We aim to develop an agent for understanding the distribution of public opinions and preferences; thus, this agent needs to have causality knowledge. When discussing social issues, Wikidata, a knowledge base, can provide linked data and play an important role in analyzing discussion content. However, there is a lack of causal content on Wikidata, and some content has errors. Therefore, it is necessary to automatically extract knowledge from news and add it to Wikidata. We propose a method of automatically determining causality in text and directly extracting effect from news. We collected news and used GPT-3 to infer whether a news article is causally related to the entity and further infer the effect of this entity. We also attempted to increase the reliability of extracted causality knowledge by dealing with multilingual texts.
CITATION STYLE
Jin, Y., & Shiramatsu, S. (2023). Multilingual Complementation of Causality Property on Wikidata Based on GPT-3. In Lecture Notes in Networks and Systems (Vol. 464, pp. 573–580). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-981-19-2394-4_52
Mendeley helps you to discover research relevant for your work.