Learning to Update Knowledge Graphs by Reading News

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 2632–2641, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 2632 Learning to Update Knowledge Graphs by Reading News Jizhi Tang, Yansong Feng, Dongyan Zhao Institute of Computer Science and Technology, Peking University The MOE Key Laboratory of Computational Linguistics, Peking University {tangjizhi, fengyansong, zhaody}@pku.edu.cn Abstract News streams contain rich up-to-date informa- tion which can be used to update knowledge graphs (KGs). Most current text-based KG up- dating methods rely on elaborately designed information extraction systems and carefully crafted rules, which are often domain-specific and hard to maintain or generalize. However, such methods may not pay enough attention to the implicit information that lies underneath texts, thus often suffer from coverage issues. In this paper, we propose a novel graph based neural network method, GUpdater, to tackle these problems1. GUpdater is built upon graph neural networks (GNNs) with a text-based at- tention mechanism to guide the updating mes- sage passing through the KG structures. Ex- periments on a real-world KG updating dataset show that our model can effectively broadcast the news information to the KG structures and perform necessary link-adding or link-deleting operations to ensure the KG up-to-date accord- ing to news snippets. 1 Introduction Knowledge graphs have been widely used in dif- ferent areas, where keeping the knowledge triples up-to-date is a crucial step to guarantee the KG quality. Existing works attempt to synchronize KGs with encyclopedia sources (Morsey et al., 2012; Liang et al., 2017), which basically leverage struc- tured data, while directly updating KGs using plain texts, such as news snippets, is not an easy task and still remains untouched. For example, given a KG recording the rosters of all NBA teams, we need to update this KG according to a news article which describes a trade between different teams. Given the following news snippet that re- ports a trade between Minnesota Timberwolves 1Code and data are available at: https://github. com/esddse/GUpdater and Philadelphia 76ers, one has to add and delete many links in the KG, as illustrated in Figure 1. The Minnesota Timberwolves has ac- quired forward Robert Covington, for- ward Dario ˇSari´c, guard Jerryd Bay- less and a 2022 second-round draft pick from the Philadelphia 76ers in exchange for forward Jimmy Butler and center Justin Patton. Current approaches either rely on manual up- dating, or often solve it in a two-step manner: first, using an off-the-shelf information extraction (IE) tool to extract triples from the text, and then modi- fying the KG according to certain rules predefined by domain experts. However, we should notice that besides what we can explicitly extract from the text, the implicit information, i.e., information that is not mentioned in the text but can be inferred from the text, should also be taken into considera- tion. This requires far more complicated updating rules. In the above example, if Robert Coving- ton (RC) is traded from Philadelphia 76ers …

Content truncated to 500 words