Programming environments and Tools

• Programming languages with completed projects: C, C++, Java, JavaScript, perl, awk, {ba,{{t}c}}sh, {my}sql. Pascal, LISP, BASIC, FORTRAN,
• IDE's: emacs, MSVC++, MSVB, {SG,HT,X,VoiceX}ML.
• Operating Systems: Un*xes including Solaris, several Linuxes, FreeBSD, WinXP, Win2000, Win98, WinNT, Win95, Win3.1, DOS 6.2
• Hardware Platforms: Wintel and Unix PC's, Mac's, Sun/SGI workstations
• Relevant API's: Microsoft Win32, MFC, SAPI, TAPI, MAPI; Novell SRAPI. Entropic's TrueTalk TTS API and HTK speech recognition API. Linkon Teravox and Device Driver Interface. Sun XTL telephony API. Some familiarity with messaging/networking protocols.
• IVR systems: designed and implemented high-speed IVR application generator.
• Voice Email: design and implementation currently in progress.
• Lip Synch Machine: built the system.

Theory

• Automatic speech recognition: TV is fully grounded in the fundamentals of HMM training and decoding; can teach it with or without the basic equations. He is a competent user of the HTK HMM Tool Kit, used by its authors to build the most accurate systems in the world for continuous, large vocabulary, speaker independent ASR (cf. ARPA Benchmark Tests). He is capable of building state-of-the-art HMM models, and effective recognition-based systems.
• Text to speech synthesis: TV has deep scientific and broad engineering understanding of text to speech synthesis technology. he wrote the first type-and-talk TTS web page anywhere (Dec. 1994), for Entropic, using AT&T's TTS. (TV's thesis adviser, Mark Liberman, wrote the first version of AT&T's TTS.) TV is competent at optimizing synthesized speech parameters from intonation to phoneme transcriptions. He is an active consultant on large-scale synthesis-based telephone apps (references available on request). He is capable of porting tts engines to new platforms and APIs as well as using stable APIs in new applications. An ongoing development project uses TTS for reading email over the phone.
• Speech science: TV's thesis was titled "English Vowels: Their Surface Phonology and Phonetic Implementation in Vernacular Dialects", and included chapters on acoustics, computational analysis of speech data, and a number of dialects (LA Chicano English, Alabama and Chicago White English, and Jamaican English Creole). TV taught phonetics, speech recognition, social/dialect variation and methodology at Stanford University in 1991-93. He has presented a large number of competitively reviewed as well as invited papers at the Acoustical Society of America, the Linguistic Society of America, the NWAVE sociolinguistics conference, etc. He has some skill in identifying the native geographical dialect of English speakers.
• Other areas: TV has considerable background in:
  • Electrical engineering: linear systems, FFT methods, stochastic signal processing, information theory, Hidden Markov models, &c.
  • Computer science: several programming languages and methodologies, artificial intelligence, computational linguistics (recognition, synthesis, parsing), &c.
  • Statistics: exploratory data analysis, hypothesis testing, a variety of uni- and multi-variate analysis and modeling techniques, statistical natural language processing, &c.
  • Dialog models: discourse analysis (e.g. Polanyi's models)