Hello,
Last week I posted an inquiry about the availability of automatic
phonetic transcription systems for English. Thanks to everyone who
responded. Here's a brief summary.
First off, there are many text to speech systems out there some of which
may provide transcription indirectly. Also, lots of research has gone on
in these area, but not all systems may be available online. The
following are systems that I've been made aware of that provide text to
phoneme transcription directly and are currently available for use.
Antal van den Bosch informed me of TreeTalk, which does grapheme-phoneme
conversion for English and Dutch. There's an online demo here:
http://ilk.uvt.nl/g2p-www-demo.html
It is a decision-tree-based system and papers are referenced on the
site. Further information is available here:
http://ilk.uvt.nl/~antalb/
It was built using TiMBL (http://ilk.uvt.nl/software.html) and is a part
of NeXTeNS (http://nextens.uvt.nl/), a Dutch text to speech system.
Kevin Lenzo at CMU has a rule-based, decision-tree system called t2p.
http://www-2.cs.cmu.edu/~lenzo/t2p/
I believe it is extendable to other languages, given a pronunciation
dictionary. It's used by several full text to speech systems. It
produces phonemes in the same encoding as the CMU Pronunciation
Dictionary. Papers are referenced on the web site.
The speech group at NIST has a website with several tools including one
by William Fisher called addttp4. It uses a combination of statistics
and rules.
http://www.nist.gov/speech/tools/index.htm
It's an English system and generates Pronlex entries in CMU format. The
algorithm is described in:
"A Statistical Text-to-Phone Function Using Ngrams and Rules", by
William M. Fisher, IEEE International Conference on Acoustics, Speech,
and Signal Processing 1999, pp. II-649-652.
The comp.speech newsgroup FAQ mentions an English system based on the
Naval Research Lab's set of text to phoneme rules and put together by
John A. Wasser.
http://www.speech.cs.cmu.edu/comp.speech/Section5/Synth/text.phoneme.3.h
tml
The links on the FAQ aren't up to date. The system is now available
here:
ftp://svr-ftp.eng.cam.ac.uk/pub/comp.speech/synthesis/english2phoneme.ta
r.gz
The readme is dated 1985, and the files' modification dates are 1994,
but it built OK with gcc 3.3. The phonetic codes it produces are
documented in the source files. I couldn't find out much more
information about this one.
Here's a site that has a CGI transcription system. Additional
information and links on the site.
http://www.foreignword.com/dictionary/truespel/transpel.htm
Also, of interest are dictionaries and wordlists that provide
pronunciation guides. There are several dictionaries available on-line.
However, CMU's Pronunciation Dictionary for English appears to be a
good, machine-readable one. It's included with t2p, mentioned above.
http://www.speech.cs.cmu.edu/cgi-bin/cmudict
Thanks again for the help!
_________________________________
Daniel Walker, Software Engineer
Bowne Global Solutions
Office 5095 Murphy Canyon Road
San Diego, CA 92101 USA
Phone +1 858 737 5247
Mobile +1 619 251 4068
Fax +1 858 737 5297
daniel.walker@bowneglobal.com
www.bowneglobal.com
This archive was generated by hypermail 2b29 : Thu Nov 27 2003 - 03:37:45 MET