When Yashashree Chandak read her first research paper as a BTech sophomore at Pune Institute of Computer Technology, she didn’t expect to be mind-blown. “It was the ground breaking study on the use of the Transformer model in attention mechanism, ‘Attention is all you need’ which was co-authored by one of my seniors from college,” she relates. It piqued her curiosity into the world of algorithms and machine learning, drawing her towards research. She began scouting for research internship opportunities and as luck would have it, chanced upon the call for summer research interns by iHub-Data, a DST-funded technology innovation hub at IIITH. Under the Summer Research Internship on Technological Innovations (SRISHTI) program, iHub-Data has been inviting applications from engineering undergrads across the country to help enhance their academic and research skills.
Making Summers Count
A 2.5 month-long residential program, SRISHTI requires students to undertake research or any allied activities under the guidance of a qualified research supervisor and a team of researchers at IIITH. “From the very first day of SRISHTI, I was determined to make the most of this invaluable opportunity and immerse myself in the world of research,” says Yashashree who took her definitive steps at the Memory And NeuroDynAmics (MANDA) Lab. The MANDA Lab which is part of the Cognitive Science lab is focused on neurocognitive studies of human memory, learning and decision-making. Hence Yashashree’s work at the lab fittingly revolves around unravelling the intricate interaction between memory and language.
Work At The MANDA Lab
Yashashree’s current work is focused on understanding the narrative flow quantification of autobiographical memory retrieval. Inspired by a Microsoft research project that explored human memory and cognition based on recollected (autobiographical) events versus imagined stories, Yashashree under the guidance of Dr. Vishnu Sreekumar set out to understand underlying cognitive differences behind the writing styles of biographies versus autobiographies. “In autobiographies, when someone is writing his or her own story, retrieved memories can cue other memories and these jumps between events may not always follow a sequential order. But when someone is writing another person’s life history, the narrative flow is more likely to be orderly and sequential due to greater emphasis on narrativization by the biographers,” she says. In order to test the hypothesis, they analysed 210 paragraphs of the same event from autobiographies and biographies of 22 different personalities with the help of Large Language Models. With a validation of their initial hypothesis, the team has now moved on to performing a linguistic analysis on the selected texts. “We’re using a readability test that can validate whether biographies which are written by professional biographers are written at a higher level of complexity than autobiographies which may be more lucid given that they are based on one’s own experiences and also given that these authors are unlikely to be professional writers,” explains Yashashree. The initial work was accepted as an abstract with a blitz presentation at the 9th Annual Conference of Cognitive Science (ACCS9) that was held at IIT Delhi in December 2022. The ongoing research has also been accepted as a member abstract at the renowned Cognitive Science Society’s annual meeting (CogSci) in Sydney in July 2023.
“From day one, it was evident that Yashashree possessed a sharp intellect and a strong work ethic,” remarks Prof. Sreekumar labelling her ability to generate novel ideas and explore uncharted territory minus a prompt, as a testament to her innate research abilities. He goes on to add that submitting a full 6-page paper to the CogSci conference is a remarkable milestone coming as it did merely 6-7 months into their collaboration. “This accomplishment is particularly noteworthy considering she managed a full course load at her home institute simultaneously. Additionally, her contribution stands out as the sole paper submission from our lab,” says the professor. Unlike her other peers whose research-related activities came to a halt with the end of the internship program, Yashashree continues to collaborate with the MANDA lab. “In addition to our work on coherence of textual passages in books, we are also trying to do a recurrence analysis of events in the biographies and autobiographies,” says the ML enthusiast while adding that a journal submission is in the works too.
Adobe STEM Scholarship
When Adobe India’s Women In Technology scholarships were recently announced, it did not come as a surprise to find that Yashashree was one of six recipients selected for 2023. The scholarship recognises talented female undergraduates and masters students studying Computer Science and provides them an opportunity to “learn, build and grow.” The recognition covers tuition fees for the remainder of the recipient’s university education, an opportunity to interview as an intern at Adobe India, mentorship by a senior technology leader and an all-expenses paid trip to the Grace Hopper Conference India. “The knowledge and skills acquired during my internship proved instrumental during the selection process,” says Yashashree. As a project lead at the MANDA lab, she had not only demonstrated team leadership skills but also revealed her technical prowess through her work on large language models. “The award has definitely reinforced the significance of my research experiences and their impact on my professional journey,” she says. For someone who defines herself as a “lover of Maths problems”, Yashashree’s other hats include that of a kathak dancer, painter and singer.