att_abstract={{Classifying web queries  to predefined target categories,also known as web query classification [10], is important to improve search relevance and also for targeted online advertising. However, web queries are typically short, ambiguous and in constant flux. Moreover, often the target categories lack semantic descriptions. These challenges make the web query classification a non-trivial challenge. In this paper, we present two complementary approaches - enrichment and reductionist - for web query classification. The enrichment method uses World Wide Web (WWW) to build word clouds of target categories, which are then compared to web queries and further uses bayesian framework to classify a given query. The reductionist approach, also referred as the centroid method, works by reducing web queries to few central tokens that rep- resents the query. The two approaches are complementary to each other as the enrichment method exhibit high recall but low precision, whereas the centroid method has high precision but low recall. We evaluate the two approaches based on 1300 human labelled web queries.}},
	att_authors={ra727a, ik089v, rz649p},
	att_categories={C_IIS.10, C_CCF.5},
	att_copyright_notice={{The definitive version was published in ICONIP2011. {{, 2011-11-17}}
	att_tags={query classification,  enrichment,  centroid,  precision,  recall},
	author={Ritesh Agrawal and Irwin King and Remi Zajac and Xiaofeng Yu},
	title={{Enrichment and Reductionism: Two Complementary Approaches for Web Query Classification}},