Exploiting Parallel News Streams for Unsupervised Event Extraction

Exploiting Parallel News Streams for Unsupervised Event Extraction Congle Zhang, Stephen Soderland & Daniel S. Weld Computer Science & Engineering University of Washington Seattle, WA 98195, USA {clzhang, soderlan, weld}@cs.washington.edu Abstract Most approaches to relation extraction, the task of extracting ground facts from natural language text, are based on machine learning and thus starved by scarce training data. Man- ual annotation is too expensive to scale to a comprehensive set of relations. Distant super- vision, which automatically creates training data, only works with relations that already populate a knowledge base (KB). Unfortu- nately, KBs such as FreeBase rarely cover event relations (e.g. “person travels to loca- tion”). Thus, the problem of extracting a wide range of events — e.g., from news streams — is an important, open challenge. This paper introduces NEWSSPIKE-RE, a novel, unsupervised algorithm that discovers event relations and then learns to extract them. NEWSSPIKE-RE uses a novel probabilistic graphical model to cluster sentences describ- ing similar events from parallel news streams. These clusters then comprise training data for the extractor. Our evaluation shows that NEWSSPIKE-RE generates high quality train- ing sentences and learns extractors that per- form much better than rival approaches, more than doubling the area under a precision-recall curve compared to Universal Schemas. 1 Introduction Relation extraction, the process of extracting struc- tured information from natural language text, grows increasingly important for Web search and ques- tion answering. Traditional supervised approaches, which can achieve high precision and recall, are lim- ited by the cost of labeling training data and are un- likely to scale to the thousands of relations on the Web. Another approach, distant supervision (Craven and Kumlien, 1999; Wu and Weld, 2007), creates its own training data by matching the ground instances of a Knowledge base (KB) (e.g. Freebase) to the un- labeled text. Unfortunately, while distant supervision can work well in some situations, the method is limited to rela- tively static facts (e.g., born-in(person, location) or capital-of(location,location)) where there is a cor- responding knowledge base. But what about dy- namic event relations (also known as fluents), such as travel-to(person, location) or fire(organization, person)? Since these time-dependent facts are ephemeral, they are rarely stored in a pre-existing KB. At the same time, knowledge of real-time events is crucial for making informed decisions in fields like finance and politics. Indeed, news stories report events almost exclusively, so learning to ex- tract events is an important open problem. This paper develops a new unsupervised tech- nique, NEWSSPIKE-RE, to both discover event rela- tions and extract them with high precision. The in- tuition underlying NEWSSPIKE-RE is that the text of articles from two different news sources are not independent, since they are each conditioned on the same real-world events. By looking for rarely de- scribed entities that suddenly “spike” in popularity on a given date, one can identify paraphrases. Such temporal correspondence (Zhang and Weld, 2013) allow one to cluster diverse sentences, and the re- sulting clusters may be used to form training data in order …

Content truncated to 500 words