A few general questions (about speech recognition)
Thread poster: Samuel Murray
Samuel Murray
Samuel Murray  Identity Verified
Netherlands
Local time: 15:14
Member (2006)
English to Afrikaans
+ ...
Nov 5, 2007

G'day everyone

I would like to add a starter article about speech recognition to the ProZwiki. Here is what I have so far:

===========================================

Speech recognition

Generally speaking, speech recognition software turns human speech into text. Software that recognise spoken commands can also be referred to as speech recognition software, but the ones that turn any spoken words into text are the ones usually used by Transla
... See more
G'day everyone

I would like to add a starter article about speech recognition to the ProZwiki. Here is what I have so far:

===========================================

Speech recognition

Generally speaking, speech recognition software turns human speech into text. Software that recognise spoken commands can also be referred to as speech recognition software, but the ones that turn any spoken words into text are the ones usually used by Translators. The best known speech recognition software is Dragon Naturally speaking.

There is a forum on ProZ.com specifically dedicated to speech recognition.

Commercial software (todo: write objective reviews)

* Dragon NaturallySpeaking, by Nuance Communications
* ViaVoice, by IBM
* L&H Voice Xpress Professional, by Lernout & Hauspie
* Voice Studio, by Ultimate Interactive Desktops

===========================================

And to add some more meat to it, would you please help me answer the following questions?

* How common/possible is it that you can simply play an audio file and let the program extract the text?

* How well do such programs work for interviews (not carefully dictated speech)?

* What languages do these programs generally support (eg there is no list of languages on the Dragon web site)?

* Do you know of not so well known programs for specific purposes or not so common languages?

* Are there any viable free (open source or otherwise) speech recognition programs or projects?

Any other questions which you think is highly relevant to speech recognition that one might answer in a wiki article?

Thanks!

[Edited at 2007-11-05 07:25]
Collapse


 
Steven Capsuto
Steven Capsuto  Identity Verified
United States
Local time: 10:14
Member (2004)
Spanish to English
+ ...
Some of the answers Nov 5, 2007

Samuel Murray wrote:
* How common/possible is it that you can simply play an audio file and let the program extract the text?


Speech recognition software can only achieve relatively good accuracy after it has been "trained" to recognize a particular person's speech pattern. If you play another person's voice into it, accuracy will plummet. It also can be thrown off by things like background noise. So just playing any old recording of a person speaking won't work.

Also, even once the software has been trained, it makes errors, particularly with homonyms, hastily pronounced words, and unfamiliar proper nouns. If you were dictating "live," you might notice these errors and dictate corrections on the fly. But if you have the program recognize language from a recording, you'll have to do serious quality control at the end, possibly by listening to the recording as you proofread.


* How well do such programs work for interviews (not carefully dictated speech)?


Badly in most cases, since they involve more than one voice. The packages I've used (which are a few years old) can only have one voice profile loaded at a time. In any case, usually the people in the interview recording won't be available to train your software.


* What languages do these programs generally support (eg there is no list of languages on the Dragon web site)?


The last time I checked, the versions sold in English-speaking countries only recognize English. But all of the Dragon versions for other languages also include English recognition.

Good luck with the article.

[Edited at 2007-11-05 08:20]


 
Suzanne Blangsted (X)
Suzanne Blangsted (X)  Identity Verified
Local time: 07:14
Danish to English
+ ...
one voice Nov 5, 2007

I agree with the article from Steven. I would like to add that to avoid the many mistakes that can occur with other voices than your own is to listen to the text in a sentence with the software "off", then turn the software "on" (I used the + key on the keyboard), and repeat the sentence with your own voice. The accuracy of recognizition should be 99%, if you have trained the voice recognition software correctly, and if you are speaking clearly.

If you on a regular basis receive ma
... See more
I agree with the article from Steven. I would like to add that to avoid the many mistakes that can occur with other voices than your own is to listen to the text in a sentence with the software "off", then turn the software "on" (I used the + key on the keyboard), and repeat the sentence with your own voice. The accuracy of recognizition should be 99%, if you have trained the voice recognition software correctly, and if you are speaking clearly.

If you on a regular basis receive material with the same voice, you might be able to train your program to recognize that voice also, but I doubt it.

I only have the English version of the Dragon but this Dragon is also available in a few other languages, the major ones.
Collapse


 


To report site rules violations or get help, contact a site moderator:


You can also contact site staff by submitting a support request »

A few general questions (about speech recognition)






Trados Business Manager Lite
Create customer quotes and invoices from within Trados Studio

Trados Business Manager Lite helps to simplify and speed up some of the daily tasks, such as invoicing and reporting, associated with running your freelance translation business.

More info »
CafeTran Espresso
You've never met a CAT tool this clever!

Translate faster & easier, using a sophisticated CAT tool built by a translator / developer. Accept jobs from clients who use Trados, MemoQ, Wordfast & major CAT tools. Download and start using CafeTran Espresso -- for free

Buy now! »