Executive Summary : | Nowadays there has been a growing tendency of using chatbots for various applications. India is a multilingual country with great linguistic and cultural diversities. Unfortunately, to the best of our knowledge, none of the available chatbots supports Indian languages. This project aims at developing a chatbot- Sevak, which can understand the written and spoken Indian languages (Hindi, Bengali, Tamil, and Telugu) along with English. Indian population is habituated to chat in the mix (code-mixed) languages i.e. mixing of Hindi in English (Hinglish), mixing of Bengali in English (Benglish) etc. To reach out to India’s heart investigator’s proposed chatbot will have the flexibility to operate in a code-mixed environment. The engine will have three major components, viz. Natural Language Understanding (NLU), Dialogue Manager (DM), and the Natural Language Generation (NLG). Each of these modules will be developed using the recent methods and tools of Natural Language Processing (NLP) and Machine Learning (ML). At the end of 3 years, investigators will develop Sevak - an open-source, pluggable, chatbot engine, which supports Indian languages. The chatbot is targeted to serve the following three very crucial sectors of society: a). Sevak for Railways, and b). Sevak for Primary Health Care, and c). Sevak for the Judiciary Domain the aim is to launch Sevak service through Facebook messenger, WhatsApp services, android app-based service, and a voice call-based service for the users who do not have smartphone facilities. The primary beneficiary would be the common man - seeking railway, healthcare and judiciary-related information. Moreover, this will be useful to the various government organizations related to railways, health and judiciary. The Ministry of Health and Family Welfare, the Ministry of Railways, and the Department of Law and Justice can use it further to reach out to the common people they are targeting to serve. |
Co-PI: | Prof. Pushpak Bhattacharyya, Indian Institute of Technology (IIT) Patna, Dr Amitava Das, Assistant Professor, Indian Institute of Information Technology (IIIT), Andhra Pradesh, Dr Manish Shrivastava, Assistant Professor, International Institute of Information Technology (IIIT) Hyderabad, Dr Dipankar Das, Assistant Professor, Jadavpur University, Kolkata |
Total Budget (INR): | 98,64,300 |
Achievements : | (i). A prototype model for Sevak has been developed in the Railways domain;
(ii). The chatbot, Sevak has been integrated to Facebook and Whatsapp;
(iii). A Chit-chat model based on deep learning has been developed;
(iv). There are 10 versions of Sevak are currently available: English, Hindi-Native, Hindi-Roman, Hinglish (Hindi-English code-mixed), Bengali-Native, Bengali-Roman, Bengalish (Bangla-English code-mixed), Telugu-Native, Telugu-Roman, Tenglish (Telugu-English Code-mixed). |
Publications : | 6 |
