X
    Categories: FacultyProfiles

IIITH’s Language Forensic Sherlock – Why Chiranjeevi Yarra loves his English accents

Chiranjeevi Yarra’s research  in Spoken English and Indian language forensics and informatics could help millions in Rural India to speak English better, by understanding phonemes and accents in their local language and dialect, for a more productive ML-based user experience.

Mix a dollop of Sherlock Holmes’ Spidey-sense to identify accents, with Prof. Higgins’ talent (from My Fair Lady) for syllabic correction. That’s the space where Chiranjeevi Yarra and his team work.

When Yarra left his Telugu-medium village school in Sompeta on the Andhra-Orissa border, little did he realize that he would one day be contributing significant research in the Indian English domain and in speech signal processing, machine learning, and time-varying signal analysis. The Prime Minister’s Fellowship awardee for Doctoral Research is an alumnus of IISc Bangalore, IIT Kharagpur and NIT Warangal.

When the scholar secured the 2nd rank in both his M.Tech and Ph.D, it was a sweet victory for the youngster who had once struggled with English. Coming from Telugu medium instruction till his 10th graduation, it was a big challenge for him in his early diploma years, particularly in English-heavy subjects like electrical materials. “But the challenge spurred me on and from the 2nd year, I was among the top 5 students. I couldn’t have reached where I did without the support of my parents, wife, and Ph.D advisor Dr. Prasanta Kumar Ghosh”, notes the researcher.

A Parsing acquaintance with Prosody
An Assistant Professor at IIIT Hyderabad’s Speech Lab –Language Technologies Research Center (LTRC), Yarra’s team is one of the few Indian researchers working on Prosody. For instance, a common-place greeting like ‘Good Morning’ can be loaded with significance, to convey anything from frosty anger, a punch in the gut, to jaunty happiness or the warmth of freshly brewed coffee. Just shows how tonal variations and emphasis on different syllables, can communicate emotional range and offer a wealth of tonal prosodic information!

Chiranjeevi with his Speech Lab group

What makes Yarra’s research in statistical signal analysis for speech information retrieval most significant, is its application to educational technology and under-resourced rural linguistic development.

Leading among his work are voice tutor apps for syllable detection, like VoisTUTOR (IISc-Xerox), In-SPIRE / SPIRE-SST, English Gyani for Pronunciation quality analysis and an Online tool for transcribing the phonemes of spoken data, among other small yet significant tools for creating different databases.

VoisTutor demo

Against all odds
The first in his village to secure a Ph.D, it was discipline and a rugged routine that helped Yarra to juggle his Ph.D work and a small infant, two milestones that would happen in the same month. He completed his Ph.D in 6 years, winning the Best thesis award (Prof. D J Badkas  Gold medal).

Looking back, Yarra muses that he has come a long way from playing cricket on green fields, surrounded by the famous coconut trees, that gave the area the title of second Konaseema. “Being a mere 5kms from the Orissa border, we enjoyed a hybrid culture of Andhra and Orissa. I was inspired by my father, a self-made man who rose from selling garments on a cycle to co-owning a small garment business”. The turning point was when he took tuitions and cracked the 7th standard state scholarship exam that opened the world out to him. After acing the education entrance test for Andhra’s residential schools, he was assigned to Simhachalam near Visakhapatnam where he completed his schooling. He went on to pursue a 3-year diploma in Electrical and Electronics Engineering from Government Polytechnic, Visakhapatnam, followed by B.Tech from NIT Warangal (2002-06).

Yarra secured the 17th national rank in GATE but due to financial constraints, he would work in Hewlett Packard for a year. He soon realized that his interests were more oriented towards academic research and took the decision to pursue his M.Tech in signal processing and Instrumentation at IIT Kharagpur (2007-09). “I enjoyed my TAship under Prof. S. Maka  and realized that I had what it takes to be a teacher.  In 2009, I joined KIIT, a deemed university in Bhubaneshwar as an assistant professor”.

Three apps and a baby
Yarra was toying with the idea of pursuing his Ph.D, when he got married to Anem Bhagya Lakshmi. Their daughter was born in the same month that he got admission into his Ph.D program at IISc Bangalore. She was named Lahiri, a shout out to his research work in speech signal processing. “The stipend was lesser than the emoluments offered by the two IITs but I fortunately got a 4-year Prime Minister fellowship for doctoral research, that nicely stabilized my financial situation. I was able to enjoy my education at IISc.”, he observes.

Bonding with his daughter Lahiri since his PhD days

His Ph.D in System Science and Signal Processing (2013 – 20) under Dr. Prasanta Kumar Ghosh resulted in two significant apps – Vois Tutor (Xerox Research, DST-Imprint and CIST collaboration) and English Gyani – an intelligent assistant for tutoring English via learner-tutor interactions. “We also worked on Imprint, a GOI project on Indian English pronunciation and comprehension, on phoneme transcriptions with Central Institute of Indian Languages (CIIL) and built a 250 hours’ pronunciation quality database on English-speaking student population and its dynamics”. For nearly three years, he worked on an automated feedback system for minimizing the nativity influences of language in learners of spoken English. It was an eventful doctoral tenure, with 26  conference publications and journals, development of three web-tools and mentoring 25  students from eminent colleges.

“Money was always tight but my entire education was funded through government scholarships that really helped me achieve my career goals” says the rank-holder in GATE-2006, the Andhra Pradesh Engineering Common Entrance Test-2002 and the district level Class X board exams (1998). Yarra has been the recipient of several Travel grants for various top tier conferences like ICASSP, Interspeech and an invitation to  Indo -UK summit and 7th BRICS youth summits. He has over 40 publications in different topics in NLP, speech recognition and processing, convolutional neural nets, deep learning (AI), computational linguistics, including a novel  publication that garnered over 50 citations in 4 years.

Guess that accent!
“I had delayed applying to IIITH, since I was skeptical about getting into the Institute, that had such a great  senior faculty in the speech team. But I took the chance and was delighted to get an immediate response”, smiles Yarra who joined LTRC in December 2020. His team is currently focusing on regional pronunciations of English and how the speaker’s native language influences his accent.  “We have a few DST projects in the pipeline, with different IIITH Centers, to identify regional accents and understand how prosodic cues can have applications in text to speech, speech to speech translation in the Indian context, emotion recognition and for detecting symptoms of depression, Parkinson’s or ALS from speech”, he explains. The scholar takes courses in speech signal processing and time varying analysis and has recently introduced a course called speech analysis and linguistics, on the linguistic components in speech.  “The faculty that I have interacted with, like Prof. Dipti Misra, Prof. Anil Vuppala, Prof. Jayanthi Sivaswamy , Prof. Radhika Mamidi, Prof. C V Jawahar, Prof. Bapi Raju, Prof. Priyanka Srivastava, Prof. Vineet Gandhi, Prof. Vikram Pudi or Prof. Kavita Vemuri,  whether for collaborations or project idea discussions, have always been very  accommodating and made me feel most welcome. I didn’t see this kind of culture elsewhere and this is one of the happiest things in IIIT!”

Celebrations with LTRC faculty

The road ahead
“The two things that I learnt, to manage my time efficiently, was discipline and a regular routine. I love the greenery on the IIIT campus and the sudden temperature dip, when I cycle to my Lab every morning. IIIT Hyderabad’s student quality and their skillsets are very good and is comparable to the old IITs. The one thing that makes the Institute stand apart, is their research to industry translation capabilities”, he adds.

Chiranjeevi with his students

Weekends are family time and cooking with  his wife and badminton sessions with his daughter. Yarra sometimes relaxes with a good Telugu novel, a detective movie or sports on television. His travels have mostly been to international conferences, in China, Austria, Moscow, Calgary or Cebu in Philippines where “we got to network with senior dignitaries in digital signal processing and listen to experiences from industry folk”, he observes.

Family time

Right now, Yarra plans to explore prosody in two major applied research areas – detection of speech depression in normal undergraduate and graduate students. The second is speech to speech machine translation between two Indian languages or with English. In the future, he hopes to explore the prosodic variations in Indian languages, which is yet an unexplored area.

Deepa Shailendra is a freelance writer for interior design publications; an irreverent blogger, consultant editor and author of two coffee table books. A social entrepreneur who believes that we are the harbingers of the transformation and can bring the change to better our world.

Deepa Shailendra :Deepa Shailendra is a freelance writer for interior design publications; an irreverent blogger, consultant editor and author of two coffee table books. A social entrepreneur who believes that we are the harbingers of the transformation and can bring the change to better our world.