Shivani Mody, February 23, 2009
Times of India (Bangalore edition)
In government offices in Karnataka, you have clerks typing in Kannada using a keyboard having the local language alphabet. But now the state government is in the process of loading software onto the computers that will allow users to input text in English but which will output words in the local language using predictive transliteration technology.
The first pilots are being carried out in Bangalore and if successfully completed, it will be implemented across Karnataka.
The use of software as a translation tool is not new. Content in local language is increasing on the Internet on a daily basis. Today, nearly one lakh words are added online per day in local Indian languages. It has also become common for web browsers to find news and information in local languages.
Initially, the only way people could write in a local language was using the alphabet keyboard. This is a tedious process, as a user requires 3 to 4 months training to actually work with it. Using a software tool such as Quillpad users can type in English and get the output in 10 Indian languages.
The tool developed by Tachyon Technologies has a great number of users accessing the website, writing the text and then copying and pasting the content onto desired websites. Most of the Indian language fonts are included in Unicode, a standard, which when installed on websites allows you to see the text as it is.
Machine Learning Technology
The Quill technology works on advanced machine learning techniques, which makes it capable of learning new patterns from examples, and does not require manual intervention. All new words from all sources are added for testing, which is continuously updated onto the software. “These are still early stages. It will take some time for the software to evolve,” says Ram Prakash H, founder of Tachyon Technologies. The machine learning in Indian languages is a complex process as there are 14 major Indian languages and 200 plus dialects. Also the pronunciations and writing styles (using mathras) differs for each language.
0 comments:
Post a Comment