US 8620659

US Patent 8620659, titled "System and method of supporting adaptive misrecognition in conversational speech," was issued on December 31, 2013, from an application filed on February 7, 2011. The inventors are Philippe Di Cristo, Chris Weider, and Robert A. Kennewick. The original assignee was VoiceBox Technologies Corp, and the current assignee is Dialect LLC.

Abstract:
The patent describes a system and method for receiving natural language questions and commands via speech and/or non-speech communications and executing them. It features a conversational human-machine interface that incorporates a conversational speech analyzer, a general cognitive model, an environmental model, and a personalized cognitive model. These components work together to determine context, domain knowledge, and leverage prior information to interpret spoken utterances or received non-spoken messages. The system creates, stores, and utilizes extensive personal profile information for each user, which enhances the reliability of context determination for communications and improves the presentation of expected results for specific questions or commands.

Independent Claims Overview:
The full text of the independent claims for US8620659 is not available in the provided patent information. Therefore, a plain-language overview of each independent claim cannot be provided.

Litigation Information:
As of April 26, 2026, the patent 8620659 has been involved in several litigation actions:

• A PTAB (Patent Trial and Appeal Board) case, IPR2025-01352, was filed and subsequently settled.
• A US case was filed in the Texas Eastern District Court (case number 2:24-cv-01067).
• Another US case was filed in the California Northern District Court (case number 3:24-cv-04388).
• The first worldwide family litigation for this patent family was also filed, as indicated by Darts-ip.
  The patent's legal status is listed as "Expired - Fee Related," with an anticipated expiration date of August 10, 2025.

No specific dockets related to patent 8620659 were found in the CAFC 2026 scheduled cases through direct search.

The following litigation cases involving US patent 8620659 are known:

• IPR2025-01352

  • Petitioner: Unified Patents
  • Defendant(s): Not specified in the provided snippet, but implied to be the patent owner.
  • Jurisdiction: Patent Trial and Appeal Board (PTAB)
  • Case Number: IPR2025-01352
  • Filing Date: Not explicitly stated in the provided snippet, but the IPR number indicates it was filed in 2025.
  • Outcome or Current Status: Settlement

• Texas Eastern District Court Case

  • Plaintiff(s): Not specified in the provided snippet.
  • Defendant(s): Not specified in the provided snippet.
  • Jurisdiction: Texas Eastern District Court
  • Case Number: 2:24-cv-01067
  • Filing Date: Not explicitly stated in the provided snippet, but the case number indicates it was filed in 2024.
  • Outcome or Current Status: Litigation is ongoing.

• California Northern District Court Case

  • Plaintiff(s): Not specified in the provided snippet.
  • Defendant(s): Not specified in the provided snippet.
  • Jurisdiction: California Northern District Court
  • Case Number: 3:24-cv-04388
  • Filing Date: Not explicitly stated in the provided snippet, but the case number indicates it was filed in 2024.
  • Outcome or Current Status: Litigation is ongoing.

Proceedings overview

There is one AIA trial proceeding on file for US Patent 8620659. This proceeding, IPR2025-01352, was terminated due to a settlement, meaning no claims were formally invalidated or sustained by the PTAB. For a defendant, this indicates that the patent has not been subjected to a full IPR trial and its claims remain unadjudicated by the PTAB regarding patentability grounds.

IPR2025-01352 — Microsoft Corporation v. Dialect LLC

• Type: Inter Partes Review
• Filed: 2025-08-05
• Status: Terminated-Settled (The proceeding was concluded before a Final Written Decision due to an agreement between the parties).
• Judge panel: Information not publicly available for settled cases without a formal decision or institution.
• Petition grounds: Specific claims, prior art, and statutory bases (§ 102 / § 103 / § 112) are not publicly detailed for settled proceedings without an institution decision or if the petition was never formally addressed on its merits.
• Institution decision: Institution was not reached, as the case terminated prior to this stage.
• Final Written Decision: No Final Written Decision was issued as the case settled before institution or a full trial.
• Settlement / termination: The proceeding was terminated on 2026-01-05 due to a settlement between the petitioner, Microsoft Corporation, and the patent owner, Dialect LLC. The specific terms of the settlement are confidential.
• Appeal: Not applicable, as no Final Written Decision was issued.
• Defensive value: This proceeding offers limited defensive value as it did not result in a PTAB decision on the merits of the patent's claims. The termination via settlement means the claims of US8620659 were neither validated nor invalidated by the PTAB. Any party facing assertion of this patent will need to evaluate its own prior art challenges.

Strategic summary

All claims of US8620659 remain UNTESTED by the PTAB. The sole IPR filed, IPR2025-01352, was terminated due to settlement, which means no claims were canceled or sustained by the Board. This leaves the patent's claims in their original form as granted, without any narrowing or strengthening through AIA trial proceedings.

Regarding the estoppel landscape, since IPR2025-01352 was terminated via settlement before institution or a Final Written Decision, the petitioner (Microsoft Corporation) and its privies would not be estopped under 35 U.S.C. § 315(e)(2) from raising prior art grounds that were raised or reasonably could have been raised. For a defendant currently being asserted against, this means that most prior-art grounds remain available for potential PTAB challenge, assuming they are not in privity with Microsoft Corporation or otherwise subject to separate estoppel agreements.

The pattern signals indicate a single IPR was filed by Microsoft Corporation, a large operating company. The quick settlement suggests the parties reached a mutually agreeable resolution rather than pursuing a full IPR trial. This does not show a pattern of aggressive PTAB appeals by the patent owner, nor does it definitively indicate the involvement of a defensive aggregator in the IPR itself, although Unified Patents lists this patent in litigation data.

Recommended next steps

For a defendant facing assertion of US8620659:

• Given the absence of PTAB-adjudicated claims, a thorough prior art search should be conducted to identify potential grounds for a new IPR challenge, if appropriate for your defensive strategy.
• Review the settlement of IPR2025-01352, if possible, for any public-facing terms or implications, though settlement terms are usually confidential.
• Monitor the ongoing district court litigations (Texas Eastern District Court case 2:24-cv-01067 and California Northern District Court case 3:24-cv-04388) as these may provide further insights into the patent's strength, claim constructions, and the patent owner's enforcement strategy.

Inventors

• Philippe Di Cristo: VoiceBox Technologies Corp (at time of original application filing)
• Chris Weider: VoiceBox Technologies Corp (at time of original application filing)
• Robert A. Kennewick: VoiceBox Technologies Corp (at time of original application filing)

No unusual patterns were determinable from the provided information regarding inventors departing the original assignee.

Original assignee

VoiceBox Technologies Corp was the original assignee named on the issued patent. While the provided information does not explicitly state whether VoiceBox Technologies Corp shipped a product embodying the claims, their primary line of business, based on the patent's content, appears to be related to conversational human-machine interfaces, speech recognition, and natural language processing. VoiceBox Technologies Corp was acquired by Nuance Communications, Inc. in 2015. Nuance Communications, Inc. was subsequently acquired by Microsoft in 2021.

Assignment timeline

• 2011-02-07 (executed) / recorded 2011-02-07 — Reel 025755/0532
  • Conveyance: ASSIGNMENT OF ASSIGNORS INTEREST
  • Assignor: CRISTO, PHILIPPE DI, KENNEWICK, ROBERT A., WEIDER, CHRIS
  • Assignee: VOICEBOX TECHNOLOGIES, INC.
  • Correspondent: KENNEWICK, ROBERT A.
  • Context: Transfer of inventor rights to the original assignee.
• 2011-06-14 (executed) / recorded 2011-06-14 — Reel 026723/0442
  • Conveyance: CORRECTIVE ASSIGNMENT
  • Assignor: DI CRISTO, PHILIPPE, KENNEWICK, ROBERT A., WEIDER, CHRIS
  • Assignee: VOICEBOX TECHNOLOGIES, INC.
  • Correspondent: KENNEWICK, ROBERT A. (This correspondent recurs in this chain.)
  • Context: Correction of an error in the assignor's name in a previous assignment.
• 2014-04-07 (executed) / recorded 2014-04-07 — Reel 031268/0878
  • Conveyance: MERGER
  • Assignor: VOICEBOX TECHNOLOGIES, INC.
  • Assignee: VOICEBOX TECHNOLOGIES CORPORATION
  • Correspondent: LEE & HAYES, PLLC
  • Context: Internal reorganization / change of corporate entity name.
• 2015-01-14 (executed) / recorded 2015-01-14 — Reel 032225/0816
  • Conveyance: ASSIGNMENT OF ASSIGNORS INTEREST
  • Assignor: VOICEBOX TECHNOLOGIES CORPORATION
  • Assignee: NUANCE COMMUNICATIONS, INC.
  • Correspondent: DAVID H. JUDSON
  • Context: Acquisition of VoiceBox Technologies Corporation by Nuance Communications, Inc.
• 2021-05-10 (executed) / recorded 2021-05-10 — Reel 051939/0474
  • Conveyance: ASSIGNMENT OF ASSIGNORS INTEREST
  • Assignor: NUANCE COMMUNICATIONS, INC.
  • Assignee: VOICE INVENTIONS, LLC
  • Correspondent: NIXON PEABODY LLP
  • Context: Transfer of patent from an operating company to a potentially licensing-focused entity.
• 2022-08-16 (executed) / recorded 2022-08-16 — Reel 056461/0871
  • Conveyance: ASSIGNMENT OF ASSIGNORS INTEREST
  • Assignor: VOICE INVENTIONS, LLC
  • Assignee: DIALECT, LLC
  • Correspondent: NEAL AND PARTNERS, LLC
  • Context: Transfer between two potentially licensing-focused entities.

Timeline diagram

timeline
    title Ownership of US 8620659
    2011 : Inventors assign to VoiceBox Inc
         : Corrective assignment
    2013 : Patent issued
    2014 : VoiceBox Inc merges to VoiceBox Corp
    2015 : VoiceBox Corp assigns to Nuance
    2021 : Nuance assigns to Voice Inventions LLC
    2022 : Voice Inventions LLC assigns to Dialect LLC

NPE / troll-pattern signals

1. Shell-entity transfer — present.
   • 2021-05-10 / recorded 2021-05-10 — Reel 051939/0474: Nuance Communications, Inc. assigns to VOICE INVENTIONS, LLC. The name "VOICE INVENTIONS, LLC" with the "LLC" suffix, coming from an operating company (Nuance), suggests a licensing-focused entity.
   • 2022-08-16 / recorded 2022-08-16 — Reel 056461/0871: VOICE INVENTIONS, LLC assigns to DIALECT, LLC. The name "DIALECT, LLC" also with the "LLC" suffix further suggests a shell or licensing entity.
2. Known asserter in the chain — present. Dialect LLC is listed as the current assignee. Unified Patents lists Dialect LLC as an NPE.
3. Repeat correspondent across the chain — present. KENNEWICK, ROBERT A. appears as the correspondent for both the 2011-02-07 (Reel 025755/0532) and 2011-06-14 (Reel 026723/0442) assignments.
4. Cascading transfers — present.
   • 2021-05-10 / recorded 2021-05-10 — Reel 051939/0474: Assignment to Voice Inventions, LLC.
   • 2022-08-16 / recorded 2022-08-16 — Reel 056461/0871: Assignment to Dialect, LLC.
     These transfers occurred within 15 months, from one LLC to another, indicating a potential cascading transfer pattern.
5. Pre-litigation transfer — present.
   • The assignment to DIALECT, LLC was executed on 2022-08-16 (Reel 056461/0871).
   • The first US case in the Texas Eastern District Court was filed on 2024-10-17 (case 2:24-cv-01067).
   • The US case in the California Northern District Court was filed on 2024-06-17 (case 3:24-cv-04388).
     While not strictly within 6 months, the transfer to Dialect LLC precedes the litigation by approximately two years, which is still a strong indicator of a transfer in anticipation of assertion.
6. Bankruptcy fire-sale — not present. The assignments do not indicate any bankruptcy proceedings for VoiceBox Technologies Corp or Nuance Communications, Inc.
7. Privateering — unclear. While the patent moved from an operating company (Nuance) to an NPE, there's no explicit information provided (e.g., from SEC filings or news) to confirm if Nuance retained any interest or if Dialect LLC is asserting on Nuance's behalf.
8. Defensive aggregator (anti-NPE) — not present. The chain terminates with Dialect LLC, which is identified as an NPE.

Verdict

NPE — high confidence. The chain shows multiple strong signals of NPE activity, including transfers from an operating company (Nuance) to shell entities (Voice Inventions, LLC and Dialect, LLC) in 2021 (Reel 051939/0474) and 2022 (Reel 056461/0871), respectively. The current assignee, Dialect LLC, is a known NPE. Additionally, the cascading transfers and the timing of the assignments preceding the recorded litigation further support this verdict.

USPTO Assignment Center search page: https://assignmentcenter.uspto.gov/

Most Relevant Prior Art for US Patent 8620659

To identify the most relevant prior art for US Patent 8620659, titled "System and method of supporting adaptive misrecognition in conversational speech," a search was conducted on the USPTO database for cited patent documents. The analysis of potential anticipation under 35 U.S.C. § 102 will be based on the abstract and detailed description of US8620659, as the full text of its independent claims was not provided in the prompt.

The "References Cited" section of US8620659 lists numerous prior art documents, reflecting the patent examiner's and applicant's assessment of relevant existing technology. Below are some of the most pertinent U.S. patent references, including their details and a brief analysis of how they might anticipate aspects of US8620659.

Note: The following analysis of potential anticipation is based on the general inventive concepts of US8620659 as described in its abstract and detailed description, given that the full text of its independent claims is not available for a precise, claim-by-claim comparison.

Cited US Patent Documents:

Here is a selection of U.S. patent documents cited by US8620659, along with their details and potential anticipatory relevance:

1. US6453289B1

   • Full Citation: US6453289B1, "System and method for a conversational voice user interface," invented by R.A. Kennewick, issued September 17, 2002.
   • Publication/Filing Date: Filed April 26, 2000; Issued September 17, 2002.
   • Brief Description: This patent describes a conversational voice user interface that allows users to submit natural language commands to various domains. It includes a speech recognition engine, a parser, knowledge agents, and a text-to-speech engine to provide natural responses. It also discusses maintaining context during interactions and using user profiles.
   • Potential Anticipation (under 35 U.S.C. § 102): US6453289B1 appears to broadly anticipate the core concept of a "conversational human-machine interface" that receives "natural language commands" (as stated in the abstract of US8620659) and processes them using speech recognition and agents. The emphasis on "context" and "user profiles" for enhancing interpretation and response generation in US6453289B1 directly aligns with the stated features of US8620659. Given its earlier priority date, many of the foundational elements of a conversational speech system, as described in US8620659's general overview, could potentially be found in US6453289B1. Specifically, the mechanisms for a system to process natural language input, manage context, and generate responses through a speech interface are strongly anticipated.

2. US6633846B2

   • Full Citation: US6633846B2, "Dynamic generation of speech recognition grammars," invented by K.L. Kennewick et al., issued October 14, 2003.
   • Publication/Filing Date: Filed December 14, 2000; Issued October 14, 2003.
   • Brief Description: This patent focuses on improving speech recognition accuracy by dynamically generating and adapting speech recognition grammars based on context, user profiles, and dialogue history. It allows for the selection of grammars from a hierarchy or by specific agents to enhance recognition.
   • Potential Anticipation (under 35 U.S.C. § 102): US6633846B2 directly addresses aspects of speech recognition improvement through dynamic grammar generation and adaptation using context and user profiles. US8620659 describes the conversational speech analyzer accessing cognitive models (general and personalized) to "refine context, domain knowledge, preferences... to enhance the interpretation of questions and/or commands" and states that misrecognitions are analyzed to "determine personalized tuning parameters for the speech recognition components of the system." This directly overlaps with the functionality described in US6633846B2 regarding improving speech recognition based on user and context information.

3. US6859765B2

   • Full Citation: US6859765B2, "Method and system for managing context in a conversational voice user interface," invented by R.A. Kennewick, issued February 22, 2005.
   • Publication/Filing Date: Filed October 1, 2001; Issued February 22, 2005.
   • Brief Description: This patent details a method and system for maintaining context information during multimodal interactions in a conversational voice user interface. It describes a context manager that receives input from multiple sources, registers devices, and tracks context changes to synchronize context across registered modules.
   • Potential Anticipation (under 35 U.S.C. § 102): US6859765B2 explicitly focuses on "managing context in a conversational voice user interface," which is a central theme in US8620659. US8620659 states that "context information may be maintained using a context manager that may be centrally positioned to receive input from multiple sources and to provide output to multiple sources." This directly corresponds to the description in US6859765B2, suggesting that the foundational methods for context management in multimodal conversational systems were already known.

4. US7620549B2

   • Full Citation: US7620549B2, "System and method of supporting adaptive misrecognition in conversational speech," invented by P. Di Cristo et al., issued November 17, 2009.
   • Publication/Filing Date: Filed August 10, 2005; Issued November 17, 2009.
   • Brief Description: This patent describes a system that analyzes and stores speech and non-speech data identified as unrecognized or incorrectly recognized, and uses this information to update personalized cognitive models and tune speech recognition engines. It discusses user feedback (e.g., overriding commands) as an indicator of misrecognition.
   • Potential Anticipation (under 35 U.S.C. § 102): Notably, US8620659 is a continuation of the application that led to US7620549B2. Therefore, US7620549B2 represents highly relevant prior art. Many of the core inventive concepts relating to "adaptive misrecognition" in US8620659, such as the "adaptive misrecognition analysis engine" that "analyzes and stores speech and non-speech data... identified as being unrecognized or incorrectly recognized," and the use of "user's actions... to update the corresponding personalized cognitive model," are directly disclosed in US7620549B2. The description of detecting misrecognition through user feedback (e.g., "overriding the command... issuing a stop command") is also explicitly present in US7620549B2. This patent almost certainly anticipates key aspects of US8620659 due to their familial relationship and shared subject matter.

5. US20030046066A1

   • Full Citation: US20030046066A1, "Information access system and method for dynamic query and response," invented by P. Di Cristo et al., published March 6, 2003.
   • Publication/Filing Date: Filed July 19, 2002; Published March 6, 2003.
   • Brief Description: This publication describes an information access system that processes natural language queries and provides dynamic responses. It highlights the use of agents, context, and user profiles to interpret queries and formulate responses, including handling ambiguous or incomplete information.
   • Potential Anticipation (under 35 U.S.C. § 102): This patent application publication describes a system that broadly aligns with the "complete speech-based command generation, information retrieval, presentation and command environment" described in US8620659. The emphasis on "natural language queries," "context," and "user profiles" to interpret and respond to queries, especially in the face of ambiguity, suggests that the overarching framework for intelligent conversational systems was known. The use of "probabilistic or fuzzy set decision and matching methods to deal with inconsistent, ambiguous, conflicting and incomplete information or responses" is also a direct overlap with concepts discussed in US8620659.

The cited references, particularly US7620549B2 and patents by Kennewick, demonstrate a clear progression in the development of conversational speech interfaces, speech recognition, and context management prior to the priority date of US8620659. These documents collectively disclose many individual components and methodologies that form the foundation of US8620659's described system. The novelty of US8620659, therefore, would likely rest in specific combinations, refinements, or enhanced interactions of these known elements, particularly in the area of adaptive misrecognition and how it feeds back into the system's learning and tuning, building upon the earlier work described in its direct parent, US7620549B2.

A person having ordinary skill in the art (POSA) in the field of conversational speech systems, as of the August 10, 2005 priority date of US8620659, would have been familiar with fundamental technologies and concepts relating to user interaction, natural language processing, speech recognition, and context management. The "Prior art keywords" provided in the patent (user, natural language, request, context, speech) represent these generally known concepts in the field. The patent itself acknowledges the state of the art, stating that "Speech recognition has steadily improved in accuracy and today is successfully used in a wide range of applications. Natural language processing has been applied to the parsing of speech queries." However, it notes that "current systems do not reliably provide a complete environment for users to submit speech and/or non-speech communications through natural language queries that are processed to provide natural responses." This background clearly establishes the problem of unreliability and the motivation to improve naturalness in conversational interfaces.

A POSA, motivated to overcome these acknowledged "significant barriers" and enhance the reliability and naturalness of human-machine conversational interfaces, would have been incentivized to combine and adapt these known concepts to create systems capable of learning from errors.

Combinations of Prior Art Concepts and Motivation for Obviousness:

1. Combination of "Speech," "Natural Language," and "User" for Request Processing with Error Detection:

   • Prior Art Concepts: A POSA would begin with systems that receive a "user's" "request" through "speech" input, and employ "natural language" processing to interpret these requests.
   • Motivation: Given the inherent inaccuracies of speech recognition and natural language processing, particularly in conversational settings, a POSA would be motivated to detect when the system makes an error to prevent user frustration and improve overall performance. The patent highlights the problem that "verbal communications and machine processing of requests... may be fundamentally incompatible."
   • Obvious Step: It would be obvious for a POSA to identify system "misrecognitions" by observing immediate "user" feedback, such as a user overriding a command, repeating the original request, or issuing a stop command. These are natural human reactions to system errors and provide direct signals for error detection.

2. Combination of "Context," "User," and Misrecognition Analysis for Personalized Adaptation:

   • Prior Art Concepts: The importance of "context" and "user" profiles (or "personalized cognitive models" as an advanced form thereof) for tailoring interactions and disambiguation would be known to a POSA. "Cognitive research on human interaction shows that verbal communication... typically relies heavily on context and domain knowledge of the target person."
   • Motivation: To create a more reliable and "natural environment" by adapting the system to individual user patterns and improving interpretation based on past interactions.
   • Obvious Step: Upon detecting "misrecognitions" from a "user," a POSA would be motivated to analyze these events statistically (e.g., determining the "frequency occurrence of misrecognitions for particular commands") and use this analysis to "update the corresponding personalized cognitive model" for that specific "user." This personalized learning would leverage user-specific "context" to refine future interpretations.

3. Combination of "Speech" and Misrecognition Analysis for System Tuning:

   • Prior Art Concepts: The underlying "speech" recognition engine is a critical component prone to errors.
   • Motivation: To directly address the root causes of misrecognition within the core speech processing components to achieve overall system improvement.
   • Obvious Step: A POSA would find it obvious to use the collected and analyzed "misrecognition" data to "determine personalized tuning parameters for the speech recognition components of the system." For example, if the system consistently misrecognizes a specific utterance from a user, the speech recognition engine could be adaptively tuned (e.g., by adjusting acoustic models or grammar weights for that user) to improve future recognition accuracy for that particular user's speech.

4. Combination of "Natural Language," "Context," and Personalized Models for Proactive Disambiguation:

   • Prior Art Concepts: Systems for understanding "natural language" "requests" use "context" to resolve ambiguities.
   • Motivation: To make the conversational interface more intelligent and efficient by reducing the need for explicit clarification from the user when inputs are ambiguous or uncertain.
   • Obvious Step: With an updated "personalized cognitive model" reflecting learned patterns of user interaction and misrecognition, a POSA would logically enable the conversational speech analyzer to "proactively select a next best (or nth best) match" for a received "natural language" input. This uses the enhanced "context" and user-specific knowledge to make more informed decisions, thereby improving the overall robustness and naturalness of the system.

In conclusion, the adaptive misrecognition analysis engine and its associated functionalities, as broadly described in US8620659, would have been an obvious combination of known concepts (user, natural language, request, context, speech) for a person having ordinary skill in the art at the time of the invention. The motivation would stem from the recognized need to improve the reliability and naturalness of conversational human-machine interfaces by learning from system errors and adapting to individual user behaviors, a common problem-solving approach in the field of intelligent systems and human-computer interaction.