Selected Publications

2008

• Empiricism is Not a Matter of Faith (Pedersen), Computational Linguistics, Volume 34, Number 3, pp. 465-470, September 2008. [Journal Citation Reports Index Factor 2007: 2.367]
  
  
  [This explains why we release our software, and how we develop it so that it's both possible and productive to do so. The goal is to promote greater sharing of software so that experimental results can be reproduced reliably and quickly.]

2007

• Measures of Semantic Similarity and Relatedness in the Biomedical Domain (Pedersen, Pakhomov, Patwardhan, and Chute), Journal of Biomedical Informatics, 40(3), 288-299, June 2007. [Journal Citation Reports Index Factor 2006: 2.346]
  
  
  [This shows how WordNet::Similarity is being extended to the biomedical domain. We find that a variant of our vector measure agrees much more with medical experts than do measures based on hierarchical paths in SMOMED-CT.]

2006

• A Comparative Study of Supervised Learning as Applied to Acronym Expansion in Clinical Reports (Joshi, Pakhomov, Pedersen, and Chute) - Appears in the Proceedings of the Annual Symposium of the American Medical Informatics Association, pp. 399-403, Nov 11-16, 2006, Washington, DC. [acceptance rate 41%]
  
  
  [This shows that acronym expansion in medical text can be effectively handled using supervised learning techniques first developed for word sense disambiguation of general English text.]
  
  
  
• Unsupervised Corpus Based Methods for WSD (Pedersen), In Agirre, E. and Edmonds, P. (Editors), Word Sense Disambiguation : Algorithms and Applications, June 2006, pp. 133-166, Springer.
  
  
  [This describes the foundations of SenseClusters, which clusters written contexts based on their lexical similarity. Various related approaches are also discussed, including Latent Semantic Analysis.]

2005

• Name Discrimination by Clustering Similar Contexts (Pedersen, Purandare, and Kulkarni) - Appears in the Proceedings of the Sixth International Conference on Intelligent Text Processing and Computational Linguistics, p. 220-231, February 13-19, 2005, Mexico City. [acceptance rate 37%]
  
  
  [We show how to use SenseClusters to discover the identities associated with a name mentioned multiple times in a text.]

2004

• Word Sense Discrimination by Clustering Contexts in Vector and Similarity Spaces (Purandare and Pedersen) - Appears in the Proceedings of the Conference on Computational Natural Language Learning (CoNLL), pp. 41-48, May 6-7, 2004, Boston, MA. [acceptance rate 48%]
  
  
  [ We introduce SenseClusters, which remains a very active project. The goal is to develop unsupervised methods of clustering similar contexts based on lexical features discovered by the Ngram Statistics Package. ]

2003

• Extended Gloss Overlaps as a Measure of Semantic Relatedness (Banerjee and Pedersen) - Appears in the Proceedings of the Eighteenth International Joint Conference on Artificial Intelligence, pp. 805-810, August 9-15, 2003, Acapulco, Mexico. [acceptance rate 21%]
  
  
  [ We introduce the extended gloss overlap measure (aka lesk in WordNet::Similarity). The goal is to measure how related two concepts are based on the similarity of their definitions as found in WordNet. ]

• Using Measures of Semantic Relatedness for Word Sense Disambiguation (Patwardhan, Banerjee and Pedersen) - Appears in the Proceedings of the Fourth International Conference on Intelligent Text Processing and Computational Linguistics, pp. 241-257, February 17-21, 2003, Mexico City. [acceptance rate 46%]
  
  
  [ We introduce WordNet::Similarity and WordNet::SenseRelate, both of which remain very active projects. The goal is to measure the similarity of concepts based on WordNet, and to use that information to assign each word in a running text the sense that is most related to its neighbors. ]

By: Ted Pedersen - tpederse AT d umn edu