Deriving Ontology from Folksonomy and Controlled Vocabulary

Chen, Miao; Qin, Jian

Permalink

https://hdl.handle.net/2142/15143 Copy

Description

Title

Deriving Ontology from Folksonomy and Controlled Vocabulary

Author(s)

• Chen, Miao
• Qin, Jian

Issue Date

2008-02-28

Keyword(s)

• Ontology
• folksonomy
• tag
• controlled vocabulary
• vector space

Abstract

Popular online tagging websites, such as Flickr, Technorati, and Del.icio.us, allow users to tag objects freely without constraints of any controlled vocabulary. The word “folksonomy” has been used to describe this type of grass-root taxonomies, which provides a rich source for building ontologies. Research has experimented with building ontology purely from user-generated tags, and the approaches have been focused primarily on statistical methods. Schmitz (2006) conducted a study of inducing an ontology from Flickr tags, which used statistical methods to detect subsumption pairs based on co-occurrences of tags. Heymann Carcia-Molina (2006) established a hierarchical taxonomy based on tags from Delicious and CiteULike respectively by using cosine similarity of tag vectors. Although a good source for building ontology, folksonomy has its disadvantage in representing object content. For example, the tags are subjective and their hyponyms are usually not indicated (Mathes, 2004). Relations between tags are unknown and useful information is missed in tags, therefore the unorganized status of tags affects retrieval of objects negatively. Folksonomies alone are not sufficient for building a comprehensive and high-quality ontology. Researchers have suggested using other sources such as WordNet and Wikipedia to assist ontology building from folksonomies (Damme et al., 2007). Contrary to folksonomies, controlled vocabulary is characterized by rigid structures and slow responsiveness to new terminology. But its systematic organization and careful formulation of terms and relationships would be complementary to the disadvantages of folksonomies. Online lexical resources and gazetteers as instances of controlled vocabulary appear to be promising approaches in using folksonomies for generating ontologies (Schmitz, 2006; Damme et al., 2007). While researchers are speculating this method, little research has been done to actually implement it to build ontology. In addition, building ontologies from a combination of both folksonomies and controlled vocabulary is rarely mentioned in previous works. This study is intended to fill the gaps between methodologies in using folksonomies to produce ontologies.

Type of Resource

text

Permalink

http://hdl.handle.net/2142/15143

Owning Collections