Self-organizing documentary maps for information retrieval

This paper presents a neural-network-based approach for information retrieval, an important issue in natural language processing. We first describe a method for creating self-organizing documentary maps - visible and continuous representations in which all queries and documents are mapped in topological order according to their similarities. We then show that documents related to queries can be retrieved by merely calculating the Euclidean distances between the positions at which queries and documents are placed and choosing the N closest documents in the ranking order for each query. Small-scale computer experiments have demonstrated that the proposed method is capable of high precision.