Summary extraction from chinese text for data archives of online news

  • Authors:
  • Nozomi Mikami;Lukáš Pichl

  • Affiliations:
  • International Christian University, Tokyo, Japan;International Christian University, Tokyo, Japan

  • Venue:
  • DNIS'11 Proceedings of the 7th international conference on Databases in Networked Information Systems
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

Electronic news media consistently use a specific language frame for efficient knowledge delivery and opinion formation. Since machine representation of logographs, and their derived forms, such as ideograms, and the Chinese characters in general, enumerates to a large set of symbols, the information content of particular text sequence interconnects context patterns across various scope ranges. Here we concern with the enumerated form of sinogram reflecting on the characters not only historically and culturally, but also educationally. Logographs visually invoke mutual functional relations by design and through their usage in overlaping scopes. Here we study the procedural summarization of text originally intended for online news distribution and the preferable evaluation method of its usability. Sinogrammatic electronic news sentences are analyzed for mutual similarity patterns both inward and outward, in order to facilitate sentence extraction for summary inclusion while reflecting on the principle of characters. Traditional partition of linguistic knowledge representation is aided by invocation of bypass routes in logographic text similar to software pictograms, for which design and usage frames are coeducational. Machine extracted summaries are compared with human chosen sentences while employing the Turing test to ascertain cohesion of Human - Human and Human - Machine comparison. The implementation of popularity-based summarization algorithm is available as a Java program.