PTAB

IPR2025-00655

Microsoft Corp v. Dialect LLC

Key Events
Petition
petition

1. Case Identification

2. Patent Overview

  • Title: System and Method for Natural Language Understanding
  • Brief Description: The ’160 patent discloses systems and methods for retrieving information and processing commands through a speech interface. The technology provides an integrated environment allowing users to submit natural language queries, which are processed to determine user intent and correct recognition errors, even when inputs are incomplete or ambiguous.

3. Grounds for Unpatentability

Ground 1: Claims 12 and 13 are obvious over Maes in view of Ross

  • Prior Art Relied Upon: Maes (Patent 6,964,023) and Ross (Application # 2002/0133354).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner argued that Maes disclosed the core limitations of a "knowledge-enhanced speech recognition engine" for interpreting natural language utterances, as required by independent claim 12. Maes taught a multi-modal system that performs speech recognition and uses a semantic module to disambiguate a user's intent and correct ambiguities, analogous to determining intent and correcting false recognitions. Maes further disclosed using a "context stack" to store historical interaction data to aid in this process. Petitioner contended that Ross remedied a gap in Maes by teaching the use of specific "context description grammars" for different applications. Ross disclosed comparing a user's transcribed utterance against these grammars to identify the correct context and application. The combination of Maes's intent-disambiguation system with Ross's grammar-based context matching was alleged to render the claims obvious. For claim 13, Petitioner argued Maes disclosed a "multi-pass" speech recognition module by describing an iterative process to determine user intent.
    • Motivation to Combine: A Person of Ordinary Skill in the Art (POSITA) would combine Maes and Ross because they are analogous art from the same assignee (IBM) addressing the same problem of interpreting user utterances. Petitioner asserted it would have been a predictable design choice to incorporate Ross's method of using context grammars to improve the ability of Maes's system to determine target applications. This combination was characterized as a simple arrangement of known elements (a speech recognition system, context stack, and context grammars) to achieve a predictable result—an improved system for interpreting user utterances.
    • Expectation of Success: A POSITA would have a reasonable expectation of success because the combination involved applying conventional, commercially available speech recognition techniques. Modifying Maes's system to include Ross's teachings on context grammars would be a simple implementation predictably resulting in a system with improved accuracy, which was a stated goal of Maes.

Ground 2: Claims 12 and 13 are obvious over Maes, Coffman, and Ross

  • Prior Art Relied Upon: Maes (Patent 6,964,023), Coffman (WO 00/20962), and Ross (Application # 2002/0133354).
  • Core Argument for this Ground:
    • Prior Art Mapping: This ground was presented as an alternative to Ground 1, to be considered if Coffman is not deemed fully incorporated by reference into Maes. The arguments largely mirrored those in Ground 1. Petitioner relied on Coffman's explicit teachings of searching a "context stack" to find entries matching a user's transcribed input. Coffman disclosed comparing new queries against entries down the stack until an acceptable match is found. This teaching was asserted to reinforce the functionality of the context stack disclosed in Maes. The combination of Maes (providing the overall system), Coffman (providing explicit details on context stack searching), and Ross (providing the use of context grammars) allegedly disclosed every limitation of the challenged claims.
    • Motivation to Combine: A POSITA would be motivated to combine Maes and Coffman because Maes expressly incorporated Coffman by reference, they shared a common inventor, and originated from the same assignee (IBM). The motivation to then add Ross's teachings was the same as in Ground 1: to enhance the system’s ability to determine target applications by using a known, grammar-based approach. The references were argued to teach strikingly similar techniques for using a context list or stack, making their combination logical and straightforward.
    • Expectation of Success: The expectation of success was based on the same reasoning as Ground 1. The combination involved analogous art using conventional software and hardware, and the result—improved recognition accuracy—was predictable and desirable.

4. Arguments Regarding Discretionary Denial

  • §325(d) Arguments: Petitioner argued against discretionary denial under 35 U.S.C. §325(d), asserting the primary references, Maes and Ross, were never cited or substantively considered by the examiner during the original prosecution. Although a related application to Coffman was cited, Petitioner contended the examiner materially erred by not considering the specific combination of Maes and Ross, which discloses the claimed "knowledge-enhanced" and "multi-pass" features that the patent owner previously argued were absent from the prior art of record.
  • §314(a) (Fintiv) Arguments: Petitioner argued the Fintiv factors weighed against discretionary denial. Key arguments included: the trial date in the parallel district court litigation is scheduled for January 2026, which is far enough in the future that a Final Written Decision (FWD) in this inter partes review (IPR) would issue close to that date; investment in the court proceeding has been minimal; and Petitioner stipulated that it would not pursue in district court any grounds raised or that reasonably could have been raised in this petition, which should be dispositive. Finally, Petitioner asserted the merits of the petition are particularly strong and compelling.

5. Relief Requested

  • Petitioner requested the institution of an IPR and cancellation of claims 12 and 13 of Patent 7,640,160 as unpatentable.