A Method of Detecting KATAKANA Variants in a Document

• Kubota Jun'ich

  Information System Res. Lab. Matsushita Electric Industrial Co., Ltd

• Shoda Yukie

  Information System Res. Lab. Matsushita Electric Industrial Co., Ltd

• Kawai Masahiro

  Information System Res. Lab. Matsushita Electric Industrial Co., Ltd

• Tamagawa Hirofumi

  Word Processor Division Matsushita Electric Industrial Co., Ltd

• Sugimura Ryoichi

  Information System Res. Lab. Matsushita Electric Industrial Co., Ltd

Bibliographic Information

Other Title

• カタカナ表記の統一方式

Abstract

A method to detect equivalents for approximate KATAKANA expressions from a Japanese document is proposed. As a Japanese phonetic symbol system (KATAKANA) cannot express precise pronunciation, a KATAKANA loan word can have many expressions. It is also difficult to maintain a dictionary with all KATAKANA expression items. To detect them without a dictionary, this algorithm transforms KATAKANA strings to directed graphs based on rewrite rules, then checks whether they have the same labeled path or not. This method can recall alternative KATAKANA expressions with an accuracy of 97.4%.

Journal

• IPSJ SIG Notes

  IPSJ SIG Notes 111-117, 1993

  Information Processing Society of Japan (IPSJ)

Details

• CRID: 1571417127176544128
• NII Article ID: 110002934673
• NII Book ID: AN10115061
• Text Lang: ja
• Data Source:
  • CiNii Articles