Nicht alle Hemden sind blau: Diversifying search results in online shopping
(master thesis or bachelor project)
This thesis is in collaboration with FINDOLOGIC, a search engine for online shops. FINDOLOGIC motivates the problem as follows:
Eine Suche nach "hemd" in einem Bekleidungs-Shop findet massenhaft Hemden, die alle gleich relevant sind. Im schlechtesten Fall kommen auf der ersten Seite der Ergebnisse nur blaue Herren-Hemden. Um für solche allgemeinen Anfragen einen guten Überblick über das Artikelsortiment zu liefern ist eine Durchmischung der Artikel wünschenswert, so dass z.B. Hemden in verschiedenen Farben gelistet werden.
In large online shops, each time a query is issued, a search result consists of thousands of items. Assuming that all of them are relevant, one has to decide how to display them to the user. Simple sorting by relevance causes to display very similar items together. One way of diversifying the results is to group similar items and provide the user with the groups (or group representatives) only.
The goal of this thesis is to develop techniques to increase the diversity of the search results in online shopping. The following tasks are to be performed:
- study the related work on diversifying search results,
- find and developing techniques suitable in online shopping domain,
- implement a prototype search engine.
Contact: Nikolaus Augsten
In cooperation with:
