Den interaktiva IR-forskningens genomslag i söktjänsterna på webben
SöktjänsterInformationsåtervinningDatorbaserad informationssökning - bibliotekDocumentationInformationLibrary scienceArchivisticsArkiv- och bibliotekskunskapDokumentationSocial Sciences
The theoretical focus in this master is information retrieval (IR), and especially the part of IR that includes interactivity. This research is by some researchers called IIR. Our principal aim was to examine if search engines on the WWW are taking advantage of research results from IIR research. A way to do this is to look at the application of special features, in this case query expansion (QE) and relevance feedback (RF). Both these methods aim for a better matching of documents and the searchers information need. Another feature of these methods is that they can include the user in the process. To accomplish this goal we performed a literature study and a small e-mail survey. We included the seven largest search engines and three others because of their special features. We determined which search engines that should be in our survey by search traffic volume. Only two of the requested search engines replied to our questionnaire, MSN Search and Terra Lycos. Our theoretical studies have shown that there are benefits with QE and RF. These benefits are foremost in the efficiency of the search engines. Theory describes many ways to incorporate these methods in online systems. We could find neither RF nor QE at MSN Search. Lycos have a feature called Suggested search , which is a kind of QE based on other searchers choices. Another common feature on the Web is more like this . The user points out a document as relevant in the same way as in RF, but we don t consider this being an interactive process. One reason to use RF is that the user is the only one that knows what is relevant, an important concept in our study. There are two main schools in IIR, one tries to simulate human behaviour and the other wants to stimulate the user. QE can be done either manual, automatic or interactive. We think that the interactive method is in line with the stimulation school and the best alternative. A problem for both users and researchers is that the information about search features given by the search engines is poor. The terminology used is also very diverse. Our thought is that a good cooperation between search companies and the academic world can result in better search services. International conferences, e.g. TREC, already takes place and can serve as platforms for a dialog. Dialog is also a keyword for the components, user and system.