Natural Language Questions for the Web of Data

Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pages 379–390, Jeju Island, Korea, 12–14 July 2012. c⃝2012 Association for Computational Linguistics Natural Language Questions for the Web of Data Mohamed Yahya1, Klaus Berberich1, Shady Elbassuoni2 Maya Ramanath3, Volker Tresp4, Gerhard Weikum1 1 Max Planck Institute for Informatics, Germany 2 Qatar Computing Research Institute 3 Dept. of CSE, IIT-Delhi, India 4 Siemens AG, Corporate Technology, Munich, Germany {myahya,kberberi,weikum}@mpi-inf.mpg.de selbassuoni@qf.org.qa ramanath@cse.iitd.ac.in volker.tresp@siemens.com Abstract The Linked Data initiative comprises struc- tured databases in the Semantic-Web data model RDF. Exploring this heterogeneous data by structured query languages is tedious and error-prone even for skilled users. To ease the task, this paper presents a methodology for translating natural language questions into structured SPARQL queries over linked-data sources. Our method is based on an integer linear pro- gram to solve several disambiguation tasks jointly: the segmentation of questions into phrases; the mapping of phrases to semantic entities, classes, and relations; and the con- struction of SPARQL triple patterns. Our so- lution harnesses the rich type system provided by knowledge bases in the web of linked data, to constrain our semantic-coherence objective function. We present experiments on both the question translation and the resulting query answering. 1 Introduction 1.1 Motivation Recently, very large, structured, and semantically rich knowledge bases have become available. Ex- amples are Yago (Suchanek et al., 2007), DBpe- dia (Auer et al., 2007), and Freebase (Bollacker et al., 2008). DBpedia forms the nucleus of the Web of Linked Data (Heath and Bizer, 2011), which inter- connects hundreds of RDF data sources with a total of 30 billion subject-property-object (SPO) triples. The diversity of linked-data sources and their high heterogeneity make it difficult for humans to search and discover relevant information. As linked data is in RDF format, the standard approach would be to run structured queries in triple-pattern-based lan- guages like SPARQL, but only expert programmers are able to precisely specify their information needs and cope with the high heterogeneity of the data (and absence or very high complexity of schema in- formation). For less initiated users the only option to query this rich data is by keyword search (e.g., via services like sig.ma (Tummarello et al., 2010)). None of these approaches is satisfactory. Instead, the by far most convenient approach would be to search in knowledge bases and the Web of linked data by means of natural-language questions. As an example, consider a quiz question like “Which female actor played in Casablanca and is married to a writer who was born in Rome?”. The answer could be found by querying sev- eral linked data sources together, like the IMDB- style LinkedMDB movie database and the DB- pedia knowledge base, exploiting that there are entity-level sameAs links between these collections. One can think of different formulations of the example question, such as “Which actress from Casablanca is married to a writer from Rome?”. A possible SPARQL formulation, assuming a user fa- miliar with the …

Content truncated to 500 words