Citation Provenance
Research Area: Digital Libraries Year: 2012
Type of Publication: Technical Report Keywords: citation analysis, citation provenance, source of citation, citation classification
  • Heng Low Wee
Low Wee's FYP report
Citations in research paper acknowledges previous work and gives the provenance to key ideas in the cited paper. However, it is difficult for a reader to locate the cited information that justifies a citation without first investing time to read through the cited paper. We investigate Citation Provenance, a new task in citation analysis, which means to discover the origin of the information embodied by a citation. We first describe the challenges in collecting annotations for our training set, and present a two-tier approach in tackling this problem. We adopt features previously used in Citation Classification and Information Retrieval tasks, and with them, we differentiate citations that refer to the whole paper in general (general) versus ones that cite specific claims, evidence or parts of the paper (specific). Given that a citation is specific, our second tier classifier localizes the cited information in the cited paper. Our first tier (GvS) obtained an accuracy of 0.786 in cross- validation evaluation and our second tier (LocateProv) showed an improvement over our baseline.
Digital version