Research Interests
, , 1900
I would love to develop systems that belong in:
- The problem of recommendation amuses me particularly. I am curious to learn more about how companies like Amazon, Netflix, Airbnb store data of millions of customers efficiently and provide instant recommendations. I have been exploring the application of Graph Neural networks to understand the various challenges in the domain.
- Developing scalable computational language models utilizing large-scale user generated text on social media platforms:
- I believe that social media platforms such as Reddit and Twitter, rich ecological sources of human interactions, can offer large volumes of unbiased conversations that can be summarized using NLP. They can help form targeted messaging and mitigation strategies, especially during crises and public health emergencies.
- Expanding on the above, the rife of hate speech and toxicity is of particular concern today. The language models like BERT and GPT2 have been found to frequently favor identity terms, such as “gay” and “black,” that describe a particular group of people. I am currently exploring techniques to mitigate such biases and build fair language models. Additionally, During crisis like COVID 19, the spread of fake news and misinformation have exacerbated the attempts by governments to distribute vaccines to fight the virus.