Query Expansion/post filtering

The need for query expansion or result filter.

The term collection preferred by users of searching engine could be quite different from the term collection pre-defined the documents. For example, a company named "crystal Jade Kitchen" is considered as "crystal Jade" by users of search engine.

One way is to break down the terms used in document into its minimal units, For example, instead of treat "crystal Jade Kitchen" as a term, break it down into 3 terms "crystal", "jade" and "Kitchen". But learn concepts (automatically or manully) and chain the small units of terms into big one, say "crystal jade" and form a concept <"crystal jade", restruant>, here the restruant is the class name for "crystal jade".

The key is to learn the terms used by users and mapped it into the term space in the collection of document.

Let td[i] 1<=i<=M is a set of terms used in the documents, while tq[j] 1<=j<=H. The tq[j]=concept[| m<=i<=M}]

references:
[1]Improving Document Retrieval by Automatic Query Expansion Using Collaborative Learning of Term-Based Concept
[2]Query Reformulation with Collaborative Concept-Based Expansion