Google Scholar or Scirus for scholarly material on the web? An evaluation and comparison of retrieval effectiveness.

This thesis evaluates and compares the retrieval effectiveness of the two search engines Google Scholar and Scirus, which both specialises in indexing scholarly material on the web as well as journal articles. Thirty search queries are used and the first twenty retrieved documents for each query are examined for topical relevance and scholarly content. The search queries are based on legitimate information needs taken from a scientific questionbox on the web. Queries are expressed with advanced search operators and are the same for both search engines. Precise relevance criteria are set up and a binary scale is used when judging. Inactive, duplicate, and mirror links are all considered non-relevant, as well as documents judged as non-scholarly. Two methods of measuring precision are used: average precision at each DCV for all search queries, and average precision at each DCV for each search query; which both credit ranking of relevant documents. As for scholarly content, both relevant and non-relevant documents are judged, and results are expressed in per cent. When judging scholarly content, guiding principles are being followed. There is not a great difference in results between the two search engines, although Google Scholar provides slightly higher results for all precision measures as well as being the best in retrieving scholarly material. However, Scirus is the better at retrieving relevant documents at DCV = 1 and DCV = 2, as showed by average precision at each DCV for all search queries.Uppsatsnivå: