En utvärdering av två pearl growing-metoder i ISI Web of Science
The purpose of this thesis is to evaluate two pearl growing methods in ISI Web of Science. The methods tested are Find Related Records FRR, an automatic service that uses bibliographic coupling, and Author Keywords AK, where searches are performed using the two first keywords from this field. Both search methods were used on twenty initial documents and the first twenty hits were evaluated for relevance using a binary scale 0,1. The relevance criteria were based on specific information needs connected to every initial document. The measures used to evaluate the effectiveness of the search methods were Precision at DCV=10, P10, and Uninterpolated Average Precision AP. Wilcoxons signed rank test was used to evaluate if there were any significant differences between the two methods. The results showed that FRR achieved the highest average values for both P10 and AP. However, neither of the methods excelled in effectiveness, something that the Wilcoxons test corroborated. When it comes to FRRs results one could state that bibliographic coupling does not always perform well when a specific information need is requested. One reason for AKs results is that keywords can be too wide or too narrow and consequently affect the retrieval of information. If a search starts with a relevant initial document, pearl growing can be a useful search method, especially if both FRR and AK are combined.