• CLaC Labs - S
• Belief
• BioKI
• Cinema
• Prepositions
• Sentiment
• Summarization

Summarization: ERSS

ERSS is a system that evolved from 2003-2007 and participated in each NIST-sponsored DUC competition. Working under the assumption that the most important entities of a newspaper article are referred to most frequently, ERSS employs a coreference-based summarization strategy. ERSS approximates noun phrase (NP) coreference resolution with a shallow processing environment to identify the most important NPs to be included in the summary. While for single documents the longest coreference chain is a good indicator for importance, for multi-document summaries we developed a more intricate cluster graph algorithm. This algorithm is robust and adapts naturally to different subtasks, in fact it underlies all of our contributions to DUC competitions. In particular, the cluster graph algorithm adapts well to introducing a focus, or bias into the summary.

2007

• Rene Witte and Sabine Bergler
  Fuzzy Clustering for Topic Analysis and Summarization of Document Collections
  In:Proceedings of the 20th Canadian Conference on Artificial Intelligence (AI 2007), Montreal, Quebec, May 28-30 2007
  Best Paper Award



2006

• R. Witte, R. Krestel, Universit\"at Karlsruhe; S. Bergler
  Context-Based Multi-Document Summarization Using Fuzzy Coreference Cluster Graphs
  In: Proceedings of DUC 2006, Document Understanding Workshop at HLT-NAACL, June 8-9, 2006, Brooklyn, New York USA

2005

• Rene Witte, Ralph Krestel, Sabine Bergler
  ERSS 2005: Coreference-Based Summarization Reloaded
  In:Workshop on Text Summarization, Document Understanding Conference (DUC 2005), October 9-10, 2005, Vancouver. NIST

2004

• Bergler, S., R. Witte, Z. Li, M. Khalife, Y. Chen, M. Doandes, A. Andreevskaia
  Multi-ERSS and ERSS 2004
  In:Workshop on Text Summarization, Document Understanding Conference (DUC 2004), May 6-7, 2004, Boston. NIST

2003

• Rene Witte and Sabine Bergler
  Fuzzy Coreference Resolution for Summarization
  In: Proceedings of the International Symposium on Reference Resolution and Its Applications to Question Answering and Summarization (ARQAS 2003), June 23-24, 2003, Universita Ca'Foscari, Venice, Italy. pp. 43-50
• Bergler, Sabine, Rene Witte, Michelle Khalife, Zhuoyan Li, and Frank Rudzicz
  Using Knowledge-poor Coreference Resolution for Text Summarization
  In:Workshop on Text Summarization, Document Understanding Conference (DUC 2003), May 31-June 1, 2003, Edmonton, Canada. NIST