Service quality now at commercial level
Powerful high speed bandwidth and new algorithm designs are leveraging Thai text-to-speech web services, enabling disabled people to enjoy easier access to data, and the service will expand to eBook users in the next two years.
pTalk program for converting Thai text to speech on Windows Mobile, developed by Sirindhorn International Institute of Technology students using core Vaja technology.
It has taken nearly decade to develop Thai text-to-speech to a suitable quality, and commercialise use will soon become a reality thorough with "Vaja WebService", according to Chai Wutiwiwatchai, director of Human Language Technology Lab at the National Electronics and Computer Technology Centre (Nectec).
The program is now at Version 6 and uses a new algorithm to facilitate a more fluid use of Thai language, without disruption between transition sentences and allowing a good quality of voice synthesis.
The lab already signed agreement with Internet Thailand to host its service in data center to support concurrent users.
The service allow users to copy plain text from a website and play the speech synthesis or to type their own.
With more high-speed and wireless networks available, using a web browser is a convenient way to use the service, whether by mobile phone or computer.
Furthermore, the lab plans to develop software that allows users to download a plug-in for Android phones and allow users to have documents such as email read aloud.
In this way, users will not need to copy and paste the text.
Recently, students at Sirindhorn International Institute of Technology, Thammasat University develop on Vaja an application which can work on Windows Mobile Phone called "pTalk".
However, this version is still limited by a slow text-to-voice process delay of around five to seven seconds for plain short messages.
Currently, the program has been distributed to 46 Thammasat university students, with co-operation from Advanced Info Service, C&C INFO Advance and Microsoft (Thailand).
Previously, Vaja only worked on PCs because it consumes lots of memory, but the students have reduced the size of their program so that it could run on mobile phones for the first time.
Chai continued that by end of this year, the service will add male and children's voices to read the text and offer a greater variety of choice for customers.
At present there is only a female voice, with a selection of reading speeds.
Currently, Thai text-to-speech services are mostly available on PCs and high-end model mobile phones, with screen reader software costing as much as a few hundred dollars.
Over the next two years, Chai expects the lab will develop voice synthesis with tones or emotions that will help enrich the reading out of eBooks.
The lab expects to spend two million baht on compiling a voice database to collect different tones and sentences.
"The service will open free for public use, with a charging model in place if users need to add or use more features," he added.
"There will also be a licensing agreement if developers are interested in building commercial applications to add on to the service."
No comments:
Post a Comment