Web 2.0 dictionary

How might we benefit from the billions of tagged multimedia files (e.g. image, video, audio) available on the Internet? This paper presents a new concept called Web 2.0 Dictionary, a dynamic dictionary that takes advantage of, and is in fact built from, the huge database of tags available on the Web. The Web 2.0 Dictionary distinguishes itself from the traditional dictionary in six main ways: (1) it is fully automatic because it downloads tags from the Web and inserts this new information into the dictionary; (2) it is dynamic because each time a new shared image/video is uploaded, a "bag-of-tags" corresponding to the image/video will be downloaded, thus updating Web 2.0 Dictionary. The Web 2.0 Dictionary is literally updating every second, which is not true of the traditional dictionary; (3) it integrates all kinds of languages (e.g. English, Chinese), as long as the images/videos are tagged with words from such languages; (4) it is built by distilling a small amount of useful information from a massive and noisy tag database maintained by the entire Internet community, therefore the relatively small amount of noise present in the database will not affect it; (5) it truly reflects the most prevalent and relevant explanations in the world, unaffected by majoritarian views and political leanings. It is a real, free dictionary. Unlike Wikipedia" [5] which can be easily revised by even a single person, the Web 2.0 Dictionary is very stable because its contents are informed by a whole community of users that upload photo/videos; (6) it provides a correlation value between every two words ranging from 0 to 1. The correlation values stored in the dictionary have wide applications. We demonstrate the effectiveness of the Web 2.0 Dictionary for image/video understanding and retrieval, object categorization, tagging recommendation, etc, in this paper.