Compared with search queries, which are usually composed of a few keywords, natural language questions can demonstrate detailed information needs through searchers' richer expressions. This study aims to provide understandings of ordinary people's image needs in their daily life, by analyzing 474 questions obtained from a social question and answer (social Q&A) site. The study found that image needs reflected through the natural language questions contain several components: context of image needs (motive and intervening variables), image attributes (descriptive metadata, syntactic, and semantic attributes), and associated information (information on known/similar/comparative images and related stories). Characteristics of each component of image needs were analyzed, and accordingly image-indexing guidelines were suggested. Because image needs comprise diverse attributes, a single indexing approach might not support all complex needs for images. Therefore, this study proposes that different indexing approaches should be integrated for enhancing keyword search and browsing effectiveness. Such approaches include descriptive metadata assigned by a creator and/or automatic algorithms, user-assigned tags (or users' reactions), indexing through associated text, and content-based image retrieval.
|Number of pages||13|
|Journal||Journal of the American Society for Information Science and Technology|
|State||Published - Nov 2011|