EventWiki: A Knowledge Base of Major Events

EventWiki: A Knowledge Base of Major Events Tao Ge1,2, Lei Cui2, Baobao Chang1, Zhifang Sui1, Furu Wei2, Ming Zhou2 1MOE Key Laboratory of Computational Linguistics, Peking University, Beijing, China 2Microsoft Research Asia, Beijing, China {tage, lecu, fuwei, mingzhou}@microsoft.com, {chbb, szf}@pku.edu.cn Abstract This paper introduces a new resource called EventWiki which is, to the best of our knowledge, the first knowledge base resource of major events. In contrast to most existing knowledge bases that focus on static entities such as people, locations and organizations, our EventWiki concentrate on major events, in which all entries in EventWiki are important events in mankind history. We demonstrate that EventWiki is a very useful resource for information extraction regarding events in Natural Language Processing (NLP), knowledge inference and automatic knowledge base construction. Keywords: EventWiki, event knowledge base, major event, resource, event-driven, information extraction, knowledge base con- struction, knowledge inference 1. Introduction Knowledge bases (KBs) are large collections of facts about entities such as people, organizations and locations, which have attracted growing interest in recent years because they are useful in various studies and applications such as question-answering, semantic parsing, information extrac- tion and knowledge inference. Most existing knowledge bases (e.g., Freebase) focus on static entities (e.g., people, locations, and organizations) and their relations. Few of them, however, pay much atten- tion to events and event-related knowledge; instead, they just treat events as normal entities and do not specially dis- tinguish them, which makes it hard to identify event en- tries. Even though some facts in KBs can be treated as events, they tend to entity-centric and insufficient to de- scribe major and complicated events (e.g., 2011 Tohoku earthquake and 2014 World Cup). In fact, event knowl- edge is very important for many applications and its related tasks (e.g., event extraction and event detection) have been extensively studied in both NLP and Data mining commu- nity in the past decades. Unfortunately, due to the lack of rich event-related knowledge and annotations, studies re- lated to event knowledge discovery are either limited (e.g., in event extraction evaluation under Automatic Content Ex- traction project1, only events of 33 predefined event types are studied) or shallow (e.g., clustering-based event detec- tion). To provide rich event knowledge and a dataset for event- related studies, this paper introduces a new knowledge base resource – EventWiki is a knowledge base of events, which collects 21,275 major events of 95 types from Wikipedia2. We will show in the following section that EventWiki con- tains rich information about events (e.g., event taxonomy, event schemas and event relations), which makes it useful for a variety of studies and applications such as informa- tion extraction, knowledge inference and knowledge base construction. To the best of our knowledge, EventWiki is the first knowledge base of events which covers such large 1http://projects.ldc.upenn.edu/ace/ 2https://en.wikipedia.org numbers of major event entries. 2. EventWiki Compared to static entities (e.g., person, location and orga- nization), there is less work focusing on event knowledge discovery due to a lack of event-centric knowledge bases. Although …

Content truncated to 500 words