kapil thadani

Hi there.

You've wandered onto a webpage for a PhD student at Columbia University. I'm part of the natural language processing group and am also involved with the machine learning group and the center for computational learning systems. My advisor is Kathy McKeown. Here's a brief research CV 

• Email:
• Call: +1.212.939.7191
• Drop by: 724 CEPSR, Columbia University, NYC

Research

I spend most of my time being fascinated and frustrated by the fields of computational linguistics and machine learning. My research focus is in structured prediction for text-to-text generation tasks. For instance, I'm currently working towards building models for manipulating text under formal set-theoretic operations for combination and decomposition. Specifically, we attempt to generate fluent sentences with valid parses without explicitly relying on any one particular semantic representation of text. This work is relevant to areas like automated paraphrasing, summarization, question answering and translation.

Other projects that I'm currently involved in include the unsupervised retrieval of latent structure from text, topic models for web summarization and semi-automated annotation strategies. I've previously worked on problems like redundancy reduction in text, semi-parametric density estimation, selectional preference discovery and time-series clustering.

Refereed publications

• A Joint Phrasal and Dependency Model for Paraphrase Alignment 
  Kapil Thadani, Scott Martin and Michael White
  In proceedings of the 24th International Conference on Computational Linguistics (COLING), Dec 2012, Mumbai, India.
• On-the-fly Topic Adaptation for YouTube Video Transcription 
  Kapil Thadani, Fadi Biadsy and Daniel M. Bikel
  In proceedings of Interspeech, Sep 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, Nov 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, June 2011, Portland, Oregon.
• Optimal and Syntactically Informed Decoding for Monolingual Phrase-Based Alignment    Slides 
  Kapil Thadani and Kathleen McKeown
  In proceedings of the Annual Meeting of the Association of Computational Linguistics (ACL-HLT) Short Papers, June 2011, Portland, Oregon.
• A Hierarchical Model for Web Summarization 
  Yves Petinot, Kathleen McKeown and Kapil Thadani
  In proceedings of the Annual Meeting of the Association of Computational Linguistics (ACL-HLT) Short Papers, June 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 the 11th Annual North American Meeting of the Association of Computational Linguistics (NAACL-HLT) Short Papers, June 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, June 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 the 7th International Conference on Language Resources and Computation (LREC), May 2010, Valletta, Malta.
• A Framework for Decreasing Textual Redundancy    Slides 
  Kapil Thadani and Kathleen McKeown
  In proceedings of the 22nd International Conference on Computational Linguistics (COLING), August 2008, Manchester, UK.
• Density Estimation under Independent Similarly Distributed Sampling Assumptions    Addendum 
  Tony Jebara, Yingbo Song and Kapil Thadani
  In proceedings of the 21st Conference on Neural Information Processing Systems (NIPS), December 2007, Vancouver, Canada.
• Spectral Clustering and Embedding with Hidden Markov Models 
  Tony Jebara, Yingbo Song and Kapil Thadani
  In proceedings of the 18th European Conference on Machine Learning (ECML), September 2007, Warsaw, Poland.

Other publications

• Decreasing Textual Redundancy
  Kapil Thadani
  Master's Thesis, December 2007, New York, New York, USA.
• Independent Similarly Distributed Sampling Assumptions for Semiparametric Density Estimation 
  Tony Jebara, Yingbo Song and Kapil Thadani
  In proceedings of the 2nd New York Academy of Sciences Symposium on Machine Learning, October 2007, New York, New York, USA.

Datasets

• A corpus of ~300 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)   Citation
• A collection of ~940 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)   Citation

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: Me → Tony Jebara → Tommi Jaakkola → Noga Alon → Paul Erdős
My Bacon number is still woefully undefined.