Browse Prior Art Database

Local-Context-Dependent Learning in AI Kana-to-Kanji Conversion

IP.com Disclosure Number: IPCOM000105105D
Original Publication Date: 1993-Jun-01
Included in the Prior Art Database: 2005-Mar-19
Document File: 2 page(s) / 48K

Publishing Venue

IBM

Related People

Nozaki, H: AUTHOR [+3]

Abstract

Disclosed is a method for changing the priority of learned words in kana-to-kanji conversion

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 90% of the total text.

Local-Context-Dependent Learning in AI Kana-to-Kanji Conversion

      Disclosed is a method for changing the priority of learned
words in kana-to-kanji conversion

      According to differences in the learning of context, the
priority of learned words is changed in AI kana-to-kanji conversion,
with the aim of a higher hit-ratio by taking account of the
connectability of two words in a co-occurrent relation.

The method addresses three cases:

1.  A word learned outside a co-occurrence context never influences a
    co-occurrence context.  Therefore, the learned word never appears
    in the conversion result of a co-occurrence context.

2.  A word learned in a co-occurrence context influences that of
    context.  Therefore, the learned word appears in that
    co-occurrence context.

3.  A word learned in a co-occurrence context has an influence on
    outside the co-occurrence context.  Therefore, the learned word
    appears outside the co-occurrence context.

      The disclosed method consists of co-occurrence analysis, that
is, an AI kana-to-kanji conversion program and a learning method that
takes account of differences in context.  This is realized by a
learning phase and a conversion phased shown in the Figure.  The
following two methods of AI learning are used currently.

o   Method 1 -  A word learned in any context(whether or not it is a
    co-occurrence context) has priority in any conversion
    context(whether or no...