| Relevance is a fundamental concept in information science. However, among the large body of relevance literature, very little of the research studied users' relevance judgments about video. The study described in this dissertation investigated what kinds of criteria people use to select videos. A semi-structured interviewing method was applied in the study to take advantage of its exploratory and descriptive nature. Twenty-six participants in three user task groups (illustration, collection, and production) were interviewed. Three categories of relevance judgment criteria were identified: textual (e.g., topicality, date, and nationality), visual (e.g., cinematography, objects/events, and style), and implicit (e.g., interest, familiarity, and appropriateness). Among the 36 relevance criteria found in this study, topicality was still considered the most important criterion for video relevance judgments, which is consistent with the findings from textual relevance literature. However, its degree of importance was much lower compared to the textual relevance literature. The participants also applied more diverse criteria to select videos, and other frequently mentioned criteria included scene-level information, author, genre, price, and review. In addition, strong task effects were found among the three user task groups, and within each group there were also other factors that affected the participants' relevance judgments, such as user experiences, user domains, and video genres. The results not only enrich the relevance literature but also provide implications for video interface design and video retrieval, as well as for video metadata research. |
