Skip to Main Content
Information is increasingly being distributed in the form of dynamic streams instead of static web pages. It began with news RSS feeds, but with the emergence of social media services such as twitter and facebook, now encompasses instant status updates as well as shared links to various types of web content. While one of the challenging tasks in using such stream based services is to search quality streams of interests, existing work has mainly focused on the retrieval models for individual posts or classification frameworks for blogs, leaving the problems arising in building a dedicated stream search engine in real-world settings largely unexplored. This paper presents a novel stream search engine, named FeedMil, that can satisfy the need for retrieving quality streams of topical relevance for the purpose of subscription. Through addressing the issues unique to the stream search problem, FeedMil is able to give a new search experience that is focused on quality and topic relevance beyond just a sim-ple query matching, enabling users to quickly discover high quality but less popular streams located in the long tail of millions of streams.