TagSense: Marrying Folksonomy and Ontology

TitleTagSense: Marrying Folksonomy and Ontology
Publication TypeThesis
Year of Publication2004
AuthorsZixin Wu
Academic DepartmentDepartment of Computer Science
Number of Pages72
UniversityUniversity of Georgia
CityAthens
Thesis TypeMS
Abstract

Tagging communities is a featured Web 2.0 phenomenon, where users describe a Web resource by using keywords (called tags). This behavior can be viewed as cooperative meta-data extraction and annotation. While these tagging communities become more and more popular, their information retrieval mechanism is still keyword based, thus it is difficult to achieve high precision and recall rates because of word ambiguity and lack of semantics. In this thesis, we combTagging communities is a featured Web 2.0 phenomenon, where users describe a Web resource by using keywords (called tags). This behavior can be viewed as cooperative meta-data extraction and annotation. While these tagging communities become more and more popular, their information retrieval mechanism is still keyword based, thus it is difficult to achieve high precision and recall rates because of word ambiguity and lack of semantics. In this thesis, we combine the approaches of folksonomy and ontology to improve recall rate and ranking of query results. We index Web resources by the meanings of tags instead of strings, and we provide context-aware multi-ontologies semantic search capability which utilizes relationships in ontologies. The evaluations performed on a subset of Flickr's photos indicate that our users spent significantly less time and effort in finding what they want on our system than on Google Desktop on the same datasets.ine the approaches of folksonomy and ontology to improve recall rate and ranking of query results. We index Web resources by the meanings of tags instead of strings, and we provide context-aware multi-ontologies semantic search capability which utilizes relationships in ontologies. The evaluations performed on a subset of Flickr's photos indicate that our users spent significantly less time and effort in finding what they want on our system than on Google Desktop on the same datasets.

Related Files: