Interview with Dr. Tanmoy Chakraborty
Dr. Tanmoy Chakraborty
Assistant Professor and Ramanujan Fellow
Dept. of CSE, IIIT Delhi, India
Director, Laboratory for Computational Social Systems (LCS2)
Project Director, Technology Innovation Hub, IIIT Delhi, India
Q1. Please share your educational and professional journey with us?
I did my BTech in Computer Science and Engineering from Kalyani Government Engineering College, West Bengal University of Technology in 2009. I was the topper among all disciplines in the college, received the best student award and a Gold Medal from Tata Consultancy Services (TCS). In the campus placement, I got offers from TCS and a few other companies. Due to poor financial conditions in my family, I decided not to pursue my higher studies and join TCS. However, during my internship at TCS in my final year of my BTech, I realized that a corporate job was not enjoyable to me. After discussing with my parents, I finally appeared for GATE and joined MTech in Jadavpur University in 2009, leaving all my corporate offers.
During my MTech, I got exposed to the world of research in Natural Language Processing (NLP). I stood the second position in my MTech and got a job offer from CDOT (Centre for Development of Telematics is an Indian Government-owned telecommunications technology development centre). The salary package was attractive, and I almost decided to join there. However, my MTech thesis advisor motivated me to go for higher studies, and by that time, I had already started liking research. My advisor’s inspiration acted as a catalyst to my inner suppressed passion to pursue higher studies, leading me to join IIT Kharagpur as a PhD student in January 2012.
A three and a half years’ journey at IIT Kharagpur was a roller coaster, mostly positive. I got exposed to a new field called “Social Network Analysis” by one of the finest research groups in India. I received the prestigious Google India PhD fellowship (usually 3-4 students receive this fellowship every year). The fellowship was almost doubled the usual stipend given to a PhD student by the Govt of India. The fellowship helped reduce the financial crisis of my family. My PhD research was recognized as the best thesis by various organizations such as the Indian National Academy of Engineering (INAE), IBM and Xerox Research. I got opportunities to visit abroad six times to present research at conferences and various research labs. I got my PhD degree in September 2015.
After my PhD, I got offers from six US universities for postdoctoral research. I ended by joining University of Maryland as a postdoctoral fellow to do research on social network analysis and cybersecurity. I stayed there for almost one and a half years before joining IIIT Delhi as an assistant professor in May 2017.
I established my research group, Laboratory for Computational Social Systems (LCS2; http://lcs2.iiitd.edu.in/), recruited PhD and masters’ students, started getting research grants from both industry and government sectors, and kept building the computational and research infrastructure. Currently, the lab hosts 9 PhD students, many masters and undergraduate students, with an overall strength of ~30 students. One of my colleagues joined the lab last year to co-lead the group with me. We broadly work in NLP and Social Computing with a special focus on trust and safety in online social media. Our lab’s alumni are placed all over the world — CMU, Stanford, Johns Hopkins, University of Maryland, Cambridge university, Microsoft Research, etc.
Q2. How Data Science is helpful to prevent fake and fraud actions in business world?
Well. Data science is the only mechanism by which you spot fake and fraudulent activities from the massive amount of data available online in the form of text, video, audio, time-series data, etc. Let’s take the example of financial transactions in a bank. There are millions of transactions happening every day. For a human investigator, it is not possible to examine every transaction manually. Even if you gather some knowledge from the past about the potential financial fraud patterns, the attackers are so clever that they change their attacking strategy dynamically every day. It is impossible for a human being to know about all possible attacking patterns. Therefore, we need data science algorithms to adapt and improvise dynamically to the new attacking patterns and detect fraud transactions (or at least filter out a few suspicious transactions for further manual investigation).
The same situation prevails in social media like Facebook, Twitter (fraud account, fake news, fake likes/follows, etc.), e-commerce services like Flipkart, Amazon (fraud accounts/transactions, fake comments to the products, etc.), video sharing platforms like YouTube, Snap, Instagram (malicious videos, fake comments, etc.), to name a few.
In short, human-in-the-loop is always needed. However, what data science can do is ease the job of humans and helps them detect fraud activities intelligently and efficiently.
Q3. Tell us something about your new book on fake news?
There have been extensive research efforts in the last 3-4 years worldwide from both academia and industry to deal with fake news. We, those who are active in this research area, are well aware of cutting-edge research on fake news detection. However, it may be not easy for a newcomer in this area to consolidate the literature and figure out the current scopes. Therefore, last year, we thought it might be good to summarize the existing research on fake news and provide an all-in-one resource to the practitioners.
Our book titled “Data Science for Fake News: Surveys and Perspectives” (available online at https://www.springer.com/gp/book/9783030626952) covers diverse areas in fake news literature – both theoretical and applications. It covers the state-of-the-art fake news detection methods, the existing datasets, tools and techniques. It also includes other ethical issues for counterfeit news detection, social science and political science views of fake news, medical misinformation and fake news in the Indian context. We hope that the readers will find this valuable resource for academic curriculum and research.
Q4. Tell us about your research in cyber-security?
We are focusing on different trust and safety issues in online social networks (OSNs). In particular, we build on data-driven AI-based technologies to combat hostile posts and fraud activities in OSNs, including fake news and misinformation, hate speech, fraud accounts, collusive attacks etc. Recently, we have curated massive social media datasets in Indian and English languages and have built systems to combat COVID-19 fake news detection. The model has been developed in collaboration with Accenture, which is currently running live. We have also developed solutions to fight hate speech and provocative online posts in partnership with Wipro.
Our recent collaboration with Logically (a UK-based fact-checking startup) has successfully stopped the spread of hate speech propagation on social media. We are about to start a new collaboration with LinkedIn towards the border goal of sanitizing the online ecosystem. As you have rightly noticed, all our research are industry-focused so that ultimately the end research gets transformed into technologies. We have recently filed a few patents out of these works. We are also exploring other aspects of online hostility, such as bias and fairness and other ethical issues.
Another line of research focuses on detecting group-based coordinated attacks in social networks. We got a Google faculty award and a grant from Flipkart to start working on this area. Given the tremendous growth of social networks, the social reputation of an entity in online media plays an important role. This has led to users choosing artificial ways to gain social reputation; employing blackmarket services as the natural way to boost social reputation is time-consuming. We refer to such artificial ways of increasing social reputation as “collusion”. We have built technologies to detect such blackmarket-based collusive attacks on Twitter, YouTube, Amazon, Flipkart and other online media platforms.
Q5. What is NLP? Would you please share with us about your research in NLP?
NLP stands for Natural Language Processing. We all know what a “language” is — it is a medium of communication. What is “natural language”? It is a language that evolves naturally without following any predefined grammar or syntax. Examples include the language of human beings, the language of dogs, etc. However, computer languages such as C, C++ are not natural languages; these are called “artificial languages” as these languages are designed following some predefined rules/grammar/syntax.
One may argue how a human language such as English is a “natural” language as it follows grammar (e.g., Subject-Verb-Object in a sentence and so on). However, it is important to note that the grammar did not come first. It is the English language which came first, and the grammar was gradually constructed to describe its syntax. The English language evolved naturally without caring about grammar. You may have seen the differences between the way William Shakespeare used to write English sentences and today’s language construction. You will find significant differences.
Now let us discuss what is “natural language process” or NLP. A Computer or a machine does not understand natural languages. It only understands 0s and 1s. NLP is a way to program a computer to process natural languages to mimic the way a human reacts in response to a natural language. NLP is a branch of computer science and linguistics that leverages artificial intelligence and human-computer interaction to process natural languages (speech, text etc.). NLP deals with several components — morphology, syntax, semantics, pragmatics, etc.
Our lab currently focuses on two subareas of NLP — conversation and dialogue modeling and text summarization, both in English and regional languages such as Hindi, Bengali and code-mixed language. In the former subarea, we are working on building an intelligent conversational dialogue generation system (aka chatbot) that can understand human responses and generate natural language sentences (which are syntactically and semantically correct). There are two major steps involved in building a chatbot – natural language understanding (NLU) and natural language generation (NLG). In the latter subarea, we focus on abstractive text summarization.
We all know that a summary of a large passage is always useful to read. It gives a brief overview of the passage. In abstractive summarization, the challenge is to generate new sentences rather than selecting important sentences from the passage to generate the summary. We are also working on “multimodal summarization”, where we leverage videos, speech and images to generate more meaningful abstractive summaries. It is helpful to summarize a lengthy video lecture, a TV series, a Television show or even a film.
Apart from these two, a few other projects are also going on in low-resource language modeling, sentiment analysis and emotion recognition, legal document processing, multimodal data processing, mental health and psycholinguistics, etc.
Q6. How do you help students for new innovation at TIH, IIIT-Delhi?
Technology Innovation Hub (TIH) is an initiative of the Dept. of Science and Technology (DST), Govt. of India, to foster collaborative research among government, academia and industry. This initiative aims to build the translational research ecosystem — to transform academic research to technology, tools or intellectual properties. Our institute, IIIT Delhi, received a 100cr grant from DST to establish a TIH in the area of “cognitive computing and social sensing”. We encourage students and academicians to start entrepreneurship activities and incubate their research in the form of startups. At the same time, we encourage fundamental research in cognitive computing and social sensing that will progressively enhance the core science.
We will fund academic projects and entrepreneurship activities towards the growth of India’s research in cognitive computing and social sensing. We have decided on two grand challenges, which would be our primary focus in the next 2-3 years — legal informatics and public health. In the former vertical, we aim to build a cognitive computing-based legal information management system that will serve as an all-in-one platform for the legal and judiciary domain. It would offer law-enforcement agencies, advocates and judiciary to systematically access historical court cases and help them take actions intelligently and efficiently. In short, it would aid the law and judiciary system of our country. In the public health domain, we aim to build cognitive computing enabled strategies for protecting and improving the health of people and their communities.
More details can be found at https://ihub-anubhuti-iiitd.org.
Q7. Which one thing do you like most about IIIT-Delhi?
IIIT Delhi is an industry-led academic institute whose vision is to establish a world-class research and teaching ecosystem. Being a young institute, it is less bureaucratic and more supportive. IIITD encourages all the faculties in growing their research and provides ample opportunities and support for career growth. The rules and regulations are flexible to accommodate all possible means so that ultimately faculties get full freedom to grow their research. Most importantly, the administrative staffs are very supportive. I am well aware of the academic culture of Indian institute and have experienced academic culture in the USA as well. I can clearly see a significant difference between their cultures and that of IIIT Delhi. This was the reason behind choosing IIIT Delhi as my first professional place.
Q8. Tell us about one of the moments which changed your perspective towards life?
Well. There are quite a few. The first one was, of course, my bitter experience during the industry internship in the final year of my BTech. I was so frustrated that even if my family suffered from a tremendous financial crisis, I chose to go for higher study. It turned out to be a great decision that indeed changed my career path.
However, the most memorable moment that changed my perspective of life was a recent incident a few years back. There was a time when some Indian academicians started backbiting about me in the closed and sensitive forums. They never approached me directly; rather took indirect ways to malice me. It was excruciating as a few of them were academically very close to me. They even did not stand beside me and provoked others to shout at me. They tried all possible ways to diminish me. I took several attempts to contact them and explained to them about the incident. They just ignored and never responded to my emails/calls. Several of my friends suggested me taking firm actions against them. However, I just remained silent and kept focusing on my work. I knew that it was not my words but my work that would eventually stand strong against them in future.
Eventually, with the help of some of my colleagues and well-wishers, I gradually came out of the situation. This incident taught me the basic instinct of human being and helped me push myself beyond my standard capacity. After recovering from this, I pushed myself so hard that today I have been able to establish such a vibrating research lab, have received several grants and fellowships from top industries like LinkedIn, Accenture and Wipro, have coauthored two books (one was published, the other is with the publisher for printing) and have taken the responsibility of a massive initiative like TIH. It was an eye-opener that has also taught me how to select academic collaborators, trust, friendship, and many more.
I also learned that when the power comes in hand, people pretend themselves as God and never hesitate to harm others, even their close ones. However, they never realize that power is temporary and unstable. It may be with you today, but no one knows tomorrow, it may move to some other’s hand. I believe in “karma.” I believe that every human being will be punished in this life in some way or the other due to his/her substandard karma in the past.
Q9. Which one thing do you want to change in yourself and why?
I would like to change quite a few things in myself.
First of all, I am very emotional and very quickly start believing others without thinking much. I have less control over managing adversarial situations and react fast without thinking much about their consequences. I have been learning tricks and strategies to gain control over these, and so far, it has been quite effective.
Secondly, I am over passionate about my research and over conscious of my career. Sometimes, both of these traits misbalance my personal and professional life and harm my family life. I am also trying to make a balance between my work and personal life.
Finally, I am over-aggressive about my research. I believe that with extreme push, one can do anything. I started believing in this as I went through different circumstances, and it was my extreme push that helped me overcome all the adversarial circumstances. This philosophy may be harmful in the case of slow learners and those who need more time to execute certain assignments.
Read more about Dr. Tanmoy Chakraborty @
Website: http://faculty.iiitd.ac.in/~tanmoy/
Lab website: http://lcs2.iiitd.edu.in/
Blog: https://medium.com/@chak.tanmoy.iit
Twitter: @Tanmoy_Chak
LinkedIn: https://www.linkedin.com/in/tanmoy-chakraborty-89553324/
Other media coverages: http://faculty.iiitd.ac.in/~tanmoy/mediacoverages.html