Show simple item record

dc.contributor.authorFarid, Mina 16:05:59 (GMT) 16:05:59 (GMT)
dc.description.abstractThe RDF data model has become a prevalent format to represent heterogeneous data because of its versatility. The capability of dismantling information from its native formats and representing it in triple format offers a simple yet powerful way of modelling data that is obtained from multiple sources. In addition, the triple format and schema constraints of the RDF model make the RDF data easy to process as labeled, directed graphs. This graph representation of RDF data supports higher-level analytics by enabling querying using different techniques and querying languages, e.g., SPARQL. Anlaytics that require structured data are supported by transforming the graph data on-the-fly to populate the target schema that is needed for downstream analysis. These target schemas are defined by downstream applications according to their information need. The flexibility of RDF data brings two main challenges. First, the extraction of RDF data is a complex task that may involve domain expertise about the information required to be extracted for different applications. Another significant aspect of analyzing RDF data is its quality, which depends on multiple factors including the reliability of data sources and the accuracy of the extraction systems. The quality of the analysis depends mainly on the quality of the underlying data. Therefore, evaluating and improving the quality of RDF data has a direct effect on the correctness of downstream analytics. This work presents multiple approaches related to the extraction and quality evaluation of RDF data. To cope with the large amounts of data that needs to be extracted, we present DSTLR, a scalable framework to extract RDF triples from semi-structured and unstructured data sources. For rare entities that fall on the long tail of information, there may not be enough signals to support high-confidence extraction. Towards this problem, we present an approach to estimate property values for long tail entities. We also present multiple algorithms and approaches that focus on the quality of RDF data. These include discovering quality constraints from RDF data, and utilizing machine learning techniques to repair errors in RDF data.en
dc.publisherUniversity of Waterlooen
dc.subjectdata qualityen
dc.subjectinformation extractionen
dc.titleExtracting and Cleaning RDF Dataen
dc.typeDoctoral Thesisen
dc.pendingfalse R. Cheriton School of Computer Scienceen Scienceen of Waterlooen
uws-etd.degreeDoctor of Philosophyen
uws.contributor.advisorIlyas, Ihab
uws.contributor.affiliation1Faculty of Mathematicsen

Files in this item


This item appears in the following Collection(s)

Show simple item record


University of Waterloo Library
200 University Avenue West
Waterloo, Ontario, Canada N2L 3G1
519 888 4883

All items in UWSpace are protected by copyright, with all rights reserved.

DSpace software

Service outages