Interesting People mailing list archives
IP: Bell Labs Launches Web Site For Text-To-Speech Synthesis
From: David Farber <farber () cis upenn edu>
Date: Wed, 05 Mar 1997 22:05:55 -0500
Sort of fun djf FYI: This Comprehensive, Multi-Lingual Interactive TTS Web Site Features German, English, Russian, Chinese, French, Romanian, Italian, and Spanish Demos ___________ Can your PC actually speak your written text? The answer, if you visit the newly-announced Bell Labs Text-to-Speech (TTS) Web site (www.bell-labs.com/project/tts/), is a resounding yes. Bell Labs, the research and development arm of Lucent Technologies -- which has led the industry in speech synthesis research over the past seven decades -- has designed the most comprehensive, interactive multi-lingual TTS site on the Web, which allows users to produce natural speech in several languages (German, French, Mexican, Romanian, Chinese, Russian, English, Italian) directly from written text. "Speech is a sound signal used for language communication. Bell Labs researchers, world experts in signal processing, speech modeling, and text analysis, have devoted decades of work to improving the quality of synthetic speech," said Joe Olive, head of the Language Modeling Research Department at Bell Labs. "The applications for text-to-speech synthesis are various, and growing, and they include e-mail readers, voice-response systems, and automatic order-verification systems, said Olive. "Such systems will require the high-quality word pronunciation and speech intelligibility that Bell Labs TTS systems deliver." The Bell Labs TTS system converts written text to speech through sophisticated linguistic analysis, prosodic modeling, and speech synthesis. Representations of language, typed on a users PC, are enriched by phrasing information, intonation and stress. This information is then generated by means of sophisticated machine synthesis into clearly articulated speech. The Bell Labs TTS system handles not only English, but has expanded its language set to include French, Italian, German, Russian, Japanese, Mandarin and Taiwanese Chinese, Spanish (Peninsular and Mexican), and Romanian. SPEAKING ON THE WEB Visitors to the Bell Labs Text-to-Speech Synthesis Web site at http://www.bell-labs.com/project/tts/, can sample speech in up to nine different languages, as well as visit a demonstration area that allows users to synthesize English, German, and Mandarin Chinese sentences using male, female, and child intonations with effects such as raspiness. The site offers the experience of high-quality interactive, on-the-fly modifications of voice samples. The Bell Labs TTS system even handles German noun compounds, which are notorious for being long and complex, and which cannot be prestored in a dictionary. According to Jens Nagel, an engineer with Mannesmann Autocom in Duesseldorf, Germany, "the (Bell Labs TTS system) is currently the qualitatively best TTS engine for German, in terms of overall intelligibility and especially the pronunciation of names." "Said Michael Tanenblatt, chief designer of the Web site, "the Bell Labs TTS Web site represents our strong feeling that TTS systems have started to play an important role in everyday communications. Users can explore a range of fun, intriguing applications on our site, which will increase the understanding of the role that TTS systems play in bridging communications barriers." Ongoing developments in the Bell Labs Multimedia Communication Research Laboratory include a "talking face" which uses the Bell Labs TTS system to provide a visual personality to a computer. These future developments will be reflected on the Bell Labs TTS Web site. "The average PC user has a new way to explore the future of speech synthesis, as reflected on the Bell Labs TTS site," said James Flanagan, director of the Center for Computer Aids for Industrial Productivity, at Rutgers University, a 1996 National Medal of Science winner and a leading expert in speech communications. "I applaud the efforts of Bell Labs in providing such a stunning resource for speech synthesis on the Web." BELL LABS TTS SYSTEM: HOW IT WORKS Currently, all Bell Labs systems for speech synthesis are concatenative. This means that natural speech segments are selected and stored in an acoustic inventory. These speech segments are units known as diphones, which contain transitions between adjacent phonetic segments. The Bell Labs TTS system determines and selects elements of speech which minimize discrepancies between adjacent segments. During synthesis, the TTS system smoothly concatenates, or links, the stored elements. Lucent Technologies designs, builds and delivers a wide range of public and private networks, communications systems and software, consumer and business telephone systems and microelectronics components. Bell Labs is the research and development arm for the company. It is a worldwide R&D community of nearly 24,000 people. About 20,000 of them work in Lucents business units, developing products, systems and technologies that keep Lucent competitive in its markets. About 4,000 Bell Labs people work in central labs, pursuing research and developing advanced technologies on behalf of all the Lucent operating units. Lucent Technologies was formed as a result of AT&Ts restructuring and became a fully independent company separate from AT&T on September 30, 1996. BELL LABS TEXT-TO-SPEECH SYNTHESIS: THEN AND NOW BELL LABS AND "TALKING MACHINES" Bell Labs first demonstrated an electronic speech synthesis device, the "Voder," developed by H.W. Dudley, at the 1939 Worlds Fair. The New York Times declared, in describing the machines operation, "My God, it talks." This early analog system was the forerunner of Bell Labs work in articulatory synthesis, conducted by Cecil Coker in the 1960s, and Joe Olives work on concatenative synthesis in the 1970s. BELL LABS: WHERE "HAL" FIRST SPOKE One of the more famous moments in Bell Labs synthetic speech research was the sample created by John L. Kelly in 1962. Kellys vocoder synthesizer recreated the song "Bicycle Built for Two" on the ILIAC, with musical accompaniment from Max Mathews. Arthur C. Clarke, then visiting friend and colleague John Pierce at the Bell Labs Murray Hill facility, saw this remarkable demonstration and later used it in the climactic scene of his novel and screenplay for "2001: A Space Odyssey," where the HAL9000 computer sings this song as he is disassembled by astronaut Dave Bowman. Joe Olive, recognized as the leading expert in text-to-speech synthesis, recently contributed a chapter, "The Talking Computer: Text to Speech Synthesis," to the book "HALs Legacy: 2001s Computer as Dream and Reality," (M.I.T. Press, 1996), edited by David Stork. LUCENT PRODUCTS AND THE BELL LABS TTS SYSTEM The Bell Labs TTS system is currently used in the product offerings of several Lucent business units. The Lucent Business Communications Systems Intuity Conversant integrated voice and information processing system uses TTS signal processing cards for applications that include, among others, an e-mail reader. Bell Labs TTS is also integrated in the Lucent Network Systems Speech Processing Solutions offers. Here, TTS allows companies to build applications such as voice dialing, voice-activated response systems, or reservations centers using the AYC speech boards. For information on availability of the PC version of Bell Labs TTS system, contact John Holmgren, business development manager, at 908-949-8864. For more information about Bell Labs TTS, call the Multimedia Communications Research Laboratory at 908-582-6435. For more information about Lucent Speech Processing Solutions, call 800-772-5785. For more information about the Lucent Intuity Conversant system, call 800-247-7000. WORLD WIDE WEB SITE: http://www.bell-labs.com/project/tts/ ________________
Current thread:
- IP: Bell Labs Launches Web Site For Text-To-Speech Synthesis David Farber (Mar 05)