@chriskoerner

Searchstrings revealing user intent: a better understanding of user perception

, , , and . ICWE '06: Proceedings of the 6th international conference on Web engineering, page 225--232. New York, NY, USA, ACM, (2006)
DOI: http://doi.acm.org/10.1145/1145581.1145628

Abstract

The evaluation of information driven websites by analysis of serverside available data is the objective of our approach. In our former work we developed techniques for evaluation of non-transactional websites by regarding the author's intentions and using only based on implicit user feedback. In several case studies we got aware that in single cases unsatisfied users had been evaluated positively. This divergence could be explained by not having considered the user's intentions. We propose in this approach to integrate search queries within referrer informaiton as freely available information about the user's intentions. By integrating this new source of information into our meta model of website structure, content and author intention, we enhance the formerly developed web success metric GPI. We apply well understood techniques such as PLSA for text categorization. Based on the latent semantic we construct a new indicator evaluating the website with respect to the user intention. By ranking all webpages with respect to the user intention manifested in the search query, we acchieve an individualized measure to evaluate a session by the user's initial intention. In contrast to manual assignments of weights by the website author, our proposed measure is purely calculated allowing a generic assessment of websites without manual intervention.In a case study we can show, that this indicator evaluates the quality and usability of a website more accurately by taking the user's goals under consideration. We can also show, that the initially mentioned diverging user sessions, can now be assessed according to the user's perception.Due to limited information on the host side, without direct access to the client side, still some assumptions remain to be made.

Description

Searchstrings revealing user intent

Links and resources

Tags

community

  • @chriskoerner
  • @dblp
@chriskoerner's tags highlighted