What is Anulekhika?

Anulekhika is an ASR (Automatic Speech Recognition) engine that that takes voice or audio as input and returns text as output. It may be considered as a voice typing or dictation tool where the user can speak to it and get their voice transcribed or written in their native language.

How many languages does Anulekhika support?

Anulekhika at present supports a total of 20 languages/mother tongues of India with various levels of accuracy. The languages supported are as follows:

Assamese, Awadhi, Bengali, Bodo Parja, Haryanvi, Gujarati, Hindi, Kashmiri, Konkani, Maithili, Malayalam, Marathi, Meitei, Marwari, Sindhi, Tamil, Telugu, Urdu, English, Kannada

Which model has Anulekhika been built own?

Anulekhika is built on top of the Massively Multilingual Speech (MMS) model of Facebook. The interface we have developed supports only the Indian languages/mother tongues given there in. Given that MMS supports only a few languages and does not give accuracy in many of these languages, LDC-IL is also making small efforts to train the ASR models on its own with the available datasets as we progress towards are main task of collecting and curating the datasets in Indian languages. As part of this effort, we have also developed the Maithili ASR engine with high accuracy (comparable to that of Hindi and English in general domain. The Maithili ASR engine is also hosted here.

Does Anulekhika have its own ASR engines?

As noted above, LDC-IL is not equipped to train large language models as of now. However, we do language specific research as part of the research and development tasks we undertake. As part of this, the Maithili ASR engine has been developed and hosted here.

Can Anulekhika be used as a dictation tool?

Yes, it can be used as a dictation tool which would require some post-editing work.

Can Anulekhika give a transcript of an audio file?

Yes, Anulekhika portal is equipped to take an audio file of a given language and it can return the transcript for it in the script of the chosen language. However, due to limited resources, there is a limit on the size of the audio file that can be requested to provide transcript of.

Can I upload large audio files to Anulekhika and get the transcript of the same in the desired languages?

As noted above, at present users can upload a file of not more than 50MB size and get a transcript of it. If any user has a file size that is greater than this, the same may be requested over email by sending an email to Dr. Narayan Choudhary at oic-ldcil@gov.in