Object-Level Vertical Search

Current web search engines essentially conduct document-level ranking and retrieval. However, structured information about realworld objects embedded in static webpages and online databases exists in huge amounts. We explore a new paradigm to enable web search at the object level in this paper, extracting and integrating web information for objects relevant to a specific application domain. We then rank these objects in terms of their relevance and popularity in answering user queries. In this paper, we introduce the overview and core technologies of object-level vertical search engines that have been implemented in two working systems: Libra Academic Search (http://libra.msra.cn) and Windows Live Product Search (http://products.live.com).