Queries in patent prior art search, being full patent applications, are very much longer than standard ad hoc search and web search topics. Standard information retrieval (IR) techniques are not entirely effective for patent prior art search because of the presence of ambiguous terms in these massive queries. Reducing patent queries by extracting small numbers of key terms has been shown to be ineffective mainly because it is not clear what the focus of the query is. An optimal query reduction algorithm must thus seek to retain the useful terms for retrieval favouring recall of relevant patents, but remove terms which impair retrieval effectiveness. We propose a new query reduction technique decomposing a patent application into constituent text segments and computing the Language Modeling (LM) similarities by calculating the probability of generating each segment from the top ranked documents. We reduce a patent query by removing the least similar segments from the query, hypothesizing that removal of segments most dissimilar to the pseudo-relevant documents can increase the precision of retrieval by removing nonuseful context, while still retaining the useful context to achieve high recall as well. Experiments on the patent prior art search collection CLEF-IP 2010, show that the proposed method outperforms standard pseudo relevance feedback (PRF) and a naive method of query reduction based on removal of unit frequency terms (UFTs).