Document representations and clues to document relevance

Abstract

Research into the role of document representations in the relevance judgment process has focused on the ability of users to predict the relevance of documents based on various document representations. Conclusions have been stated as to the comparative effectiveness of various document representations, but there has been little exploration into why certain document representations seem to enable users to better predict the relevance of documents. This examination is an attempt to identify the extent to which various document representations contain clues that allow users to determine the presence or absence of traits and/or qualities that determine the relevance of the document to the user's situation. Motivated users discussed their reasons for pursuing or not pursuing documents based on information contained within representations of those documents (i.e., titles, abstracts, indexing terms, etc.). The results are presented as the co-occurrence of respondents' mentions of various traits and/or qualities, and the document representations that led to such responses. It is concluded that document representations may differ in their effectiveness as indicators of potential relevance because different types of document representations vary in their ability to present clues for specific traits and/or qualities. Suggestions for further research are provided.