Dear all,
I am looking for a free speech to text (STT) software for writing technical documents (BSc, MSc, PhD, etc). I've found a list here: http://en.wikipedia.org/wiki/List_of_speech_recognition_software
Which STT software do you recommend and why?
Thank you very much in advance.
Kinds regards,
Fernando
Hi Fernando,
I think Kaldi is currently the best speech recognition toolkit. It supports all the most recent and most effective techniques for speech recognition. The documentation is good, but not as good as the one provided by HTK, which is amother older well-known toolkit.
Mirco
Hi Fernando,
the speech recognition software listed in the wiki page are tools for the development of a speech recognizer. They are not plug-and-play software. You have to train acoustic models, language models, etc.. which means that you need first of all to know how an ASR works: given that sphynx, htk, kaldi or julius are equivalent for close-talking speech recongition.
If you want a ready-to-use software able to cope with large vocabulary continuous speech recognition in a reliable way, then you have to buy it.
And even with the state of the art commercial speech recognition systems I'm not sure if they're suitable for domain-unrestricted speech dictation. The ASR systems are pretty good for recognition of shorter utterances (such as Siri-like commands etc.) or dictation of domain-specific texts (such as medical reports, legal texts, etc.). But I think that for such difficult texts as MSc or PhD theses, ASR will not work well. It then depends on your motivation: if you need it for a handicapped person who cannot type so that he/she is able to dictate his/her thesis, then it's OK to purchase one of the major commercially available ASR systems and the user will most likely be able to customize it to his/her needs. But if your idea is to increase productivity by allowing voice dictation of the thesis, then I think there is no such a system (and I frankly doubt there will soon be any).
It's a pity there's no ready-to-use software. Such kind of software would save a lot of working time for many people, although one can type fast. Among free software packages, did you have good results using Kaldi?
There are ready-to-use applications, but definitely not free. The good ones are often not cheap, but I think they're affordable. So instead of building own ASR from an open-source toolkit (which could be very difficult), it's probably much more efficient to purchase a commercial system for 100-200USD. The performance of the major commercial ASR systems (I don't want to name any of them here) is quite impressive considering the situation 10 years ago, but I still think that their advertisement and commercial presentations are somewhat exaggerating. It's true that you can quite quickly dictate a short work email to your colleague asking him to provide you with some documents, or text a message with a lunch appointment, but dictating more difficult and creative texts isn't possible, or at least convenient. If you really put some effort into it, you can learn how to speak to the system in order to minimize the word error rate. In some specific cases (e.g. medical report dictation), you can achieve almost 100% recognition accuracy. But this probably will not work in case of much less predictable texts such as PhD theses.
Thank you! Then, which software package do you recommend Jan?
A text to speech software I recommend is here
https://speechtech.co.za
You these references
Conference Paper Web-based Automatic Pronunciation Assistant
Conference Paper The Development of Local Synthetic Voices for an Automatic P...
My reasons are :
- The TTS system can be used on resourced languages (like English) and low-resourced languages (like most African languages).
- With limited data you can build your custom TTS voice.
I recommend Express Dictate Digital Dictation Software
https://www.nch.com.au/www/express/
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Software