The combined use of bibliographic coupling and cocitation for document retrieval

A linkage similarity measure which takes into account both the bibliographic coupling of documents and their cocitations (both cited and citing papers) produced improved document retrieval over a measure based only on bibliographic coupling. The test collection consisted of 1712 papers whose relevance to specific queries had been judged by users. To evaluate the effect of using cocitation data, we calculated for each query two measures of similarity between each relevant paper and every other paper retrieved. Papers were then sorted by the similarity measures, producing two ordered lists. We then compared the resulting predictions of relevance, partial relevance, and non‐relevance to the user's evaluations of the same papers. Over‐all, the change from the bibliographic coupling measure to the linkage similarity measure, representing the introduction of cocitation data, resulted in better retrieval performance.