Results (
Thai) 1:
[Copy]Copied!
When starting the application, the GSI tab first shows a tree of attributes and attribute values that can be used to preselect a subset of products to be used. In this way, the user can rule out products of which she is certain she is not interested in. Since the maps of the GSI are dynamically generated, that is, they are dependent on user input and the randomization process, we do not show one here. Using another MP3 player catalog, it was shown in [23] that the theoretical quality (in terms of Stress, see (17.5)) of these maps is high, mainly since they only show about eight products. Also, the recommendation algorithm was tested in a simulation study, showing promising results.In [23], also a usability study among 71 respondents was reported. Results of this study showed that compared to a traditional list-based interface, the users were not significantly less satisfied with the GSI, while they found it more complex. However, it was only the first time the respondents used the GSI. Some weaknesses of the GSI reported in that usability study have been taken into account in the GSI prototype discussed here. Since the graphical recommender tab was switched off during the experiment, users missed a way to formulate a query. Secondly, people missed a way to restrict the search space, which is facilitated by the selection tree in the new prototype.
Being translated, please wait..