Skip to main navigation Skip to search Skip to main content

The research of the maximum length n-grams priority Chinese word segmentation method based on corpus type frequency information

  • School of Management, Harbin Institute of Technology

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In order to solve the difficulties to extract words in particular domain, we formulate a method of automatic word segmentation in Chinese based on corpus type frequency information. This method can effectively extract n-gram words that are not predefined in a lexicon by setting the maximum length (n) of the n-gram word we want to extract from a sentence and the minimum threshold frequency the n-gram word appears in corpus. When the real frequency the n-gram appears in corpus is above the threshold, the n-gram word will be extracted. If there are two or more n-grams have the same length, the higher frequency one will be chosen, and then the next higher frequency one if any of its characters are not in previous one.

Original languageEnglish
Title of host publicationProceedings of the 2012 National Conference on Information Technology and Computer Science, CITCS 2012
PublisherAtlantis Press
Pages71-74
Number of pages4
ISBN (Print)9789491216381
DOIs
StatePublished - 2012
Externally publishedYes
Event2012 National Conference on Information Technology and Computer Science, CITCS 2012 - Lanzhou, China
Duration: 16 Nov 201218 Nov 2012

Publication series

NameProceedings of the 2012 National Conference on Information Technology and Computer Science, CITCS 2012

Conference

Conference2012 National Conference on Information Technology and Computer Science, CITCS 2012
Country/TerritoryChina
CityLanzhou
Period16/11/1218/11/12

Keywords

  • Corpus type frequency information
  • N-gram word
  • Word frequency
  • Word segmentation

Fingerprint

Dive into the research topics of 'The research of the maximum length n-grams priority Chinese word segmentation method based on corpus type frequency information'. Together they form a unique fingerprint.

Cite this