LinkWiper – A System For Data Quality in Linked Open Data

Abstract

Linked Open Data (LOD) provides access to large amounts of data on Web. These data sets range from high quality curated data sets to low quality sets. LOD sources often need strategies to clean up data and provide methodology for quality assessment in linked data. They allow interlinking and integrating any kind of data on the web. Links between various data sources enable software applications to operate over the aggregated data space as if it is a unique local database. However, such links may be broken, leading to data quality problems. In this thesis we present LinkWiper, an automated system for cleaning data in LOD. While this thesis focuses on problems related to dereferenced links, LinkWiper can be used to tackle any other data quality problem such as duplication and consistency. The proposed system includes two major phases. The first phase uses information retrieval-like search techniques to recommend sets of alternative links. The second phase adopts crowdsourcing mechanisms to involve workers (or users) in improving the quality of the LOD sources. We provide an implementation of LinkWiper over DBPedia, a community effort to extract structured information from Wikipedia and make this information using LOD principles. We also conduct extensive experiments to illustrate the efficiency and high precision of the proposed approach.