for natural language processing.

  • Additional Grounds: Petitioner asserted further obviousness challenges by adding Iizuka (Patent 6,424,980) and Hartono (WO 2000/011571A1) to the primary combination. Iizuka was cited to teach specific techniques for "flexibly retrieving necessary information" by extracting data items from varied sources (e.g., HTML documents), mapping to claim limitations regarding extracting values from asynchronously received results. Hartono was cited to teach selecting agent "personalities" and using text-to-speech synthesizers with varied intonation to make responses sound more natural, mapping to limitations on formatting and presenting results to the user.

4. Key Claim Construction Positions

  • Petitioner noted that several challenged claim terms were construed in a district court litigation involving a related patent. While not conceding their correctness, Petitioner argued the challenged claims are obvious even under these constructions:
    • context: "the subject matter area to which a particular user input is directed and which is used to determine the meaning of the user input"
    • domain agent: "executables associated with a specific domain that receive, process and respond to user questions, queries and commands"
    • grammar: "a set of rules governing how inputs and requests to a domain agent should be structured"

5. Relief Requested

  • Petitioner requests institution of an inter partes review and cancellation of claims 1-2, 4-9, and 12-14 of Patent 7,398,209 as unpatentable.