Determining best offers for a user

Video: demonstration video

This chain deals with user dependent search, i.e. finding most relevant objects for many actual users in parallel. Every user can have unique, different notion of suitable objects. We represent this notion as user preferences based on fuzzy sets. User preferences can be obtained directly from user, learned from ranked objects or generated from user index. They are subsequently used for finding most relevant objects. This approach is combined with fulltext searching for keywords.

Description

The chain consists of four integrated methods:

It begins with registering a new user and determining his/her preferences. The user interface displays a questionnaire with personal data and an optional applet which allows user to define preferences explicitly. Preferences are then stored in the user ontology and passed to Top-k tool which finds most relevant objects.

User interface displays these objects to the user and offers a possibility to rate them in a scale from worst to best. Submitted ratings are processed by tool IGAP that learns global preferences and thus refines user profile. New preferences are again passed to Top-K to find new list of relevant objects.

An alternative way of searching is to input keywords into a text field. Tool JDBSearch finds objects that contain these keywords. If the user decides to combine fulltext search with preferential search, the results are also ordered according to the specified preferences.

This process is depicted on the following figure.

Sequence diagram

Sequence diagram of implemented chain.

References

  1. Gurský, P., Horváth, T., Jirásek, J., Krajči, S., Novotný, R., Vaneková, V., Vojtáš, P.: Knowledge Processing for Web Search – An Integrated Model. Studies in Computational Intelligence (vol. 78), Springer, 2007, ISSN 1860-949X, ISBN 978-3-540-74929-5, pp. 95-104.
  2. Gurský, P., Horváth, T., Novotný, R., Vaneková, V., Vojtáš, P.: UPRE: User Preference Based Search System. 2006 IEEE/WIC/ACM International Conference on Web Intelligence (WI'06), pages 841-844, IEEE Computer Society 2006, ISBN 0-7695-2747-7.