Speech Recognition for Lojban

91 Aufrufe
Direkt zur ersten ungelesenen Nachricht

sorpa'as plat

ungelesen,
31.10.2014, 07:54:1331.10.14
an loj...@googlegroups.com
Hi all! Recently I am trying to make something related to speech recognition for Lojban.

I just found that it is relatively easy to talk to computers in Lojban. I found a great project written by .iocikun.juj. called lojysamban (http://homepage3.nifty.com/salamander/myblog/lojysamban.html) that implemented Prolog-like language in Lojban. I have already heard that someone started to code by voice (https://www.youtube.com/watch?v=8SkdfdXWYaI) using python and a customized version of English, and it would be cool if that can be done by Lojban, since it's a true human language and doesn't need modification. This would really make Lojban come closer to AI.

But first we need a speech recognition tool for Lojban. Any ideas? Has anyone in Lojban community done this before? And any resources related to this?

mu'o mi'e la sorpa'as

v4hn

ungelesen,
15.12.2014, 12:13:0315.12.14
an loj...@googlegroups.com
coi do

> *But first we need a speech recognition tool for Lojban.* Any ideas? Has
> anyone in Lojban community done this before? And any resources related to
> this?

This idea pops up every now and then, yes.
However, I don't know of any really successful approach.

In 2008 there has been a project by a couple of cognitive science students
to implement speech recognition for lojban at the university of Osnabrueck.
After (quite) some nagging they provided me with a folder of (really few)
training data and some scripts they used back then.

It doesn't look like they had a lot of fun with it though:
> We decided to try the impossible:
> Use the speech recognition toolkit of CMU to build a speech recognizer for a non-English language.

Also, it seems a lot has changed w.r.t. open source
speech recognition engines in the last couple of years.
ti'e CMU Sphinx got rewritten and other (potentially capable)
speech recognition engines like julius appeared.

I didn't get around to play with any of these, although I'm still
interested in a working/usable FOSS speech recognition tool for Lojban.

Btw: In theory it is, by design, much more easy to do speech recognition with Lojban
than with most natural human languages because of the audio-visual isomorphism.
It prevents hard questions like "Why is there a 'k' in knight?" .u'i


mi'e la .van. mu'o
Allen antworten
Antwort an Autor
Weiterleiten
0 neue Nachrichten