Nuance Commc'ns Inc v. Tellme Networks Inc

• Decision Date: 20 April 2010
• Docket Number: Civ. No. 06-105-SLR.
• Citation: 707 F.Supp.2d 472
• Parties: NUANCE COMMUNICATIONS INC. and Phonetic Systems Ltd., Plaintiff,v.TELLME NETWORKS INC., Defendant.
• Court: U.S. District Court — District of Delaware

707 F.Supp.2d 472

NUANCE COMMUNICATIONS INC. and Phonetic Systems Ltd., Plaintiff,v.TELLME NETWORKS INC., Defendant.

Civ. No. 06-105-SLR.

United States District Court D. Delaware.

April 20, 2010.

COPYRIGHT MATERIAL OMITTED

Frederick L. Cottrell, III, Esquire, Gregory Erich Stuhlman, Esquire, Anne Shea Gaza, Esquire, Sarah R. Stafford, Esquire and Steven J. Fineman, Esquire of Richards, Layton & Finger, P.A., Wilmington, DE, Counsel for Plaintiffs. Of Counsel: Jason D. Cassady, Esquire, Luke F. McLeroy, Esquire and Theodore Stevenson, III, Esquire of McKool Smith, Dallas, TX.

Lauren E. Maguire, Esquire, Steven J. Balick, Esquire, John G. Day, Esquire and Tiffany Geyer Lydon, Esquire of Ashby & Geddes, Wilmington, DE, for Defendant. Of Counsel: Steven A. Zalesin, Esquire, Eugene M. Gelernter, Esquire, Chad J. Peterman, Esquire and Sean R. Marshall, Esquire of Patterson Belknap Webb & Tyler LLP, New York, NY.

MEMORANDUM OPINION

SUE L. ROBINSON, District Judge.

I. INTRODUCTION

Plaintiff Nuance Communications Inc. (“Nuance”) is the owner of U.S. Patent No. 5,033,088 (“the '088 patent”), entitled “Method and Apparatus for Effectively Receiving Voice Input to a Voice Recognition System,” which is directed to methods and systems for processing spoken information. Nuance asserts that defendant Tellme Networks, Inc. (“Tellme”) has directly and indirectly infringed, inter alia, the '088 patent through the provision of telephonic directory assistance services. (D.I. 1 at ¶ 8) Tellme denies these allegations and asserts various affirmative defenses, including the noninfringement and invalidity of the '088 patent. (D.I. 34 at ¶¶ 19, 20) The parties have proposed constructions for the disputed claim limitations of the '088 patent. Concurrently pending before the court are Tellme's motions for summary judgment of noninfringement (D.I. 141) and invalidity for anticipation or obviousness (D.I. 139) of the '088 patent. The court has jurisdiction pursuant to 28 U.S.C. §§ 1331 and 1338(a). For the reasons that follow, the court grants in part and denies in part the motions.

II. BACKGROUNDA. The Parties

Nuance is a corporation formed under the laws of the State of Delaware and with its principal place of business in Burlington, Massachusetts. (D.I. 1 at ¶ 2) Nuance is engaged in the business of designing, developing, marketing and providing speech and imaging products and related services. ( Id.) Plaintiff Phonetic Systems Ltd. (“Phonetic”) is a corporation organized under the laws of the nation of Israel, having its principal place of business in Petach Tikva, Israel. ( Id. at ¶ 3) Phonetic is a wholly-owned subsidiary of Nuance. ( Id.)

Tellme is a Delaware corporation having a principal place of business in Mountain View, California. (D.I. 34 at ¶ 4) A wholly-owned subsidiary of Microsoft Corporation, Tellme's business is directed to voice technologies. Specifically, Tellme provides speech recognition services to several commercial providers of directory assistance services.

B. The Prior Art

1. Reliable computerized speech recognition

This dispute concerns the field of computerized speech recognition and, within this field, the application of speech recognition to telephone-based systems. Research in this field began in the 1930s when researchers at AT & T Bell Laboratories focused on using computers to recognize human speech. Because the computerized recognition of speech was inherently less accurate than human recognition, programmers sought to develop computerized systems characterized by more reliable and accurate speech recognition capabilities.

A May 5, 1981 article by two researchers at AT & T Bell Laboratories, entitled “Isolated and Connected Word Recognition-Theory and Selected Applications” (“the Rabiner article”), provides “a tutorial on the concepts and theories underlying modern speech-recognition systems, both practical and experimental.” (D.I. 143, ex. 2 at 621) The Rabiner article describes both the importance and availability of reliable computerized speech recognition. ( Id.) In this regard, the authors emphasize that “[t]he art and science of speech recognition have been advanced to the state where it is now possible to communicate reliably with a computer by speaking to it in a disciplined manner using a vocabulary of moderate size.” ( Id.) The authors further opine that “the power of speech recognition lies in its ability to perform a given task reliably.” ( Id. at 635)

With respect to existing speech recognition systems, the Rabiner article refers to the known practice of incorporating a reliability check. This practice is exemplified by a “canonic pattern-recognition model,” in which the “final decision of what was actually spoken” by the user (i.e., a reliability check) is often handled by a “higher level of processing in the recognition system.” ( Id. at 622) One such recognition model entails the calculation of a distance score between an utterance and a template stored in the system. ( Id. at 638) A “correct” recognition results from the distance score meeting a predetermined threshold. Conversely, an utterance “is not recognized correctly” if the distance score fails to meet this threshold. ( Id.)

The importance and availability of reliable computerized speech recognition is confirmed in a 1983 article by Mark Jones (“the Jones article”). ( Id., ex. 4) The Jones article teaches a reliability check incorporating a high threshold-a system characterized by accurate positive identifications, but also by increased rejections due to recognition failures. ( Id. at 66) According to Jones, the threshold for a given system is a relative value and can “differ with varying environment, speakers, and applications.” ( Id.)

2. The Hitachi patent

The Hitachi patent, entitled “System for Receiving Extension Connection Information,” is a Japanese patent application that was published on July 22, 1981. ( Id., ex. 2 at 6091) The Hitachi patent concerns an automated system directed to the task of connecting telephone callers with extension numbers that they verbally request.¹ ( Id. at 6113)

The parties do not dispute that the invention described by the Hitachi patent functions in the following manner: (1) the system receives and answers a call; (2) the system prompts the caller to provide the name or extension number of the party with whom the caller wishes to speak; (3) the system records the information provided by the caller and attempts to automatically recognize the caller's speech; (4) the caller's speech is compared to a group of stored patterns representing each of the possible responses; (5) if the system is able to “optimally correlate” the caller's speech, it automatically routes the caller to the requested extension number; ² (6) if the system is not able to “optimally correlate” the caller's speech, it re-prompts the caller; and (7) if a recognition failure occurs (i.e., the system is unable to subsequently “optimally correlate” the caller's speech), the system plays the recording for an operator, who assists in connecting the call by entering the extension number with a keypad. ( Id. at 6116-20)

C. The '088 Patent

The invention of the '088 patent arose from the efforts of sole inventor David Shipman (“Shipman”) to interface the innovations in speech recognition technologies with telephone-based applications. (D.I. 153 at 34) Initially, Shipman sought to avoid the need for human operators. However, the telephone-based speech recognition applications contemplated by Shipman encountered many purported unique challenges, including the background noise and poor audio quality associated with a telephone transmission, as well as the practical inability to train the recognition system to a single user's voice. ( Id.) Despite Shipman's efforts to address these issues through an enhanced automated speech recognition algorithm, the reliability of a fully-automated system could not contend with the superior speech recognition capabilities of a human operator. ( Id. at 35) Consequently, Shipman's fully-automated products were not a viable alternative to human operators for certain applications in which recognition mistakes could not be tolerated, to wit, the capture of credit card numbers or prescription information.

Instead of continuing his efforts to devise a fully-automated solution, Shipman sought to bridge the gap in reliable speech recognition through the use of his “invisible corrector.” ( Id.) The “invisible corrector,” Shipman's internal designation for the invention of the '088 patent, refers to a two-path approach in which the speech recognition system will attempt to recognize spoken information through an automated process and, if not reliably recognized, seek assistance from a human attendant. ('088 patent at col. 1; 37-52) If the system recognizes the spoken information with sufficient reliability, the system will automatically complete its task (such as a directory lookup) using the recognized information. ( Id. at col. 2:41-44) If the system does not recognize the spoken information with sufficient reliability, it will silently transmit a recording of the spoken information to a human attendant.³ ( Id. at col. 2:57-64) The human attendant can then confirm or correct the system's automated recognition, allowing the system to complete the requested task. ( Id.)

Shipman alleges that this configuration results in several improvements over the prior art. The dual-path approach purportedly allows for telephonic speech recognition that rivals purely manual systems, while simultaneously cutting down on the number of operators, as well as the salary and equipment costs associated with large operator banks. Likewise, it obviates the need for multiple user confirmations present in fully automated systems.

During prosecution, the examiner rejected the application leading to the ' 088 patent as rendered obvious by U.S. Patent No. 4,618,984 to Das et al. (“the Das patent”), which disclosed, inter alia, “determining...