論 文Papers

CONFERENCE (INTERNATIONAL)

Weakly Supervised Multilingual Causality Extraction from Wikipedia

Chikara Hashimoto

2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP 2019), 2019/11

Category:

自然言語処理 (Natural Language Processing)

Abstract:
We present a method for extracting causality knowledge from Wikipedia, e.g., Protectionism -> Trade war, where cause and effect entities correspond to Wikipedia articles. Such causality knowledge is easy to verify by reading corresponding Wikipedia articles, translate to multiple languages through Wikidata, and connect to knowledge bases derived from Wikipedia. Our method exploitsWikipedia article sections that describe causality and the redundancy stemming from the multilinguality of Wikipedia. Experiments showed that our method achieved precision and recall above 98% and 64% respectively. Our method could extract the causality between entities that were written distantly in a Wikipedia article and rivaled an oracle relation extractor which perfectly detected the causality between entities co-occurring in a sentence. We release codes and data for further research.

Software Download
Download:

Weakly Supervised Multilingual Causality Extraction from Wikipedia(外部サイト/External Site Link)