Test Set for Summarization Techniques
This page provides the result of our experiments in developing text
summarizing algorithm using lexical chains. The specific
technique used here is described in Using
Lexical Chains for text Summarization (Regina Barzilay and Michael
Elhadad, Mar 1997, submitted).
For each text,we give the results of 3 variants
of our algorithm and the intermediate data showing segmentation
and lexical chains in a visual manner. The results include the following
information:
- The text of the summary.
- The position of the extracted sentences in the original text.
- The position of the segment boundaries identified by our segmentation
algorithm.
- The list of lexical chains identified in the text with their score.
- For each lexical chain, its visual rendering as a graph, the
distribution of the lexical items belonging to the chain in the
original text and the senses (synset) identified in WordNet for
each lexical item.
Note that the algorithm also identifies noun-compounds using a shallow
grammar of English. Noun-compounds are identified with their heads in the
search for lexical relations in WordNet.
All the results are obtained using
WordNet 1.5.
(The Economist 96)
(Los Angeles Times 96)
(The Economist 96)
(The Economist 96)
(The Islamic Harold 95)
(The Economist 96)
(The Economist 96)
(Scientific American 96)
(The Economist 96)
(The Economist 96)
(The Economist 96)
(The Economist 96)
Text 13: News
(Reuters 97)
(The Economist 97)
(Scientific American 96)
(The Economist 97)
Text 18: Wine
(Wall Street Journal)
(The Economist 97)
(The Economist 96)
Text 21: News
(Reuter)
(The Economist 96)
(The Economist 97)
Text 24: Hope
(The Economist 96)