抄録
様々な観点から類似していると判断できる文書群から、文書評に内在する共通性を汎化された共通イベント列(極大類比)として抽出し、極大類比に包摂される任意の文書を関連文書と判断するアクセス手法を考える。極大類比を構成する際に必要な計算量を軽減せしめ、かつ同時に関連性判断の精度を向上させる目的で、所与の質問式Qの具体化で事例文書Iの汎化になっている極大類比に限定した構成法を与える。
We present here a method for constructing a common abstract event sequence from two or more documents also regarded as event sequences. To reduce the computational cost and to improve the precision of relevance judgement made by the abstract event sequences, we introduce a notion of base queries and consider a retrieval process to find an instance of base query that is similar to a given family of instance documents. By the base query and its instance documents, we can design a divide and conquer strategy for the construction of abstract common event sequence from the instances.
