Provenance for Linked Data

Assessing the

quality of linked data

currently published on the Web is a crucial need of various data-intensive applications. Extensive work on similar applications for relational data and queries has shown that

data provenance

can be used in order to compute trustworthiness, reputation and reliability of query results, based on the source data and query operators involved in their derivation. In particular,

abstract provenance models

can be employed to record information about source data and query operators during query evaluation, and later be used e.g., to assess trust for individual query results. In this paper, we investigate the extent to which relational provenance models can be leveraged for capturing the provenance of SPARQL queries over linked data, and identify their limitations. To overcome these limitations, we advocate the need for new provenance models that capture the full expressive power of SPARQL, and can be used to support assessment of various forms of data quality for linked data manipulated declaratively by such queries.