Pattern Structures for Knowledge Processing and Information Retrieval

Abstract

In this work, we consider an application of pattern structures to the task of semantic information retrieval. We review the existing techniques of lattice-based information retrieval and introduce a novel approach to defining similarity of documents and search queries. This approach is based on the combination of the well-known embedding technique and pattern structures to provide a coherent assessment of document similarity. The experimental evaluation performed on a subset of users’ reviews collected from Amazon demonstrates the effectiveness of the combination of pattern structures with the introduced modifications compared to other lattice-based techniques.