A research team from the speech technology research section of the National Electronics and Computer Technology Centre (Nectec) has made such a possibility a reality. By integrating text-to-speech technology with a mail server, the team has developed a new e-mail reader application.

"The e-mail reader technology allows you to access your mail box through a conventional telephone or mobile phone," said Chai Wutiwiwatchai, chief of the section.

Instead of accessing the Internet to check e-mail messages, he said people can call an assigned number to access their mail box in order to hear their messages over the phone.

The development has made use of the lab's development of a Thai text-to-speech software known as Vaja, which allows the server to read text to users.

The prototype is currently being tested at Nectec, but Chai said the software itself needs to be developed further.

He said the team plans to design a dialogue-based browsing system that will allow users to access their mailbox for messages more easily.

"We have to develop a browsing method that uses dialogue interaction between the system and users as a navigator, to allow users easier access to their mailbox," he said.

However, the team admitted the quality of the Vaja reading voice is not yet good enough. They hope to develop a more natural voice for Vaja by next year.

Nectec's current Vaja software is in version 3, with version 4 already under way and scheduled for launch by the end of this year.

Chai said Vaja version 4 would add more features to support client-server and Web services. The previous Vaja version is supported only by a stand-alone PC.

As it is hoped the new version will support an interactive voice response application, Chai added the new version would also include multi-stream functions to handle multiple, simultaneous requests as well as a queuing management system.

However, with developing better voice quality as the next stage, the team hopes to develop Vaja version 5, with a more natural voice. This version is expected to come out in 2007.

In addition to Vaja, the team is also working on another piece of speech software called iSpeech, which is a type of Thai speech recognition software used to recognise voice and convert voice to text. The first iSpeech version came out last year and is now in version 1.5.

Even though the existing version simply recognises voice and converts it into text word by word, it can detect a real voice command, even in a noisy environment.

Chai said the team is developing the next version of iSpeech, which will be improved to recognise continuous speech but still be restricted by the complexities of grammar.

This version is due out by the end of this year.

"As for next year, we hope iSpeech will be developed to not only serve continuous speech but also have improved grammar capabilities, which means it will be able to recognise any kind of speech and convert it into text more accurately," he said.

He said iSpeech technology could be applied to a variety of applications as it offers users a more convenient way of turning speech into text.

The centre is working with the Justice Ministry to create a form of this technology to help courts with the transcription process.

"Normally, a court's judgement has to be transcribed into text and this is currently done manually. If the technology was adopted for use as a transcription assistant in the court, it could improve efficiency and shorten the process," Chai said.

Pongpen Sutharoj

The Nation