Question Answering over Linked Data Using First-order Logic

Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1092–1103, October 25-29, 2014, Doha, Qatar. c⃝2014 Association for Computational Linguistics Question Answering over Linked Data Using First-order Logic∗ Shizhu He, Kang Liu, Yuanzhe Zhang, Liheng Xu and Jun Zhao National Laboratory of Pattern Recognition Institute of Automation, Chinese Academy of Sciences, Beijing, 100190, China {shizhu.he, kliu, yzzhang, lhxu, jzhao}@nlpr.ia.ac.cn Abstract Question Answering over Linked Data (QALD) aims to evaluate a question an- swering system over structured data, the key objective of which is to translate questions posed using natural language into structured queries. This technique can help common users to directly ac- cess open-structured knowledge on the Web and, accordingly, has attracted much attention. To this end, we propose a novel method using first-order logic. We formulate the knowledge for resolving the ambiguities in the main three steps of QALD (phrase detection, phrase-to- semantic-item mapping and semantic item grouping) as first-order logic clauses in a Markov Logic Network. All clauses can then produce interacted effects in a unified framework and can jointly resolve all am- biguities. Moreover, our method adopts a pattern-learning strategy for semantic item grouping. In this way, our method can cover more text expressions and answer more questions than previous methods us- ing manually designed patterns. The ex- perimental results using open benchmarks demonstrate the effectiveness of the pro- posed method. 1 Introduction With the rapid development of the Web of Data, many RDF datasets have been published as Linked Data (Bizer et al., 2009), such as DBpedia (Auer et al., 2007), Freebase (Bollacker et al., 2008) and YAGO (Suchanek et al., 2007). The grow- ing amount of Linked Data contains a wealth of knowledge, including entities, classes and rela- tions. Moreover, these linked data usually have ∗Shizhu He and Kang Liu have equal contribution to this work. complex structures and are highly heterogeneous. As a result, there are gaps for users regarding ac- cess. Although a few experts can write queries us- ing structured languages (such as SPARQL) based on their needs, this skill cannot be easily utilized by common users (Christina and Freitas, 2014). Thus, providing user-friendly, simple interfaces to access these linked data becomes increasingly more urgent. Because of this, question answering over linked data (QALD) (Walter et al., 2012) has recently received much interest, and most studies on this topic have focused on translating natural lan- guage questions into structured queries (Freitas and Curry, 2014; Yahya et al., 2012; Unger et al., 2012; Shekarpour et al., 2013; Yahya et al., 2013; Bao et al., 2014; Zou et al., 2014). For example, with respect to the question “Which software has been developed by organi- zations founded in California, USA?”, the aim is to automatically convert this utterance into an SPARQL query that contains the follow- ing subject-property-object (SPO) triple format: ⟨?url rdf:type dbo:Software, ?url dbo:developer ?x1, ?x1 rdf:type dbo:Company, ?x1 dbo:foundationPlace dbr:California⟩1. To fulfill this objective, existing systems (Lopez et al., 2006; Unger et al., 2012; Yahya et al., …

Content truncated to 500 words