LanguageTechnology.org is a not-for-profit open source foundation working towards the development of Indian Language Technologies.

There are more than 10 million visually challenged people in India. Only very few of them are fortunate enough to have formal education in blind schools, that too which are mainly situated in major cities. The main factor, hindering their exposure and education is their lack of ability to read printed books. Developing a tool which can automatically read the text and books printed in their native languages, coupled with a speech interface opens a new horizon for the visually challenged people to quench their thirst for knowledge. The aim of this project is to research and develop a system which can read the printed books in a local language and synthesize corresponding speech output. We are developing software, which reads monolingual Printed Tamil or Kannada books and later, from bilingual electronic text involving Tamil or Kannada with interspersed English words and generates speech.

Automatic book reader is a two tier process. The first module is known as Optical Character Recognizer (OCR). This module converts the printed Tamil or Kannada text to the corresponding electronic text on scanning the book or paper using any standard scanner. The converted text is normally stored in standard codes (ISCII or Unicode) and later it will be used for online or offline usage. This process involves many image processing and pattern recognition techniques for recognizing and classifying the printed text. The first version of Tamil OCR is already ready in a demonstrable form. Now, it is only required to make it time efficient and parallelize with other concurrent processes.

The second module is called Text to Speech Synthesizer (TTS). This module converts the stored, monolingual Tamil or Kannada electronic text to the corresponding speech in the accent of a native Tamil or Kannada speaker. This involves development of Rich text corpus and the corresponding Speech corpus for Tamil, Kannada and English, Natural Language Processing of Tamil, Kannada and English, Speech signal Processing, and synthesis. Finally, the output of this module will be an intelligible and natural sounding speech corresponding to the bilingual electronic text given as the input to the module. First versions of TTS in both Tamil and Kannada are already ready. Improving the quality of TTS in both Tamil and Kannada are the activities being pursued, as well as integrating them with the OCRs already being developed.

Most of our projects are hosted on code.google.com. Please see book-reader for the project on Automated Book Reader for the Visually challenged.