Hierarchical Trivia Fact Extraction from Wikipedia Articles

Proceedings of the 28th International Conference on Computational Linguistics, pages 4825–4834 Barcelona, Spain (Online), December 8-13, 2020 4825 Hierarchical Trivia Fact Extraction from Wikipedia Articles Jingun Kwon1, Hidetaka Kamigaito1, Young-In Song2 and Manabu Okumura1 1Tokyo Institute of Technology 2Naver Corporation kwon.j.ad@m.titech.ac.jp {kamigaito,oku}@lr.pi.titech.ac.jp song.youngin@navercorp.com Abstract Recently, automatic trivia fact extraction has attracted much research interest. Modern search en- gines have begun to provide trivia facts as the information for entities because they can motivate more user engagement. In this paper, we propose a new unsupervised algorithm that automati- cally mines trivia facts for a given entity. Unlike previous studies, the proposed algorithm targets at a single Wikipedia article and leverages its hierarchical structure via top-down processing. Thus, the proposed algorithm offers two distinctive advantages: it does not incur high compu- tation time, and it provides a domain-independent approach for extracting trivia facts. Experi- mental results demonstrate that the proposed algorithm is over 100 times faster than the existing method which considers Wikipedia categories. Human evaluation demonstrates that the proposed algorithm can mine better trivia facts regardless of the target entity domain and outperforms the existing methods. 1 Introduction Modern search engines offer a rich knowledge panel that appears on the right side of a user screen as a search result for a given entity to improve user engagement; See an example in Figure 1. One piece of information presented in the knowledge panel can be a trivia fact about the given entity, which contributes to the effective user engagement (Tsurel et al., 2017). For example, the Google search engine provides a trivia fact for a given entity to attract users’ attention (Korn et al., 2019). Successfully attracting users’ attention can facilitate the users to revisit the search engine (O’Brien and Toms, 2008). In contrast, poor user engagement would result in the users’ switching to a competing search engine (White and Dumais, 2009). Trivia facts are, therefore, intended to improve the search quality to increase user engagement by making the search process more interesting. In contrast to common, expected, or normal information, a trivia fact is an interesting fact that is unusual, unexpected, or unique (Prakash et al., 2015). Tsurel et al. (2017) demonstrated that trivia facts are usually little-known facts, thus, well-known facts (e.g., basic background information) about a given entity cannot be considered as trivia. For example, taking Confucius as an example, the following fact in the Wikipedia article can be considered as trivia: “There is a grand-scale memorial ceremony, called Seokjeon Daeje, twice a year in South Korea for Confucius.” It is unusual and unexpected for such a memorial ceremony to be held in South Korea because Confucius is a popular Chinese philosopher and politician. In addition, surprising facts can be also considered as trivia because surprise and trivia worthiness are strongly correlated (Tsurel et al., 2017). For example, the sentence in the Wikipedia article for the film ‘The Dark Knight Rises’ states: “On July 20, 2012, during a midnight showing of The Dark Knight Rises at the Century …

Content truncated to 500 words