Abstract. This paper presents a short survey of fuzzy and neural approaches to Information Retrieval. The goal of such approaches is to de ne exible Information Retrieval Systems able to deal with the inherent vagueness and uncertainty of the retrieval process. In this survey we address if and how some approaches met their goal. 1.

Abstract: Most models for Information Retrieval (IR) using neural networks are simple spreading activation models. Some of them were successfully applied to real world document collections. Nevertheless, they do not exploit the subsymbolic paradigma of neural processing. In this paper a model using a simple backpropagation network for IR is proposed. The COSIMIR model implements the central process in IR. It is a backpropagation network which calculates the similarity between a document and a query representation. The similarity function is learned through examples. Hence, it implements a cognitive similarity function. The first evaluation demonstrates that COSIMIR works well for short vectors. 1

Ordering principles of digital libraries expressed in ontologies may be highly heterogeneous even within a domain and especially over different cultures. Automatic methods for mappings between different ontologies are necessary to ensure successful retrieval of information stored in virtual digital libraries. Text categorization has discussed learning methods to map between full text terms and thesaurus descriptors. This article reports some experiments for the mapping between different ontologies and shows further that fusion methods which have been successfully applied to ad-hoc information retrieval can also be employed for text categorization.

Abstract: Neural networks are well suited for Information Retrieval (IR) from large text or multimedia databases. Their capacity for tolerant and intuitive processing offers new perspectives in IR where the vague nature of human relevance judgements has confronted theory and systems with considerable problems. Most models use the keyword representation vector as input or output. However, fulltext indexing brings forth large vectors which are difficult to handle for neural networks. This article discusses methods for dimensionality reduction used in IR and applies one of them, Latent Semantic Indexing (LSI) to information retrieval using a neural backpropagation network. The transformation between two representation schemes is enabled through preprocessing by LSI which is based on Singular Value Decomposition (SVD). 1

This paper presents partial results of an experimental investigation concerning the use of Neural Networks in associative adaptive Information Retrieval. The learning and generalisation capabilities of the Backpropagation learning procedure are used to build up and employ application domain knowledge in the form of a sub-symbolic knowledge representation. The knowledge is acquired from examples of queries and relevant documents of the collection. In this paper the architecture of the system is presented and the results of the experimentation are briefly reported. 1 Query Adaptation using a Subsymbolic Representation of the Application Domain Knowledge Recent research work in Information Retrieval (IR) suggests that significant improvements in retrieval performance requires techniques that, in some sense, "understand" the content of documents and queries. Recently IR researcher have tried to use application domain knowledge to determine relevant relationships between documents and quer...