OXYMORON COLLECTING DEVICE AND COMPUTER PROGRAM FOR THE SAME

PROBLEM TO BE SOLVED: To provide a device that can efficiently collect oxymorons in a unit smaller than a sentence.SOLUTION: A oxymoron collecting device includes: a first stage contradiction pattern classification unit for extracting a pattern pair comprising patterns contradicting each other, by t...

Full description

Saved in:
Bibliographic Details
Main Authors HASHIMOTO CHIKARA, JULIEN KLOETZER, OTAKE KIYOTAKA, TORISAWA KENTARO, SANO MOTOKI, OH JONG HOON
Format Patent
LanguageEnglish
Japanese
Published 20.04.2015
Subjects
Online AccessGet full text

Cover

Loading…
More Information
Summary:PROBLEM TO BE SOLVED: To provide a device that can efficiently collect oxymorons in a unit smaller than a sentence.SOLUTION: A oxymoron collecting device includes: a first stage contradiction pattern classification unit for extracting a pattern pair comprising patterns contradicting each other, by the machine learning using the pattern pair comprising the patterns of the form such as "X performs of Y" as learning data; an additional contradiction pattern pairs derivation unit 130 for renewing one pattern and deriving new pattern pairs by applying an implication relationship about each of extracted pattern pairs; a learning data extension unit 136 for extending the learning data by adding the pattern pairs having high possibility comprising the patterns contradicting to each other among derived new pattern pairs to the learning data; and a SVM 142 for performing the second stage classification of classifying the pattern pairs given by the machine learning using the extended learning data into the pattern pairs comprising the patterns contradicting to each other and other patterns. 【課題】文より小さな単位で矛盾表現を効率よく収集できる装置を提供する。【解決手段】矛盾表現収集装置は、「XがYを する」という形式のパターンからなるパターン対を学習データとする機械学習により、互いに矛盾するパターンからなるパターン対を抽出する第1段階矛盾パターン分類部と、抽出されたパターン対の各々について、含意関係を適用して、一方のパターンを書換えて新たなパターン対を導出する追加矛盾パターン対導出部130と、導出された新たなパターン対の内で、互いに矛盾するパターンからなる可能性が高いパターン対を学習データに追加することで学習データを拡張する学習データ拡張部136と、拡張された学習データを用いた機械学習により、与えられたパターン対を互いに矛盾するパターンからなるパターン対とそれ以外とに分類する第2段階分類を行なうSVM142とを含む。【選択図】図4
Bibliography:Application Number: JP20130210793