An approach to extraction of causal chains among events in multiple documents

摘要

In this paper, we aim to extract causal relations in multiple documents and then overlook the whole causal chains among the events described in the documents. In order to extract causal relations from documents, we focus on the words recognized as clue expressions for causal relations. By connecting multiple causal relations to make a network of the causal chains, we have introduced matching methods based on word similarity and topic similarity. We have verified our proposed method is useful for extracting causal chains from actual newspaper articles.