PTAB

IPR2019-00800

Nuance Communications, Inc. v. MModal Services Ltd.

Log In

1. Case Identification

2. Patent Overview

  • Title: Verification of extracted data
  • Brief Description: The ’040 patent discloses a computer-implemented method for verifying the accuracy of structured codings (e.g., XML) extracted from documents like medical speech transcripts. The method involves rendering text associated with a coding to have a visual characteristic (e.g., boldface) based on a specific "feature" of that coding, without displaying the underlying code itself, to facilitate user review and correction.

3. Grounds for Unpatentability

Ground 1: Obviousness over Boone and Friedman - Claims 1, 2, 6, 7, 11, 20-22, and 24 are obvious over Boone in view of Friedman.

  • Prior Art Relied Upon: Boone (Application # 2004/0243614) and Friedman (Patent 6,182,029).
  • Core Argument for this Ground:

    • Prior Art Mapping: Petitioner asserted that Boone discloses a computer-implemented system for validating extracted medical concepts (e.g., medications, problems) from electronic documents. Boone’s system renders text in a GUI and applies "font effects" such as underlining or highlighting to distinguish text related to an extracted concept. It also includes a user interface for a reviewer to validate the accuracy of extractions, for example, by using a checkbox to mark an extraction as correct or incorrect. However, Boone did not explicitly disclose how the concepts are encoded in a markup language or that the font effects are based on a specific feature of the underlying code.

      Petitioner argued that Friedman remedies these deficiencies. Friedman describes the MedLEE system, which extracts clinical information from natural-language text and encodes it in a structured XML format. Friedman expressly teaches using nested XML tags to represent features, such as encoding a "problem" and its "body location" via nested tags. Crucially, Friedman discloses applying visual characteristics (e.g., highlighting in different colors) to the text based on features of the underlying coding (e.g., using one color for medications and another for allergies) to facilitate manual review, all without rendering the underlying XML tags themselves.

      The combination of Boone's validation framework and GUI with Friedman's feature-rich XML encoding and feature-based highlighting allegedly renders all limitations of independent claim 1 obvious. For dependent claims, Friedman's teaching of nested XML codings (e.g., a bodyloc tag nested within a problem tag) was argued to satisfy the "specified relationship" required by claim 2. Boone's user input mechanisms (e.g., lack of user edits implying correctness) were mapped to claim 11. Boone’s disclosure of using speech recognition to generate the initial text document was mapped to claim 21.

    • Motivation to Combine: A POSITA would combine Boone and Friedman because both address the same problem of improving the processing and verification of medical reports. Petitioner contended that a POSITA implementing Boone’s validation system would have looked to Friedman to incorporate its more sophisticated XML coding and document-tagging schema. Friedman's use of custom tags, nested codings, and feature-based highlighting provides a more efficient and powerful way to encode and review clinical concepts than Boone's general approach. Friedman explicitly provides its techniques to "facilitat[e] manual review" and augment textual documents, which would have been seen as a direct improvement to Boone's validation-focused system.

    • Expectation of Success: A POSITA would have had a reasonable expectation of success in combining the references. Both teach using markup languages for electronic documents, and Friedman's use of XML, a well-known and portable language, would present no technical barriers to integration with a system like Boone's. The combination would have yielded the predictable result of an improved validation system where visual cues are directly and automatically tied to the underlying data structure, as taught by Friedman.

4. Key Claim Construction Positions

  • The Petitioner asserted that the claim term "rendering, by a processor, the first data to have a visual characteristic that is based on the first feature" (claim 1) should be construed to mean "modifying a visual characteristic of the displayed text based on the feature of the underlying coding."
  • This construction was argued to be critical because it links the visual change (e.g., highlighting) directly to a specific "feature" of the data's underlying structure (e.g., an XML tag indicating the concept is an "allergy"). Petitioner supported this construction with statements from the ’040 patent’s specification and prosecution history, where the patentee distinguished prior art by emphasizing that the visual characteristic is based on features of the underlying coding.

5. Relief Requested

  • Petitioner requested the institution of an inter partes review and the cancellation of claims 1, 2, 6, 7, 11, 20, 21, 22, and 24 of the ’040 patent as unpatentable.