Natural Language Processing

In my time at Microsoft Research and at IIT Kanpur, I have had the chance to explore both practical and interesting problems in Natural Langauge Processing

  • Text Categorization using Sparse Composite Vectors
    In this project, I worked on the novel approach of topic-based document representation (distributional semantics representation) which outperforms state-of-the-art models in multi-class and multi-label classification tasks. We also showed that fuzzy GMM clustering on word-vectors leads to more coherent topics than LDA and can be used to detect Polysemic words. (Joint work with Dheeraj Mekala, IIT Kanpur, Bhargavi Paranjape, Microsoft Research Lab, India & Prof. Harish Karnick, IIT Kanpur), [Paper] [PPT]

  • Product Classification using Distributional Semantics
    In this project, I worked on the problem of hierarchal product classification for a given ontology(taxonomy) tree using a novel two-level ensemble approach based on a path-wise, node-wise and depth-wise classifier for product classification with respect to the taxonomy.(Joint work with Prof. Harish Karnick, IIT Kanpur, Ashendra Bansal, & Pradhuman Jhala,[Paper] [Poster] [PPT]

  • Text Summarization using Abstract Meaning Representation
    In this project, we explored a full-fledged pipeline for text summarization with an intermediate step of Abstract Meaning Representation (AMR). Our proposed method achieves state-of-the-art results compared to the other text summarization routines based on AMR. (Joint work with Shibhansh Dohare, IIT Kanpur & Prof. Harish Karnick, IIT Kanpur). [Paper]