This paper presents the semantic vector space model, a text representation and searching technique based on heuristic syntax parsing and distributed representation of semantic case structures. In this model, documents and queries are represented as semantic matrices, and relevancy prediction is achieved by computing the similarity of such matrices. A prototype system was built to implement this model and used in an experimental study. The preliminary results of the study showed that with longer documents and queries, especially when original documents were used as queries, the system based on our technique had significantly better performance than the VSM based SMART system in terms of recall, precision, and effectiveness of relevance ranking.