Semantic Question Answering System over Linked Data using Relational Patterns

Titelaufnahme

Titel
Semantic Question Answering System over Linked Data using Relational Patterns
Verfasser
Hakimov, Sherzod ; Tunc, Hakan ; Akimaliev, Marlen ; Dogdu, Erdogan
Erschienen
2013
Sprache
Englisch
Dokumenttyp
Konferenzband
URN
urn:nbn:de:0070-pub-26574549

Zugriffsbeschränkung

Links

Dateien

Klassifikation

Abstract

Question answering is the task of answering questions in naturallanguage. Linked Data project and Semantic Web communitymade it possible for us to query structured knowledge bases likeDBpedia and YAGO. Only expert users, however, with theknowledge of RDF and ontology definitions can build correctSPARQL queries for querying knowledge bases formally. In thispaper, we present a method for mapping natural languagequestions to ontology-based structured queries to retrieve directanswers from open knowledge bases (linked data). Our tool isbased on translating natural language questions into RDF triplepatterns using the dependency tree of the question text. Inaddition, our method uses relational patterns extracted from theWeb. We tested our tool using questions from QALD-2, QuestionAnswering over Linked Data challenge track and found promisingpreliminary results.

Statistik