Biography-Dependent Collaborative Entity Archiving for Slot Filling

Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 665–675, Lisbon, Portugal, 17-21 September 2015. c⃝2015 Association for Computational Linguistics. Biography-Dependent Collaborative Entity Archiving for Slot Filling Yu Hong† Soochow University & Rensselaer Polytechnic Institute tianxianer@gmail.com Xiaobin Wang†, Yadong Chen, Jian Wang Soochow University Suzhou, CHN (215006) czwangxiaobin@gmail.com Tongtao Zhang, Heng Ji† Rensselaer Polytechnic Institute NY, USA (12180) jih@rpi.edu Abstract Knowledge Base Population (KBP) tasks, such as slot filling, show the particular importance of entity-oriented automatic relevant document acquisition. Rich, diverse and reliable relevant documents satisfy the fundamental require- ment that a KBP system explores the nature of an entity. Towards the bottleneck problem be- tween comprehensiveness and definiteness of acquisition, we propose a collaborative archiv- ing method. In particular we introduce topic modeling methodologies into entity biography profiling, so as to build a bridge between fuzzy and exact matching. On one side, we employ the topics in a small-scale high-quality relevant documents (i.e., exact matching re- sults) to summarize the life slices of a target entity (i.e., biography), and on the other side, we use the biography as a reliable reference material to detect new truly relevant docu- ments from a large-scale partially complete pseudo-feedback (i.e., fuzzy matching results). We leverage the archiving method to enhance slot filling systems. Experiments on KBP cor- pus show significant improvement over state- of-the-art. 1 Introduction Entity archiving is an entity-oriented document retrieval task. Towards a target entity of a specif- ic type, such as the ones discussed in this paper, a person or an organization, the goal of entity archiving is to search and collect all relevant documents from large-scale data sets under lim- ited prior knowledge of the entity. We limit our study to the regular English entity archiving, in which the prior knowledge contains the com- monly used full name (formatted by English enti- ty naming criteria) along with a gold-standard reference document, such as a news story on President “George W. Bush”. Entity archiving plays a fundamental role in KBP tasks. It narrows down the range of the data source for knowledge discovery to small-scale closely related documents. Such documents, on one hand, contain informative content on a target entity, which is extremely favorable for back- ground knowledge extraction. On the other hand, the documents provide definitive evidence for verifying the claimed identity of the entity. As for KBP slot filling and verification tasks (Surdeanu and Ji, 2014), the archived relevant documents to an entity provide sufficient con- texts (provenances) of the concrete instances (fillers) of the entity attributes (slots). See Figure 1, in which both the provenances support filler extraction, meanwhile the provenance 1 addi- tionally provides the evidence to verify the fillers (e.g., is episcopalism the true religion of Bush?). Figure 1: Use of entity archiving in slot filling The main challenges of entity archiving are as Global data (e.g., all web docu- ments) on PERSON entities KBP Slot Filling task Target entity: George W. Bush Slots: Title, Religion, Nation, etc. Filler Extraction …

Content truncated to 500 words