LOD Query

SINA: Semantically INterpreting user query towards question-Answering

Motivation

Knowledge bases published on the Web of Data can be used to answer challenging questions about a wide range of topics. They usually offer a SPARQL endpoint, which can be queried by the user. However, a novice user needs to acquire both knowledge about the underlying ontology and proficiency in formulating SPARQL queries in order to actually formulate a SPARQL query. In this project, we are working on constructing SPARQL queries based on user-supplied keywords. As a result ordinary users are empowered to formulate relatively expressive queries without knowledge of SPARQL and the underlying ontology structure.

Overview

The following figure shows a birds-eye-view of the envisioned research. Based on a set of user-supplied keywords, candidate IRIs (Internationalized Resource Identier) for each of the keywords issued by the user is computed. Then, by using an inference mechanism, a subgraph based on the identied IRIs is extracted and represented to the user as the answer.

First Phase

In an early phase, a novel method for generating SPARQL queries based on two user-supplied keywords was proposed. Since this method is based on simple operations, it can generate SPARQL queries very efficiently. Also it is completely agnostic of the underlying knowledge base as well as its ontology schema. We currently use DBpedia as the underlying knowledge base, but this method is easily transferable to the whole Data Web.
The implementation of this method is publicly available at: http://lod-query.aksw.org/

Contributers in this phase:

• Saeedeh Shekarpour
• Sören Auer
• Axel Ngonga Ngomo
• Daniel Gerber
• Sebastian Hellman
• Clause Stadler

Second Phase

In the next phase, we proposed a method for generating SPARQL queries for arbitrary number of keywords. Its base is an inference mechanism embedded in a graph traversal approach which both accurately and efficiently builds SPARQL queries for the straightforward interpretations of the user's queries. Implementation of this method is also available at: http://sina.aksw.org.
This application generats SPARQL query for those keyword-based queries which can be converted to a conjunctive query. A conjunctive query is a conjunction of triple patterns.

Publishing Datasets used in Evaluation

we compared the accuracy of our approach against the top-k exploration on a data set composed of 30 SPARQL queries extracted from the DBpedia query log3 and the QALD-1 benchmark4.
* SPARQL queries dataset: http://aksw.org/Projects/lodquery/files?get=PublishedDBpediaQueries.xlsx
In our second experimental setting, we evaluated the precision and recall of our framework on the QALD-1 benchmark.
* Keyword-based queries dataset: http://aksw.org/Projects/lodquery/files?get=PublishedQualdkeywordqueries.xlsx

Contributers in this phase:

• Saeedeh Shekarpour
• Sören Auer
• Axel Ngonga Ngomo

Further Information

* Do not hesitate to contact me for any question (shekarpourATinformatik.uni-leipzig.de – sa.shekarpourATgmail.com)
* SINA is a Persian name: http://en.wikipedia.org/wiki/Sina
* SINA is the name of a Persian physician and philosopher which his Latinized name is Avicenna:

• http://www.iranchamber.com/personalities/asina/abu_ali_sina.php
• http://en.wikipedia.org/wiki/Avicenna

Contact

Saeedeh Shekarpour Johannisgasse 26, Zimmer 5-22 04103 Leipzig Tel.:   +49 341 97 32341 E-Mail, Research Group, Workpage

Dr. Axel-C. Ngonga Ngomo Johannisgasse 26, Zimmer 5-22 04103 Leipzig Tel.:   +49 341 97-32341 E-Mail, Workpage

Dr. Sören Auer Johannisgasse 26, Zimmer 5-09 04103 Leipzig Tel.:   +49 341 97-32367 E-Mail, Homepage, Research Group, rdfs:seeAlso