Microsoft Corp v. Dialect LLC

1. Case Identification

• Case #: IPR2025-00659
• Patent #: 9,495,957
• Filed: March 21, 2025
• Petitioner(s): Microsoft Corporation
• Patent Owner(s): Dialect, LLC
• Challenged Claims: 7-8

2. Patent Overview

• Title: Natural Language Human-Machine Interactions
• Brief Description: The ’957 patent discloses a computer-implemented method for processing natural language utterances on mobile devices using a speech interface or a combination speech and non-speech interface. The system generates and uses a "context stack" of prior interactions and user information to improve the reliability of interpreting user commands or questions.

3. Grounds for Unpatentability

Ground 1: Claims 7 and 8 are obvious over Maes.

• Prior Art Relied Upon: Maes (Patent 6,964,023).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner argued that Maes, which discloses a multi-modal conversational computing system, teaches every limitation of claims 7 and 8. Maes’s system processes natural language utterances and uses a "context stack" to store historical interaction data, satisfying the "generating a context stack" limitation of claim 7. Its speech recognition engines determine words from utterances, and its dialog manager searches the context stack to identify context entries corresponding to those words. Petitioner asserted that Maes’s system generates probability or "likelihood scores" for phonemes in a given context, which maps to the claimed "rank scores." For claim 8, Maes was argued to teach prompting a user for more information when intent is ambiguous (e.g., "what device do you want to have turned on?") and receiving a "non-speech input" (e.g., a gesture pointing to a radio) in response to disambiguate the command.

Ground 2: Claims 7 and 8 are obvious over Maes in view of Coffman.

• Prior Art Relied Upon: Maes (Patent 6,964,023) and Coffman (International Publication No. WO 00/20962).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner contended that Maes expressly incorporates Coffman by reference, making its teachings effectively part of Maes. To the extent they are considered separate, the argument largely mirrors Ground 1, with Coffman’s disclosures used to bolster the teachings of Maes. Specifically, Coffman was cited for its detailed description of searching a context stack by comparing transcribed user inputs against stored entries until an acceptable match is found, which Petitioner argued confirms the "identifying" and "comparing" limitations of claim 7.
  • Motivation to Combine: A POSITA would combine Maes and Coffman for several reasons. Maes expressly incorporates Coffman. The references share a common inventor and original assignee (IBM) and address nearly identical problems using similar context stack techniques. Combining Coffman's efficient context stack management (placing the most recent context at the top) with Maes's system would improve performance, a stated goal of Maes.
  • Expectation of Success: A POSITA would have a reasonable expectation of success because the combination involves applying known, compatible techniques from analogous art to a similar system. Both references teach using conventional hardware and software, indicating a straightforward integration without undue experimentation.

Ground 3: Claims 7 and 8 are obvious over Kennewick and Weissman in further view of Maes.

• Prior Art Relied Upon: Kennewick (Application # 2004/0044516), Weissman (Patent 6,453,315), and Maes (Patent 6,964,023).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner first argued that the Patent Owner is collaterally estopped from relitigating the unpatentability of claim 7 over the combination of Kennewick and Weissman. This is based on a final rejection of claim 7 on these references during a prior ex parte reexamination of the ’957 patent. For dependent claim 8, Petitioner argued that Maes supplies the limitations of prompting for additional information and receiving non-speech input in response.
  • Motivation to Combine: A POSITA would combine the teachings of Maes with the system of Kennewick and Weissman to enhance its functionality. Kennewick discloses prompting a user to verify a command if a confidence level is low, and a POSITA would look to a similar system like Maes to implement more advanced, multi-modal prompting and input techniques (e.g., receiving gestures) to predictably improve the reliability of processing user requests.
  • Expectation of Success: The combination would have been a straightforward application of conventional technologies, as Maes provides a known solution to the exact problem of ambiguity resolution faced by the system in Kennewick.

4. Arguments Regarding Discretionary Denial

• §325(d) - Same or Substantially the Same Art: Petitioner argued against discretionary denial under §325(d), asserting that the primary references (Maes and the incorporated Coffman) were never substantively evaluated by the Examiner during prosecution. Though present in a 28-page Information Disclosure Statement with over 100 other references, they were not used as a basis for any rejection. Petitioner contended this failure to consider such highly relevant art constituted a material error by the USPTO.
• §314(a) - Fintiv Factors: Petitioner argued that the Fintiv factors weigh against discretionary denial. Key arguments included that Petitioner Microsoft is not a defendant in the parallel litigation involving the ’957 patent; the trial for the related party is scheduled for January 2026, nearly two years away, and is likely to be delayed; and Petitioner has stipulated that it will not pursue the IPR grounds in district court if the petition is instituted, which weighs strongly against denial.

5. Relief Requested

• Petitioner requests institution of an inter partes review and cancellation of claims 7 and 8 of Patent 9,495,957 as unpatentable.