DCT
1:17-cv-10054
Word To Info Inc v. Nuance Communications Inc
I. Executive Summary and Procedural Information
- Parties & Counsel:
- Plaintiff: Word to Info, Inc. (Texas)
- Defendant: Nuance Communications, Inc. (Delaware, with a principal place of business in Massachusetts)
- Plaintiff’s Counsel: GARDNER & ROSENBERG P.C.; FARNEY DANIELS PC
- Case Identification: 1:17-cv-10054, D. Mass., 01/11/2017
- Venue Allegations: Venue is alleged to be proper in the District of Massachusetts because Defendant's principal place of business is located there, and it allegedly conducts substantial business and has committed acts of infringement within the district.
- Core Dispute: Plaintiff alleges that Defendant’s natural language understanding products and services, including its LinkBase ontology and Dragon Medical software suite, infringe a family of seven U.S. patents related to systems for storing and retrieving knowledge from natural language.
- Technical Context: The technology concerns advanced natural language processing systems that can parse, understand, and store information from human language to build a retrievable knowledge base, a foundational technology for voice recognition, dictation software, and artificial intelligence.
- Key Procedural History: The complaint alleges that at least one of the patents-in-suit was cited during the prosecution of U.S. patents assigned to Defendant Nuance. This allegation may be used to support claims of pre-suit knowledge of the patented technology.
Case Timeline
| Date | Event |
|---|---|
| 1994-09-30 | Earliest Priority Date for all Patents-in-Suit |
| 1998-02-03 | U.S. Patent No. 5,715,468 Issued |
| 2000-10-24 | U.S. Patent No. 6,138,087 Issued |
| 2003-08-19 | U.S. Patent No. 6,609,091 Issued |
| 2008-03-25 | U.S. Patent No. 7,349,840 Issued |
| 2011-01-18 | U.S. Patent No. 7,873,509 Issued |
| 2012-12-04 | U.S. Patent No. 8,326,603 Issued |
| 2014-04-01 | U.S. Patent No. 8,688,436 Issued |
| 2017-01-11 | Complaint Filed |
II. Technology and Patent(s)-in-Suit Analysis
U.S. Patent No. 5,715,468
- Patent Identification: U.S. Patent No. 5715468, “Memory System for Storing and Retrieving Experience and Knowledge with Natural Language,” issued February 3, 1998.
- The Invention Explained:
- Problem Addressed: The patent's background section identifies limitations in prior art natural language processing systems, including their inability to process multiple natural language sentences, expand their knowledge base through conversation, or resolve ambiguities effectively, particularly when selecting the correct meaning ('word sense') of a word ('468 Patent, col. 1:15-2:67).
- The Patented Solution: The invention discloses a comprehensive system that processes natural language through syntactic and semantic methods. It converts language into "electronically encoded data" using data structures like "word sense numbers" and "state representation data," which are stored in a complex memory architecture. This architecture, depicted in Figure 1, includes distinct memory components for concrete nouns, abstract nouns, and context, allowing the system to build a retrievable base of "experience and knowledge" from the language inputs ('468 Patent, Abstract; col. 4:26-5:4).
- Technical Importance: The described approach represents a method for moving beyond simple, command-based language recognition toward more sophisticated, context-aware systems capable of learning from ongoing interactions.
- Key Claims at a Glance:
- The complaint asserts independent claims 1 and 21, among others (Compl. ¶18).
- Claim 1 recites a method of processing natural language with the following essential elements:
- Providing electronically encoded data which is representative of said natural language;
- Providing a dictionary data base having a plurality of entries, where entries are comprised of syntax usage data, associated word sense numbers, and/or function codes;
- Lexically processing the encoded data to access the dictionary data base;
- Providing a grammar specification;
- Utilizing syntax usage data from dictionary entries associated with words of the natural language with reference to the grammar specification to produce output data representative of a grammatical parse of the natural language, with the output data including selected syntax usage data.
U.S. Patent No. 6,138,087
- Patent Identification: U.S. Patent No. 6138087, “Memory System for Storing and Retrieving Experience and Knowledge with Natural Language Utilizing State Representation Data, Word Sense Numbers, Function Codes and/or Directed Graphs,” issued October 24, 2000.
- The Invention Explained:
- Problem Addressed: As a continuation of the application leading to the '468 Patent, this patent builds on the same foundational problem of enabling computers to understand and learn from natural language. It focuses on the challenge of accurately selecting the correct "word sense" from multiple possibilities based on context and stored knowledge ('087 Patent, col. 3:20-33).
- The Patented Solution: The '087 Patent further refines the memory and processing system by explicitly introducing the use of "state representation data" associated with word sense numbers and the organization of knowledge into "directed graphs" ('087 Patent, Abstract). This graphical structure allows the system to represent and traverse relationships between concepts, with access to different paths in the graph determined by "access conditions," thereby improving the ability to disambiguate meaning and interpret the purpose of a natural language input ('087 Patent, col. 12:15-13:21).
- Technical Importance: The introduction of directed graphs and state representation data provided a more structured and scalable way to organize complex knowledge, a key step in developing more powerful natural language understanding engines.
- Key Claims at a Glance:
- The complaint asserts independent claim 17, among others (Compl. ¶27).
- Claim 17 recites a method of processing natural language with the following essential elements:
- Providing electronically encoded data which is representative of said natural language;
- Providing a dictionary data base containing a plurality of entries having syntax usage data and associated word sense numbers;
- Lexically processing said electronically encoded data to access said dictionary data base;
- Utilizing said syntax usage data and word sense numbers from entries of said dictionary data base which are associated with words of said natural language with reference to associated state representation data to select a word sense number for words of said natural language.
U.S. Patent No. 6,609,091
- Patent Identification: U.S. Patent No. 6609091, “Memory System for Storing and Retrieving Experience and Knowledge with Natural Language Utilizing State Representation Data, Word Sense Numbers, Function Codes and/or Directed Graphs,” issued August 19, 2003.
- Technology Synopsis: This patent continues to build on the core system, adding a focus on using a "database of requirements" to govern the selection of word sense numbers. The system must meet these requirements, which are based on the state representation data, in order to select a particular word sense, introducing a more rule-based constraint on the interpretation process ('091 Patent, Abstract).
- Asserted Claims: Independent claims 1 and 12 (Compl. ¶37).
- Accused Features: The complaint alleges that Nuance's LinkBase ontology and associated products and services infringe by providing a database where entries must meet "relationship requirements" to be selected (Compl. ¶¶ 37, 43).
U.S. Patent No. 7,349,840
- Patent Identification: U.S. Patent No. 7349840, “Memory System for Storing and Retrieving Experience and Knowledge with Natural Language Utilizing State Representation Data, Word Sense Numbers, Function Codes, Directed Graphs and/or Context Memory,” issued March 25, 2008.
- Technology Synopsis: This patent introduces the concept of a "context data base" as a distinct element. The system uses this context database, in addition to the grammar and dictionary, to select the appropriate word sense numbers. This emphasizes the importance of the conversational or situational context in disambiguating language ('840 Patent, Abstract).
- Asserted Claims: Independent claim 1 (Compl. ¶47).
- Accused Features: The complaint alleges infringement by Nuance's products, which are said to use "context recognition" and a "context data base" to recognize words based on the surrounding context (Compl. ¶¶ 47, 53).
U.S. Patent No. 7,873,509
- Patent Identification: U.S. Patent No. 7873509, “Memory System for Storing and Retrieving Experience and Knowledge with Natural Language Utilizing State Representation Data, Word Sense Numbers, Function Codes, Directed Graphs, Context Memory, and/or Purpose Relations,” issued January 18, 2011.
- Technology Synopsis: This patent adds the concept of "purpose relations" to the system. It describes an "experience and knowledge database" using directed graphs where nodes have "access conditions." The system performs "relation path identification processing" to find paths through this graph that satisfy the access conditions, allowing it to infer the purpose or intent behind a natural language clause ('509 Patent, Abstract).
- Asserted Claims: Independent claims 9 and 16 (Compl. ¶57).
- Accused Features: Nuance's LinkBase is accused of infringing by utilizing directed graphs with paths and conditions for accessing each path and performing "relation path identification processing" to find paths from nodes (Compl. ¶¶ 57-59).
U.S. Patent No. 8,326,603
- Patent Identification: U.S. Patent No. 8326603, “Memory System for Storing and Retrieving Experience and Knowledge with Natural Language Queries,” issued December 4, 2012.
- Technology Synopsis: This patent appears to focus on applying the underlying system architecture to handle "natural language queries." It reiterates the use of a knowledge database with directed graphs, nodes, and access conditions to identify paths in response to a query, thereby retrieving stored experience and knowledge ('603 Patent, Abstract).
- Asserted Claims: Independent claims 14 and 16 (Compl. ¶64).
- Accused Features: The accused features are consistent with the other patents, focusing on LinkBase's alleged use of directed graphs with nodes organized into paths that have access conditions for determining accessibility (Compl. ¶¶ 64, 66).
U.S. Patent No. 8,688,436
- Patent Identification: U.S. Patent No. 8688436, “Memory System for Storing and Retrieving Experience and Knowledge by Utilizing Natural Language Responses,” issued April 1, 2014.
- Technology Synopsis: This patent adds a "natural language plausibility and expectedness processor." This component is used to evaluate and select from among different possible interpretations or responses based on stored knowledge and context. The system uses this processor to initiate access to the dictionary database ('436 Patent, Abstract).
- Asserted Claims: Independent claims 1 and 7 (Compl. ¶70).
- Accused Features: Nuance's systems are accused of infringing by providing a "natural language plausibility and expectedness processor." The complaint points to functionality like providing alternative choices or "typeahead" suggestions as evidence of this processor (Compl. ¶¶ 70, 74-75).
III. The Accused Instrumentality
- Product Identification: The accused instrumentalities are Defendant Nuance's "LinkBase ontology" and the software applications and services that use data from it (Compl. ¶18). These include Nuance's Dragon Medical suite of products, Nuance's 360 Development Platform, and its clinical language understanding (CLU) services (Compl. ¶18).
- Functionality and Market Context:
- The complaint alleges that the accused products provide natural language understanding capabilities. The core accused component, LinkBase, is described as an "ontology" and a database for natural language processing acquired by Nuance (Compl. ¶20). It allegedly contains over 5 million "knowledge entries" including concepts, relationships, and synonymous terms (Compl. ¶20).
- Functionally, the accused services are alleged to encode natural language inputs (speech or text) into electronic data (Compl. ¶19). They then allegedly use "grammatical analysis" and a "GUM-based (Generalized Upper Model) grammatical specification" to convert the input into a structured format to access the relationships stored in LinkBase, ultimately producing a structured output like XML or HTML (Compl. ¶¶ 22, 23, 24). The complaint alleges these services are used in the commercially significant healthcare technology and clinical language understanding markets (Compl. ¶¶ 18, 24).
- No probative visual evidence provided in complaint.
IV. Analysis of Infringement Allegations
'468 Patent Infringement Allegations
| Claim Element (from Independent Claim 1) | Alleged Infringing Functionality | Complaint Citation | Patent Citation |
|---|---|---|---|
| A method of processing natural language, which comprises steps: providing electronically encoded data which is representative of said natural language; | Nuance software encodes natural language inputs into audio or text files. | ¶19 | col. 366:1-3 |
| providing a dictionary data base wherein said dictionary data base contains a plurality of entries which are comprised of one or more of syntax usage data, associated word sense numbers, and/or function codes; | Nuance's LinkBase is alleged to be a dictionary database containing over 5 million entries, including concepts, relationships, word sense numbers, and synonyms (syntax usage data). | ¶¶20, 21 | col. 367:4-9 |
| lexically processing said electronically encoded data to access said dictionary data base; | Nuance software allegedly lexically processes the encoded data to access the LinkBase dictionary database, for example, by utilizing grammatical analysis. | ¶22 | col. 367:10-12 |
| providing a grammar specification; | LinkBase is alleged to use a GUM-based (Generalized Upper Model) grammatical specification. | ¶23 | col. 367:13-14 |
| utilizing said syntax usage data which are from entries of said dictionary data base and which are associated with words of said natural language with reference to said grammar specification to produce output data representative of a grammatical parse of said natural language, said output data including selected syntax usage. | Nuance software allegedly utilizes syntax usage data from LinkBase with reference to its grammar to perform syntactic parsing and deduce concepts, creating output data such as XML or HTML. | ¶24 | col. 367:15-22 |
'087 Patent Infringement Allegations
| Claim Element (from Independent Claim 17) | Alleged Infringing Functionality | Complaint Citation | Patent Citation |
|---|---|---|---|
| A method of processing natural language, which comprises steps: providing electronically encoded data which is representative of said natural language; | Nuance software encodes natural language speech or text into electronic files. | ¶28 | col. 368:32-34 |
| providing a dictionary data base containing a plurality of entries having associated word sense numbers and having associated state representation data; | LinkBase is alleged to be a dictionary database that includes entries with associated word sense numbers and associated state representation data, including medical codes for medical states. | ¶29 | col. 368:35-40 |
| lexically processing said electronically encoded data to access said dictionary data base; | Nuance software is alleged to lexically process the encoded data to access LinkBase, using grammatical analysis to convert an encoded sentence into a structure. | ¶32 | col. 368:41-43 |
| utilizing said syntax usage data and said word sense numbers which are from entries of said dictionary data base and which are associated with words of said natural language with reference to associated state representation data to select a word sense number for words of said natural language. | Nuance software allegedly utilizes syntax usage data and word sense numbers from LinkBase with reference to state representation data to select a word sense number for words. An exemplary workflow is cited where tokens are mapped to the ontology and syntactic parsing is performed. | ¶34 | col. 368:44-51 |
- Identified Points of Contention:
- Scope Questions: The complaint's allegations for each patent are very similar and rely on high-level descriptions of the accused products. A central issue may be whether Nuance’s "ontology" and "grammatical analysis," as described in public-facing documents, fall within the scope of the patent's more specific architectural elements like "dictionary data base," "grammar specification," and "lexically processing." The defense may argue that its modern, ontology-based system operates in a fundamentally different way than the specific, multi-step procedural methods disclosed in the patents.
- Technical Questions: The complaint repeatedly alleges that LinkBase includes "word sense numbers" and "state representation data." A key technical question for the court will be whether the data structures used by Nuance to represent word meanings and medical concepts are structurally and functionally equivalent to the specific data formats for "word sense numbers" and "state representation data" described in the patents' detailed specifications (e.g., '468 Patent, Fig. 17A).
V. Key Claim Terms for Construction
The Term: "word sense number" (appears in claims of '468, '087, and subsequent patents)
Context and Importance: This term is a foundational data structure in the claimed inventions, used to disambiguate the meaning of words. The outcome of the case may depend on whether Nuance's method for representing different meanings of a word is found to be a "word sense number" as claimed. Practitioners may focus on this term because the patent specification provides a highly detailed, multi-component definition that may be narrower than a generic concept of word disambiguation.
Intrinsic Evidence for Interpretation:
- Evidence for a Broader Interpretation: The claims themselves do not define the structure of the "word sense number," referring to it simply as a number associated with a word in a dictionary entry that represents a particular meaning (e.g., '468 Patent, col. 367:6-9).
- Evidence for a Narrower Interpretation: The specification provides detailed formats for word sense numbers, describing them as being composed of specific components such as a "Class Number," "Member number," "Type Number," "Specificity Number," and "Experience Number" ('468 Patent, col. 147:1-68; Fig. 17A). Nuance may argue that to be a "word sense number," an accused data structure must contain these specific components.
The Term: "lexically processing" ('468 Patent, Claim 1)
Context and Importance: This term defines the core action of using the encoded language to access the dictionary. The dispute will likely focus on whether the general "grammatical analysis" alleged in the complaint (Compl. ¶22) constitutes "lexically processing."
Intrinsic Evidence for Interpretation:
- Evidence for a Broader Interpretation: A plaintiff could argue that "lexically processing" should be given its plain and ordinary meaning, covering any computational step that analyzes words or morphemes to prepare for a database lookup.
- Evidence for a Narrower Interpretation: The specification describes a specific sequence of steps for processing, including "Word Isolation Step 12," "Dictionary Look Up Step 14," and "Syntax Parse Step 16" ('468 Patent, Fig. 2; col. 26:65-35:1). A defendant may argue that "lexically processing" should be construed to require these, or analogous, distinct steps, rather than any generic form of parsing.
VI. Other Allegations
- Indirect Infringement: The complaint alleges only direct infringement under 35 U.S.C. § 271(a) for each asserted patent (Compl. ¶¶ 18, 27, 37, 47, 57, 64, 70).
- Willful Infringement: The complaint seeks a finding of willful infringement in its prayer for relief (Compl. p. 23, ¶B). The factual basis for this appears to be the allegation that Nuance had pre-suit knowledge of the patent family because at least one of the patents-in-suit was cited during the prosecution of patents assigned to Nuance (Compl. ¶9).
VII. Analyst’s Conclusion: Key Questions for the Case
- A primary issue will be one of architectural correspondence: do the accused Nuance products, particularly the "LinkBase ontology," implement the specific multi-part memory and processing architecture (e.g., distinct dictionary, grammar, and processing modules) recited in the claims, or do they utilize a more integrated or fundamentally different system for natural language understanding? The complaint's reliance on high-level, public descriptions of the accused products may create an evidentiary challenge in demonstrating a feature-by-feature mapping to the detailed systems disclosed in the patents.
- The case will also turn on a question of definitional scope during claim construction. Will core technical terms like "word sense number" and "state representation data" be interpreted broadly to cover any data structure that serves to disambiguate meaning, or will they be limited to the specific, multi-component data formats detailed extensively in the patent specifications? The resolution of this question will likely determine whether the functionality of the accused products can be found to meet these limitations.