| Summary: | 博士 === 國立臺灣師範大學 === 圖書資訊學研究所 === 101 === With the wide application of Web 2.0, various social networking platforms allow taggers to use uncontrolled, free keywords (i.e., social tags) to organize information. In library and information science, professional indexers are guided by the principles of authority control and thesaurus control to organize information with controlled vocabularies. Both social taggers and professional indexers regard keywords as concepts that represent their cognitions and mental models of information content, according to their prior experience and knowledge. Existing studies have focused on examining the sources and usage of individual keywords, and comparing the similarity between tags and controlled vocabularies. However, the results of such studies only reflect scattered debris rather than a whole picture of the mental models used by social taggers and professional indexers for article indexing. A better understanding of the mental models of taggers and professional indexers and their usage gap may inspire better selection of appropriate keywords for organizing information, facilitating resource discovery, and guiding users to find the right information.
This study explores the mental models used by taggers and professional indexers to designate keywords for article indexing. Using a dataset of 3,972 CiteULike tags and 6,708 Library and Information Science Abstracts (LISA) descriptors from 1,489 scholarly articles in 13 library and information science journals, this study attempts to analyze the keyword usage of taggers and professional indexers to capture and build up their mental models for article indexing, and generalize their structures and patterns. To achieve this end, in this study social network analysis and frequent-pattern growth methods were employed. When measured with respect to terms used, power law distribution, a comparison of terms used as tags and descriptors, social network analysis (including centrality, overall structure and role equivalence) and frequent-pattern growth analysis (including frequent-pattern tree), little similarity was found between the mental models of taggers and professional indexers in article indexing.
The results of this study are summarized as follows:
Taggers’ mental models for article indexing are more diverse than those of professional indexers.
Social taggers have a higher preference than professional indexers to select terms for article indexing from title keywords.
There is little similarity between social tags and controlled vocabularies and they complement each other.
Keywords in content-related categories were not used independently by social taggers, but they were often used with those from topic-related categories. On the other hand, keywords of other-related categories were often co-used with those of title-, topic- or content-related categories by professional indexers.
Social taggers may prefer to assign co-occurring keywords with more sets of fewer facets’ viewpoints (almost always two-facets); however, professional indexers may be inclined to offer keywords with fewer sets of more facets’ viewpoints (i.e., two-, three- and seven-facets).
Social taggers may be inclined to assign keywords with fewer path-based rules comprising fewer keyword categories. Professional indexers may tend to offer keywords with more path-based rules comprising more keyword categories.
According to the research results mentioned above, the key contributions of this study are as follows:
Development of a generic model of mental models of social taggers and professional indexers for article indexing.
Analysis of the structures and patterns embedded in maps of mental models of social taggers and professional indexers in article indexing.
Analysis of the characteristics of keyword usage and co-occurring keywords’ associations.
Presentation of a theoretical basis to explain the reason why social tags complement controlled vocabularies.
Extension of the tag category model by feasibility examination and explanation.
Furthermore, the results of this study also inform the design of information systems, including term recommendations and user interfaces for indexing, as well as frequent-pattern based classification trees for browsing and navigation.
|
|---|
