I am a fourth year CS PhD student in the MURGe Lab (part of the bigger UNC NLP Lab) at the University of North Carolina at Chapel Hill, advised by Prof. Mohit Bansal. My PhD is supported by a Google PhD Fellowship and a Rebecca and Munroe Cobey Fellowship.

My broad research interests are in interpretable machine learning and natural language processing. I am generally interested in multi-step reasoning problems (over text and semi-structured data), often referred to as the System 2 Reasoning. In the past, I have developed interpretable models that can generate Natural Language Proofs for formal reasoning (EMNLP 2020, NAACL 2021), Explanation Graphs for structured commonsense reasoning (EMNLP 2021, ACL 2022), Summarization Programs for abstractive summarization (arxiv 2022), and Multi-step Reasoning Paths for Text Generation from semi-structured data (arxiv 2022). I also wrote a paper on connecting explainability to data hardness (EMNLP 2022). During my PhD, I have spent two wonderful summers interning at FAIR Labs, Meta AI Research and Salesforce AI Research.

Before starting my PhD, I was a Research Engineer at IBM Research - India building industry-scale Intelligent Tutoring Systems (IJCAI 2019, CIKM 2018, AIED 2018). Even before that, I was an M.Tech. student in the CS department at IIT, Delhi, having worked with Prof. Mausam and developed the state-of-the-art Open Information Extraction system (Open IE 5.0) (ACL 2017, COLING 2018).

Recent News

• January 2023: Summarization Programs is accepted to ICLR 2023.
• December 2022: New pre-print on Multi-step Reasoning for Text Generation from Semi-structured Data on arXiv.
• October 2022: New paper on connecting Data Hardness to Explainability is accepted to EMNLP 2022.
• September 2022: New pre-print on Interpretable Abstractive Summarization via generating and executing modular programs on arXiv.
• August 2022: I am one of eight Google PhD fellows in NLP worldwide. Thanks, Google, for the prestiguous award.
• May 2022: Started an internship at Meta AI Research in Seattle with Asli Celikyilmaz.
• February 2022: New paper on Contrastive Learning for Explanation Graph Generation is accepted to ACL 2022, Dublin, Ireland.

Industry Experience

• Research Intern (with Asli Celikyilmaz), Meta AI Research (Summer 2022)
• Research Intern (with Nazneen Rajani), Salesforce AI Research (Summer 2021)
• Research Engineer, IBM Research (2017-2019)
• Member of Technical Staff, Adobe Systems (2014-2015)

Selected Publications [Google Scholar] [dblp]

• MURMUR: Modular Multi-Step Reasoning for Semi-Structured Data-to-Text Generation
  Swarnadeep Saha, Xinyan Velocity Yu, Mohit Bansal, Ramakanth Pasunuru, and Asli Celikyilmaz
  Pre-print on arXiv, 2022
  [Long] [paper] [code]

• Summarization Programs: Interpretable Abstractive Summarization with Neural Modular Trees
  Swarnadeep Saha, Shiyue Zhang, Peter Hase, and Mohit Bansal
  International Conference on Learning Representations (ICLR) 2023
  [Long] [paper] [code]

• Are Hard Examples also Harder to Explain? A Study with Human and Model-Generated Explanations
  Swarnadeep Saha, Peter Hase, Nazneen Rajani, and Mohit Bansal
  Conference on Empirical Methods in Natural Language Processing (EMNLP), 2022
  [Short] [Oral] [paper] [data]

• Explanation Graph Generation via Pre-trained Language Models: An Empirical Study with Contrastive Learning [Accept Rate: 21%]
  Swarnadeep Saha, Prateek Yadav, and Mohit Bansal
  Annual Meeting of the Association for Computational Linguistics (ACL), 2022
  [Long] [Poster] [paper] [code]

• ExplaGraphs: An Explanation Graph Generation Task for Structured Commonsense Reasoning [Acceptance Rate: 23%]
  Swarnadeep Saha, Prateek Yadav, Lisa Bauer, and Mohit Bansal
  Conference on Empirical Methods in Natural Language Processing (EMNLP), 2021
  [Long] [Oral] [paper] [data/code] [website]

• multiPRover: Generating Multiple Proofs for Improved Interpretability in Rule Reasoning [Acceptance Rate: 26%]
  Swarnadeep Saha, Prateek Yadav, and Mohit Bansal
  Conference of the North American Chapter of the Association for Computational Linguistics (NAACL), 2021
  [Long] [Oral] [paper] [code]

• PRover: Proof Generation for Interpretable Reasoning over Rules [Acceptance Rate: 22%]
  Swarnadeep Saha, Sayan Ghosh, Shashank Srivastava, and Mohit Bansal
  Conference on Empirical Methods in Natural Language Processing (EMNLP), 2020
  [Long] [Oral] [paper] [code]

• ConjNLI: Natural Language Inference over Conjunctive Sentences [Acceptance Rate: 22%]
  Swarnadeep Saha, Yixin Nie, and Mohit Bansal
  Conference on Empirical Methods in Natural Language Processing (EMNLP), 2020
  [Long] [Poster] [paper] [data/code]

• Pre-Training BERT on Domain Resources for Short Answer Grading [Acceptance Rate: 23%]
  Chul Sung, Tejas Dhamecha, Swarnadeep Saha, Tengfei Ma, Vinay Reddy, and Rishi Arora
  Conference on Empirical Methods in Natural Language Processing (EMNLP-IJCNLP), 2019
  [Short] [Poster] [paper]

• Aligning Learning Objectives to Learning Resources: A Lexico-Semantic Spatial Approach [Acceptance Rate: 17%]
  Swarnadeep Saha, Malolan Chetlur, Tejas Indulal Dhamecha, W M Gayathri K Wijayarathna, Red Mendoza, Paul Gagnon, Nabil Zary, and Shantanu Godbole
  28th International Joint Conference on Artificial Intelligence (IJCAI), 2019
  [Long] [Oral+Poster] [paper]

• Creating Scoring Rubric from Representative Student Answers for Improved Short Answer Grading [Acceptance Rate: 17%]
  Smit Marvaniya, Swarnadeep Saha, Tejas I. Dhamecha, Peter Foltz, Renuka Sindhgatta, and Bikram Sengupta
  27th ACM International Conference on Information and Knowledge Management (CIKM), 2018
  [Long] [Oral] [paper]

• Open Information Extraction from Conjunctive Sentences [Acceptance Rate: 37%]
  Swarnadeep Saha, and Mausam
  27th International Conference on Computational Linguistics (COLING), 2018
  [Long] [Oral] [paper] [code] [slides]

• Sentence Level or Token Level Features for Automatic Short Answer Grading?: Use Both [Acceptance Rate: 25%]
  Swarnadeep Saha, Tejas I. Dhamecha, Smit Marvaniya, Renuka Sindhgatta, and Bikram Sengupta
  19th International Conference of AI in Education (AIED), 2018
  [Long] [Oral] [paper] [slides]

• Balancing Human Efforts and Performance of Student Response Analyzer in Dialog-based Tutors [Acceptance Rate: 25%]
  Tejas I. Dhamecha, Smit Marvaniya, Swarnadeep Saha, Renuka Sindhgatta, and Bikram Sengupta
  19th International Conference of AI in Education (AIED), 2018
  [Long] [Oral] [paper] [slides]

• Bootstrapping for Numerical Open IE [Acceptance Rate: 18%]
  Swarnadeep Saha, Harinder Pal, and Mausam
  55th Annual Meeting of the Association for Computational Linguistics (ACL), 2017
  [Short] [Poster] [paper] [code] [poster]

Teaching

• COL 774: Machine Learning - Spring 2016
  Instructor: Prof. Parag Singla

• COL 333/COL 671: Artificial Intelligence - Fall 2016
  Instructor: Prof. Mausam

Awards

• Google PhD Fellowship in Natural Language Processing (2022).
• Munroe and Rebecca Cobey fellowship award at UNC (2019).
• Research appreciation award by director of IBM Research India for contributions to Watson Tutor (2018).
• Manager’s choice award at IBM Research - India for research contributions (2018).
• Best Masters thesis award at IIT Delhi (2017).

Professional Service

• Reviewer for EMNLP 2022, ARR 2022, ARR 2021, EMNLP 2021, AIJ 2021, NAACL 2021, AAAI 2020, AIED 2019, NAACL 2019, EMNLP 2018.