1. GoRelations: an Intuitive Query System for DBPedia Lushan Han and Tim Finin 15 November 2011 http://ebiq.org/j/93

2. Overview Introduction Related Work Query Interface Translation Evaluation Examples

3. Semantic web technologies allow machines to share data and knowledge using common web language and protocols. ~ 1997 2007 2008 2009 Linked Open Data Cloud 2007-11 2010 2011: 31B facts in 295 datasets interlinked by 504M assertions on ckan.netckan.net LOD is the new Cyc: a common source of background knowledge Uses Semantic Web Technology to publish shared data & knowledge Data is inter- linked to support inte- gration and fusion of knowledge

4. Dbpedia as LOD DBpedia is an important example of Linked Open Data – Extracts structured data from Infoboxes in Wikipedia – Stores in RDF using custom ontologies & more (e.g., Yago terms) The major integration point for the entire LOD cloud DBpedia

5. Hard for People to Query Querying DBpedia requires a lot of a user: – Understand the RDF model – Master SPARQL, a formal query language – Explore the large number of ontology terms, 320 classes and 1,600 properties – Deal with term heterogeneity (Place vs. PopulatedPlace) – Know relevant URIs (~1M entities !) Querying a large LOD is overwhelming Natural language query systems still a research goal

6. Goal Develop a system allowing a user with a basic understanding of RDF to query DBpedia and ultimately distributed LOD collections – To explore what data is in the system – To get answers to question – To create SPARQL queries for reuse or adaptation Desiderata – Easy to learn and to use – Good accuracy (e.g., precision and recall) – Fast

7. Related Work Natural Language Interfaces (NLI) systems – ORAKEL user interactions, small, closed-domain ontologies – FREyA user interactions – PANTO automatic, small, closed-domain ontologies – PowerAqua automatic, open domain, no SPARQL All have difficulties in understanding complex questions and rely on lexical matching to find candidate terms Keyword interface (and Watson?) – No support for complex queries

8. Key Idea Reduce the complexity of the problem by: – Having user enter a simple graph, and – Annotating the nodes and arcs with words and phrases

9. Semantic Graph Interface Nodes denote entities and links binary relations between them Entities described by two unrestricted terms: its name or value and its concept in the query context Users flag entities for results with a ? and those not with a * A compromise between NLI and SPARQL – Users provide compositional structure of the question – Free to use their own terms in annotating the structure

10. Translation – Step One finding semantically similar ontology terms For each concept or relation in the semantic graph, generate a list of the k candidate ontology classes or properties that are most semantically similar

11. Another Example

12. Measure of reasonableness for an interpretation is the degree to which its ontology terms associate in the way that their corresponding user terms connect in the semantic graph Joint disambiguation Resolving direction Compute the reasonableness for a single link Translation – Step Two disambiguation algorithm

13. The translation of a semantic graph query to SPARQL is straightforward given the mappings SPARQL Generation Concepts Place => Place Author => Writer Book => Book Relations born in => birthPlace wrote => author

14. Preliminary Evaluation 33 test questions from 2011 Workshop on Question Answering over Linked Data answerable using DBpedia Three human subjects unfamiliar with DBpedia translated the test questions into semantic graph queries Compared with two top NL QA systems: PowerAqua and True KnowledgePowerAqua True Knowledge

15. http://ebiq.org/GOR

21. Conclusion and Future Work Baseline system works well for DBpedia Ongoing and future work – Better Web interface – Allow user feedback and advice – Add entity matching – Extend to a distributed LOD collection For more information, see http://ebiq.org/93