Kapil Thadani

• Who? PhD candidate in the Department of Computer Science at Columbia University. Here's a brief CV .
• Mail: 450 Computer Science Bldg
  1214 Amsterdam Ave
  Mailcode: 0401
  New York NY 10027-7003
• Email:
• Phone: +1.212.939.7191
• Bring pizza: 724 Schapiro CEPSR

Research

I'm a member of the natural language processing group at Columbia advised by Kathy McKeown and have been involved with the machine learning group and the center for computational learning systems. My current research focuses on structured prediction approaches to problems such as text alignment and text-to-text generation. I'm particularly interested in inference strategies that incorporate multiple structural representations of text. This work is relevant to applications like summarization, question answering and machine translation.

Other projects that I'm currently involved in include approaches for automatically defining and surveying scientific concepts and graph-based models for summarizing web pages. I've previously worked on problems like redundancy reduction in text, semi-parametric density estimation, selectional preference discovery, time-series clustering, unsupervised syntactic language models, adaptive topic-based language modeling for speech recognition and semi-automated corpus annotation.

Refereed publications

• Sentence Compression with Joint Structural Inference 
  Kapil Thadani and Kathleen McKeown
  In Proceedings of CoNLL 2013, Sofia, Bulgaria.
• A Joint Phrasal and Dependency Model for Paraphrase Alignment    Dataset    Code available soon
  Kapil Thadani, Scott Martin and Michael White
  In Proceedings of COLING 2012, Mumbai, India.
• On-the-fly Topic Adaptation for YouTube Video Transcription 
  Kapil Thadani, Fadi Biadsy and Daniel M. Bikel
  In Proceedings of Interspeech 2012, Portland, Oregon.
• Identifying Event Descriptions using Co-training with Online News Summaries 
  William Yang Wang, Kapil Thadani and Kathleen McKeown
  In Proceedings of IJCNLP 2011, Chiang-Mai, Thailand.
• Towards Strict Sentence Intersection: Decoding and Evaluation Strategies    Slides 
  Kapil Thadani and Kathleen McKeown
  In Proceedings of the Workshop on Monolingual Text-to-Text Generation at ACL-HLT 2011, Portland, Oregon.
• Optimal and Syntactically Informed Decoding for Monolingual Phrase-Based Alignment    Slides 
  Kapil Thadani and Kathleen McKeown
  In Proceedings of ACL-HLT 2011, Portland, Oregon.
• A Hierarchical Model for Web Summarization 
  Yves Petinot, Kathleen McKeown and Kapil Thadani
  In Proceedings of ACL-HLT 2011, Portland, Oregon.
• Time-Efficient Creation of an Accurate Sentence Fusion Corpus    Dataset 
  Kathleen McKeown, Sara Rosenthal, Kapil Thadani and Coleman Moore
  In Proceedings of NAACL-HLT 2010, Los Angeles, California.
• Corpus Creation for New Genres: A Crowdsourced Approach to PP Attachment    Slides    Dataset 
  Mukund Jha, Jacob Andreas, Kapil Thadani, Sara Rosenthal and Kathleen McKeown
  In Proceedings of the Workshop on Creating Speech and Text Language Data with Amazon's Mechanical Turk at NAACL-HLT 2010, Los Angeles, California.
• Towards Semi-Automated Annotation for Prepositional Phrase Attachment 
  Sara Rosenthal, William J. Lipovsky, Kathleen McKeown, Kapil Thadani and Jacob Andreas
  In Proceedings of LREC 2010, Valletta, Malta.
• A Framework for Decreasing Textual Redundancy    Slides 
  Kapil Thadani and Kathleen McKeown
  In Proceedings of COLING 2008, Manchester, UK.
• Density Estimation under Independent Similarly Distributed Sampling Assumptions    Addendum 
  Tony Jebara, Yingbo Song and Kapil Thadani
  In Proceedings of NIPS 2007, Vancouver, Canada.
• Spectral Clustering and Embedding with Hidden Markov Models 
  Tony Jebara, Yingbo Song and Kapil Thadani
  In Proceedings of ECML 2007, Warsaw, Poland.

Patents and other publications

• Speech Recognition with Topic-Specific Language Models
  Daniel M. Bikel, Kapil Thadani, Fernando Pereira, Maria Shugrina and Fadi Biadsy
  US Patent (pending).
• Decreasing Textual Redundancy
  Kapil Thadani
  Master's Thesis, 2007.
• Independent Similarly Distributed Sampling Assumptions for Semiparametric Density Estimation 
  Tony Jebara, Yingbo Song and Kapil Thadani
  In Proceedings of the 2007 New York Academy of Sciences Symposium on Machine Learning, New York City.

Datasets

• A corpus of phrase-based alignments derived from the Edinburgh paraphrase corpus including tokenization fixes, dependency graphs, named entity annotations and baseline alignments generated by METEOR. See Scott Martin's description and the README for more details.
  Download  (1.6 MB)    Cite 
• A small corpus of pairs of related newswire sentences with multiple human-generated fusion annotations (5 intersections, 5 unions) of varying accuracy collected via Mechanical Turk users.
  Download  (91 KB)    Cite 
• A collection of prepositional phrase attachment cases over unstructured blog text. Candidates were chosen automatically and final judgments were made by humans responding to multiple-choice questions on Mechanical Turk.
  Download  (130 KB)    Cite 

Miscellany

The papers that I covered for my candidacy exam on text-to-text generation are available here.

My Erdős number is at most 4 but my Bacon number is still woefully undefined.