CPL - Chalmers Publication Library

Calendar-Based Relevance Boosting

Författare och institution:
Marcus Christiansson (Institutionen för data- och informationsteknik (Chalmers)); Peter Wintzell (Institutionen för data- och informationsteknik (Chalmers))
Antal sidor:
56
Publikationstyp:
Examensarbete på avancerad nivå
Program:
Masterarbete ()
Förlag:
Chalmers University of Technology
Förlagsort:
Göteborg
Publiceringsår:
2012
Språk:
engelska
Fulltextlänk (lokalt arkiv):
Sammanfattning (abstract):
Information retrieval systems have become a more central component in modern Software Engineering. It is hence important to improve the models that determine relevance, both concerning what data is used and how to use the data, particularly to support more personalized systems. Among the multiplicity of models to determine relevance, contextbased search aims to improve the relevance of retrieved information by considering in what context a search is being carried out. This study evaluates how improved search relevance can be achieved by considering the context of the calendar. Hence, a model concept was constructed, implemented in a typical enterprise search environment and furthermore evaluated using statistical testing. The results showed a statistically significant improvement on the search relevance using keyword matching. Nevertheless, temporal methods of determining context had no significant impact on the search relevance. Future works include evaluation of the concept in an authentic environment to determine concept applicability.
Ämne (baseras på Högskoleverkets indelning av forskningsämnen):
NATURVETENSKAP ->
Data- och informationsvetenskap
Nyckelord:
information retrieval, context-based search, calendar context search calendar-based relevance, information retrieval evaluation
Postens nummer:
164577
Posten skapad:
2012-10-10 16:25
Posten ändrad:
2013-04-04 09:59

Visa i Endnote-format