Help Building a Medical Speech Recognizer

[umair h.]

I am looking into building a STT app for medical professionals. As you guys already know, the words can be difficult to pronounce and often misinterpreted by systems. I was looking at Sphinx4 and PocketSphinx.

I need real time processing, and will ideally have it on a server.
Should I modify dictionary, acoustic model, language model or all of them?
How do I get started and where to find basic help?

Thanks!

[Nickolay V. Shmyrev]

Should I modify dictionary, acoustic model, language model or all of them?

all of them

How do I get started

http://cmusphinx.sourceforge.net/wiki/tutorial

[umair h.]

Thanks. Which one would you recommend, Sphinx4 or Pocketsphinx.

[Nickolay V. Shmyrev]

I do not know details of your project to give you recommendation.

[umair h.]

Serverside ASR, with an Emphasis on speed

[Nickolay V. Shmyrev]

For serverside processing it is better to use Kaldi.

[umair h.]

What are the differences between sphinx and Kaldi. I was under the impression the sphinx is better