Rocchio’s Formula

Definition

Rocchio’s formula is used to determine the query term weights of the terms in the new query when Rocchio’s relevance feedback algorithm is applied.

Key Points

In 1971, Rocchio proposed a classical query expansion algorithm based on the Vector Space model [1]. The basic algorithm assumes that the user identifies a set R of relevant documents and a set N of non relevant documents and the improved query is the result of a linear combination of the mean frequencies tf of the terms in the original query and in these two sets (the centroids of R and N), that is the weight of each term in the new query is:

$$qtf_m = \alpha .qtf + \beta .\sum\limits_{d \in R} {tf - {\rm \gamma }{\rm .}\sum\limits_{d \in N} {tf} }$$

Feedback variations assume that positive feedback exhibits a much clear impact on the reformulation of the query. Also positive feedback can be applied to expand query without the explicit feedback from the user (blind relevance feedback). Blind relevance feedback can be...