Meta Platforms Inc v. Dialect LLC

1. Case Identification

• Case #: IPR2025-01332
• Patent #: 8,447,607
• Filed: July 25, 2025
• Petitioner(s): Meta Platforms, Inc.
• Patent Owner(s): Dialect, LLC
• Challenged Claims: 12 and 13

2. Patent Overview

• Title: Method for Processing Natural Language Inputs
• Brief Description: The ’607 patent discloses a system for processing multi-modal natural language inputs on mobile devices. The technology uses a combination of speech and non-speech interfaces to enable human-machine interactions, allowing users to submit commands or questions across a wide range of domains.

3. Grounds for Unpatentability

Ground 1: Obviousness over Maes - Claims 12 and 13 are obvious over Maes.

• Prior Art Relied Upon: Maes (Patent 6,964,023), which Petitioner argued expressly incorporates by reference Coffman (WO 00/20962) and Ittycheriah (Patent 5,937,383).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner asserted that Maes, when read to include the incorporated teachings of Coffman and Ittycheriah, discloses every limitation of claims 12 and 13. For independent claim 12, Petitioner argued Maes teaches a method for processing multi-modal inputs, including receiving an audio signal (natural language utterance) and a corresponding video signal of the user’s face/lip movements (non-speech input). Maes allegedly generates a non-speech transcription by creating a sequence of visual phonemes (visemes) from the video signal and a speech-based transcription by producing decoded text from the audio signal. Petitioner contended that the “cognitive model” limitation is met by the incorporated teachings of Ittycheriah, which describes a user-specific cache database storing keywords and scores from prior interactions to improve speech recognition. Maes was said to generate a “merged transcription” by aligning or synchronizing the decoded text with the viseme sequence. The system then identifies an entry in a context stack that matches the merged transcription to determine user intent, a process detailed in both Maes and the incorporated Coffman reference. Finally, Maes’s dialog manager determines and launches the appropriate application program (“domain agent”) to execute the user’s request and generate a response. For dependent claim 13, Petitioner argued Maes teaches a second cognitive model based on a plurality of users by describing the creation of language models using multiple speakers, a teaching reinforced by Ittycheriah’s disclosure of a speaker classification approach using data from a class of speakers.
  • Motivation to Combine: Petitioner argued that Coffman and Ittycheriah are effectively part of the Maes disclosure because Maes expressly incorporates them by reference to provide a framework for its context stack and a method for detecting incorrectly recognized words, respectively.
  • Expectation of Success: A POSITA would have a reasonable expectation of success because the incorporated teachings from Coffman and Ittycheriah provide known solutions to enhance the functionality of the primary Maes system within the same technical field of natural language processing.

Ground 2: Obviousness over Maes, Coffman, and Ittycheriah - Claims 12 and 13 are obvious over the combination of Maes, Coffman, and Ittycheriah.

• Prior Art Relied Upon: Maes (Patent 6,964,023), Coffman (WO 00/20962), and Ittycheriah (Patent 5,937,383).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner presented this ground as an alternative, asserting that even if Coffman and Ittycheriah are not considered fully incorporated into Maes, the combination of the three references renders the challenged claims obvious. The mapping of the references to the claim limitations is identical to the mapping articulated in Ground 1.
  • Motivation to Combine: A POSITA would combine the references for several reasons. All three references share a common inventor (Stephane Maes) and a common assignee (IBM), creating a strong motivation to combine them. Technologically, a POSITA would combine Ittycheriah with Maes to improve speech recognition accuracy and system responsiveness by using a cache of user-specific vocabulary from prior interactions. Similarly, a POSITA would integrate Coffman’s teachings on context stack management and merging different input streams (e.g., keyboard and speech) to enhance the multi-modal capabilities and dialog management of the system described in Maes. These combinations represent the pursuit of known goals using prior art elements according to their established functions.
  • Expectation of Success: The combination involved applying known techniques from analogous arts to a known system to achieve predictable improvements in performance and functionality, providing a POSITA with a reasonable expectation of success.

4. Arguments Regarding Discretionary Denial

• Petitioner argued that the Board should not exercise its discretion to deny institution under 35 U.S.C. §314(a). Citing Apple Inc. v. Fintiv, Petitioner contended that institution is warranted based on efficiency, fairness, and the overwhelming strength of the invalidity arguments presented in the petition.

5. Relief Requested

• Petitioner requests institution of an inter partes review (IPR) and cancellation of claims 12 and 13 of the ’607 patent as unpatentable under 35 U.S.C. §103.