IMPLICATION RECOGNITION DEVICE, METHOD, AND PROGRAM

PROBLEM TO BE SOLVED: To improve the accuracy of implication recognition.SOLUTION: Partial sentence extraction means 22 analyzes dependency for each of t1 and t2 to extract a list of partial sentences. Synonym independent part extraction means 24 extracts, for each independent part in t2, a list of...

Full description

Saved in:
Bibliographic Details
Main Authors ASANO HISAKO, MATSUO YOSHIHIRO, BESSHO KATSUTO
Format Patent
LanguageEnglish
Japanese
Published 29.11.2018
Subjects
Online AccessGet full text

Cover

Loading…
More Information
Summary:PROBLEM TO BE SOLVED: To improve the accuracy of implication recognition.SOLUTION: Partial sentence extraction means 22 analyzes dependency for each of t1 and t2 to extract a list of partial sentences. Synonym independent part extraction means 24 extracts, for each independent part in t2, a list of synonym independent parts in t1 completely synonymous to or synonymous to the aforementioned independent part. Alignment selection means 26 selects alignment between the independent parts. Alignment similarity degree calculation means 28 calculates a degree of similarity of the selected alignment. Alignment similarity degree correction means 30 calculates a correction similarity degree of the selected alignment. Inter-text similarity degree calculation means 32 calculates a maximum value of the correction similarity degree of each alignment as a degree of similarity between the texts t1 and t2.SELECTED DRAWING: Figure 1 【課題】含意認識の精度を向上させることができるようにする。【解決手段】部分文抽出手段22が、t1,t2それぞれに対し、係り受け解析し、部分文のリストを抽出する。類義自立部抽出手段24が、t2中の各自立部に対し、該自立部と同義または類義のt1中の類義自立部のリストを抽出する。アライメント選択手段26が、自立部間のアライメントを選択する。アライメント類似度算出手段28が、選択したアライメントの類似度を算出する。アライメント類似度補正手段30が、選択したアライメントの補正類似度を算出する。テキスト間類似度算出手段32が、各アライメントの補正類似度の最大値を、テキストt1,t2間の類似度として算出する。【選択図】図1
Bibliography:Application Number: JP20170094854