Information retrieval via WWW image searching: A reality check

On the one hand, information storage and retrieval through the Internet has made spectacular progress. On the other hand, practical searching for information still confronts us with retrieval systems that are far from perfect. The purpose of this contribution is to clarify the practical state of the art of finding information through image web search systems in a particular subject domain and to derive recommendations from the outcome of the investigation. The method of investigation consisted of submitting queries to public access WWW image search engines followed by a quantitative evaluation of the 20 individual results that were ranked highest for each search query; the scoring method leads to a positive score related to precision between 0 and 20 and to a negative score related to misinformation between 0 and minus 20. This has lead to the following findings: Most positive scores were much lower than the ideal maximum, even though the queries contained very specific words and names as formulated by a user familiar with the subject domain. In other words, many image search results lead to information that is irrelevant, so that users lose time and may lose interest. Furthermore and worse, a significant number of search results may even mislead users who are unfamiliar with the selected subject domain. In conclusion, information retrieval from the WWW through image searching is attractive, simple and fast, but precision is far from perfect and users can be misinformed. Recommendations are formulated to authors, publishers, librarians and end-users of image information retrieval systems.