Clustering search results of non-text user generated content

Non-text user generated content (UGC), such as videos and images, is usually searched by metadata. Metadata, such as title, tags, and description, is created by users whenever content is uploaded. However, in many cases metadata can have multiple meanings. This requires users to spend time sifting through a long list of search results until they can find all the content for which they were actually looking. In order to address this limitation, we suggest an algorithm to cluster search results using keyword similarity. Clustering search results from YouTube are accomplished by using the Markov clustering algorithm, which helps users to quickly and easily find what they want. Finally, we conclude by evaluating the performance results of our clustering algorithm.