Analysis of Image Search Queries on the Web: Query Modification Patterns and Semantic Attributes

摘要

This study investigated query modification patterns and semantic attributes in queries executed during user searches for images on the web. Its purpose was to identify whether query modification patterns were related to users' contextual factors and content sources as well as whether the patterns characterize the use of semantic attributes expressed in users' search queries in an interactive web-searching process. To this end, a collection of 970 image search queries executed by undergraduate students in a naturalistic setting was analyzed. The study's findings showed that query modification patterns were significantly associated with content sources. Among the types of query modification employed, "Reformulation" and "New" were the most frequently used. Terms related to format, object, or place associated with an image also were found to be frequently used in search queries. Terms referring to type or genre were the most frequent attributes in web image search queries, which suggests a change from previous findings of search queries in a professional context. Implications are discussed in terms of search assistants for web image searches and semantic annotation to improve image indexing.