Subject: Ron Cole: Speech Systems and Speech Tools from OGI-CSLU From: Dan Ellis <dpwe(at)ICSI.BERKELEY.EDU> Date: Mon, 7 Apr 1997 22:44:34 PDTDear List - Ron Cole of the Center for Spoken Language Understanding at the Oregon Graduate Institute sent me the attached announcements of some summer courses they are running, and the software toolkit they are making available. DAn. ------- Forwarded Message Date: Mon, 7 Apr 97 21:55 PDT From: cole(at)cse.ogi.edu (Ron Cole) Subject: Short Courses and Speech Tools Dear Colleague, I believe the following short courses and software tools will be of interest to you. Ron Cole, Director Center for Spoken Language Understanding Oregon Graduate Institute ____________________________________________________________________ Spoken Language Systems Short Courses Summer of '97 http://www.cse.ogi.edu/CSLU The Center for Spoken Language Understanding is offering four different laboratory short courses this summer on research and development of spoken language systems. Each course provide hands-on experience using the CSLU Toolkit, a comprehensive set of software tools and technologies for research and development of spoken language systems. The courses are offered on the campus of the Oregon Graduate Institute of Science and Technology near Portland, Oregon, a short drive from the Oregon coast, the Columbia River Gorge and the Cascade Mountains (which offer summer skiing). A brief description of each course is given below. Detailed course descriptions, course dates and information on how to register are available on the CSLU Web site (http://www.cse.ogi.edu/CSLU). Attendance is limited by the number of machines in our new computer teaching laboratory (donated by Intel), so register soon to hold a spot. Building Spoken Dialogue Systems. This course teaches the principles of dialogue design and system development needed to build spoken dialogue systems for a variety of real-world applications. Through a series of exercises, students build spoken dialogue systems of increasing complexity, while learning the fundamentals skills needed to build complex systems. Students demonstrate these skills during a final project, in which they design and demonstrate their own system. (Course materials and students' project descriptions from previous courses are available on the CSLU Web site.) Modern Methods in Text-to-Speech Synthesis. This course, taught by Alan Black and Paul Taylor from the University of Edinburgh and Mike Macon from OGI, uses the Festival TTS system to teach students the principles of TTS, and how to manipulate Festival's system modules to produce natural sounding speech. Language Resources. Research and development of spoken language systems depends critically on the collection and annotation of language resources. Students will learn to design speech corpora to meet the needs of research projects and practical applications, and gain experience in each of the stages of data collection and annotation. Students will conceptualize, design, collect, annotate and document a speech database for their project. Advanced Toolkit Course. This course is intended for researchers and application developers who desire a deeper understanding of the Toolkit architecture and underlying technologies. Participants will learn to manipulate the toolkit to add new features or modify existing ones. The first part of the course covers the software architecture of the toolkit (e.g. how IO works, how to add network services). The second part covers the speech recognition and dialogue technologies used in the toolkit. ____________________________________________________________________ CSLU Toolkit Now Available The Center for Spoken Language Understanding at the Oregon Graduate Institute invites students, researchers and anyone else interested in spoken language technology to use the CSLU Toolkit, software tools and technologies supporting research and development of spoken language systems and human language technologies. The toolkit is designed to support a wide range of research and educational activities. In particular, it includes: * Tools for building and using neural network and HMM speech recognizers for small and medium vocabularies, including pre-built neural network recognizers for digits, alphabet and general English. (Work is underway on large vocabulary continuous speech recognition and these tools and recognizers will be included when ready.) * The Festival TTS system, built at the University of Edinburgh. Festival is a working text-to-speech system and comprehensive environment for research and development of concatenative TTS systems. At present, Festival speaks with a Sottish accent; over time, more voices will be added, the quality of speech will be improved, and it will be better integrated with the rest of the toolkit. * Easy-to-use graphical design tools which allow the user to build and run spoken dialogue systems over the telephone. * A variety of speech processing and visualization tools. * A software infrastructure based on dynamically loading packages into Tcl. There is powerful support built in for distributed computing (e.g. all live mic and phone IO uses network servers). The toolkit can be downloaded and used without charge for non-commercial purposes. Complete sources are included. Feedback and contributions are encouraged. The software should compile on most UNIX platforms (sometimes it takes a bit of effort, but we are working on that). Binary distributions are available for Solaris/SPARC, Solaris/i86, and Digital Unix/Alpha. This summer it will be available for Windows NT/95. For live speech input and output, we support soundblaster and Dialogic cards for Solaris/i86 and Sun audio. We hope to add support for voice modems in the future. For more information, see http://www.cse.ogi.edu/CSLU ------- End of Forwarded Message