We aim to develop an agent for understanding distribution of public opinions and preferences. For this development, the agent needs to have causality knowledge. At the same time, when discussing social issues, Wikidata, as a knowledge base, can provide linked data and play an important role in analyzing the content of the discussion. However, there is a lack of causal content on Wikidata, and even some content has errors. In our re-search, we propose a method of automatically judging causality and directly extracting cause or effect from Japanese Wikipedia articles. Moreover, we are trying to increase the reliability of extracted causality knowledge by dealing with multilingual texts.