Rajeev Sangal
Professor-Emeritus, IIIT Hyderabad, India
Keynote Title: Challenges in Speech to Speech Language Translation: Prosody and Discourse
Abstract
Many scientific problems await work for successfully building speech to speech language translation (SSLT) systems. First, one has to analyze prosody as a part of automatic speech recognition (ASR), and pass it on as 'rich text' encoding, along with the output text transcript. The current ASR systems are legacy systems originally designed for producing dictations, and must be changed/upgraded to produce rich text.
Second, text to text machine translation (MT) systems have to learn to handle 'discourse', and must not be limited to handling single sentence at a time. In addition, the prosody information has to be passed on to the target language sentence or discourse segment.
Third, text to speech (TTS) systems, have to produce prosody present in the source language, conveying the same effect in the target language.
All this requires, the development of new metrics for ASR, MT and TTS, which take into account prosody as well as discourse.
The national mission on language translation of India has to be futuristic as well as practical in delivering working systems. Combining research questions with practical systems is a challenge that can be met by three virtuous cycles, combining research on the one hand, and startups and delivery on the other hand.
Bio
Prof Rajeev Sangal along with Dr Vineet Chaitanya has developed Computational Paninian Grammar (CPG) framework which is linguistically elegant and computationally efficient. It forms the basis of most work on Indian language processing today. He has written 5 books and numerous research papers.
He set up Language Technologies Research Centre at IIIT Hyderabad as its founder-head in 2002. It is the largest academic centre for computational linguistics (CL) and Natural Language Processing research in this region of the world.
He was a member of the editorial board of Computational Linguistics in 2004-06. He has been a member of editorial boards of journals: Machine Translation, Natural Language Engineering, Trans. of Asian Language Processing, CSI Transactions, etc. He reviews for major conferences such as ACL, COLING, IJCNLP, ICON, etc.
He was the founding President of NLP Association India from 2002 to 2021, and helped establish a quality annual conference ICON.
He conceptualized the National Language Translation Mission (Bhashini), and serves as the Chairman of its Executive Committee.