Synopsi: Design and development of an automatic text summarization system

Start Date: 01/10/1999
End Date: 31/12/2000
Funding: EPET II
Project Leader: Piperidis Stelios

Synopsi aimed at the design and development of a text summarization prototype based on the text extract summarization paradigm. The developed methodology computes a score for each sentence in the text to be summarized, depending on the sentence’s importance in the text. Sentence score calculation is based on a set of sentence and text features, ranging from termhood of the sentence’s nominal chunks to actual sentence position in the text. Highly-scoring sentences are then aggregated and presented in the same order as in the text. Since, the sentences that are thus extracted present a certain degree of incoherence, methods for improving coherence, cohesion and coverage have been investigated.

The expected benefit of automatic text summarization systems is multi-dimensional. On the one hand information dissemination is largely accelerated by the reduction of the time needed to compile a summary. On the other hand, summaries help :

  • Keep people informed of the very recent developments in their areas of interest
  • Reduce reading time
  • Allow for better selection of those articles that need to be read in full
  • Make bibliographic research easier
  • Enable a less costly transmission of information
  • Help information providers overcome current size constraints, e.g. in wap enabled applications