Towards Temporal Scoping of Relational Facts based on Wikipedia Data

Proceedings of the Eighteenth Conference on Computational Language Learning, pages 109–118, Baltimore, Maryland USA, June 26-27 2014. c⃝2014 Association for Computational Linguistics Temporal Scoping of Relational Facts based on Wikipedia Data Avirup Sil ∗ Computer and Information Sciences Temple University Philadelphia, PA 19122 avi@temple.edu Silviu Cucerzan Microsoft Research One Microsoft Way Redmond, WA 98052 silviu@microsoft.com Abstract Most previous work in information extraction from text has focused on named-entity recognition, entity linking, and relation extraction. Less attention has been paid given to extracting the temporal scope for relations between named entities; for example, the relation president-Of(John F. Kennedy, USA) is true only in the time-frame (January 20, 1961 - November 22, 1963). In this paper we present a system for temporal scoping of relational facts, which is trained on distant supervision based on the largest semi-structured resource available: Wikipedia. The system employs language models consisting of patterns automat- ically bootstrapped from Wikipedia sentences that contain the main entity of a page and slot-fillers extracted from the corresponding infoboxes. This proposed system achieves state-of-the-art results on 6 out of 7 relations on the benchmark Text Analysis Conference 2013 dataset for temporal slot filling (TSF), and out- performs the next best system in the TAC 2013 evaluation by more than 10 points. 1 Introduction Previous work on relation extraction (Agichtein and Gravano, 2000; Etzioni et al., 2004) by sys- tems such as NELL (Carlson et al., 2010), Know- ItAll (Etzioni et al., 2004) and YAGO (Suchanek et al., 2007) have targeted the extraction of en- tity tuples, such as president-Of(George W. Bush, USA), in order to build large knowl- edge bases of facts. These systems assume that relational facts are time-invariant. However, this assumption is not always true, for example ∗ This research was carried out during an internship at Microsoft Research. president-Of(George W. Bush, USA) holds within the time-frame (2001-2009) only. In this paper, we focus on the relatively less explored problem of attaching temporal scope to relation between entities. The Text Analysis Conference (TAC) introduced temporal slot filling (TSF) as one of the knowledge base population (KBP) tasks in 2013 (Dang and Surdeanu, 2013). The in- put to a TAC-TSF system is a binary relation e.g. per:spouse(Brad Pitt, Jennifer Aniston) and a document assumed to contain supporting evidence for the relation. The required output is a 4-tuple timestamp [T1, T2, T3, T4], where T1 and T2 are normalized dates that provide a range for the start date of the relation, and T3 and T4 provide the range for the end of the relationship. Sys- tems must also output the offsets of the text men- tions that support the temporal information ex- tracted. For example, from a text such as “Pitt married Jennifer Aniston on July 29, 2000 [...] the couple divorced five years later in 2005.”, a sys- tem must extract the normalized timestamp [2000- 07-29, 2000-07-29, 2005-01-01, 2005-12-31], to- gether with the entity and date offsets that support the timestamp. In this paper, we describe TSRF, a system for temporal …

Content truncated to 500 words