Modelling a Query Space Using Associations

We all use our associative memory constantly. Words and concepts form paths that we can follow to find new related concepts; for example, when we think about a car we may associate it with driving, roads or Japan, a country that produces cars. In this paper we present an approach for information modelling that is derived from human associative memory. The idea is to create a network of concepts where the links model the strength of the association between the concepts instead of, for example, semantics. The network, called association network, can be learned with an unsupervised network learning algorithm using concept co-occurrences, frequencies and concept distances. The possibility to create the network with unsupervised learning brings a great benefit when compared to semantic networks, where the ontology development usually requires a lot of manual labour. We present a case where the associations bring benefits over semantics due to easier implementation and the overall concept. The case focuses on a business intelligence search engine where we modelled its query space using association modelling. We utilised the model in information retrieval and system development