[
https://issues.apache.org/jira/browse/TIKA-94?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12775155#action_12775155 ]
David Woollard commented on TIKA-94:
------------------------------------
I've used Sphinx for a couple of projects, to great success, but its probably not what you are looking for here... fundamentally, there a two problems with general speech recognition. The first is training. Most speech recognition systems are trained in some way. Sphinx is "pre-trained" with a number of different english-language accents (mostly American), or you can retrain it yourself. The second problem is a grammar. In order to recognize a good collection of words in the english language, you need to provide a very large grammar (60,000ish words in the english language comes out to be a 200Mb+ file... probably many orders of magnitude larger than you would want in a dependency).
Long and short of it is that sphinx is great out of the box if you speak with a Mid-western American accent (as I happen to) and you are detailing with a command and control situation where you can produce a small, tailored grammar.
This is something that could be done, but i would advocate that it be left outside the core parsers.
> Speech recognition
> ------------------
>
> Key: TIKA-94
> URL:
https://issues.apache.org/jira/browse/TIKA-94> Project: Tika
> Issue Type: New Feature
> Components: parser
> Reporter: Jukka Zitting
> Priority: Minor
>
> Like OCR for image files (TIKA-93), we could try using speech recognition to extract text content (where available) from audio (and video!) files.
> The CMU Sphinx engine (
http://cmusphinx.sourceforge.net/) looks promising and comes with a friendly license.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
