G2P4J Grapheme to Phoneme Forge

* G2P4J

G 2 P 4 J Grapheme to Phoneme Forge

Use our tools to hand edit phonetic word dictionaries for speech recognition engines. The new G2P4J format supporting SAMPA and Kirshenbaum IPA is portable to Sphinx, Julius and others. Demo medical, legal and technical dictionaries are featured.

Warning! This is in early development. Things may suddenly stop working.
Test these scripts on a live DVD boot such as KNOPPIX 3.5.1
which does not write to your drives.

* News

+ 25 August 2009 add file conversion and patch utilities.
See the Sample Screens
+ 9 August 2009 added the h key to show help for phonemes
+ 2 August 2009 developing a process to add the correction 'patch'
to the dictionary. Caution, very little testing has been done.
+ 23 July 2009 early test are comparing 'out of dictionary words'
with phonemes generated by Sequitur G2P and Festival.
+ 19 July 2009 after some documentation our project is accepted at Sourceforge.

* Mission

+Assist open source speech recognition.
+G2P4KIDS Phoneme demonstration for young school children.

* How to run the program

+Required: linux, expect program language, and festival speech.
+Open a shell in an empty folder. Paste the following.
wget http://g2p4j.sourceforge.net/g2p4j
chmod 777 g2p4j
+Download or create your own input word list or dictionary called in.p4j
wget http://g2p4j.sourceforge.net/in.p4j
+Enter g2p4j or ./g2p4j Make sure you have /usr/bin/expect and festival.
Sample Screens

* Description

==== What is G2P4J ?
(c) D. M. Tan Creti 2009
gnu copyleft
free documentation license
(fdl) http://www.gnu.org/copyleft/fdl.txt
http://g2p4j.sourceforge.net

* Author and Links

David M.
Tan Creti

IRC nick paradocs
Public key:

Other Projects:
email

Optimized layout
for the blind
with ELinks and screen readers.

Documents,
Web Pages and
G2P4J Texts
are licensed (CC)
Creative Commons
License

VoxForge
Free Speech
making possible
'Open Source'
speech recognition.
Make a gift
of your voice.

LibriVox
acoustical
liberation
of books
in the
public domain

Hosting courtesy of