On the Frameworks for Information Retrieval Modeling

Relevance, as shown by various cognitive and experimental studies, depends on a number of situational factors other than topicality. Formal IR models, on the other hand, only consider relevance from a topicality point of view. There is an urgent need of appropriate models which are able to consider relevance in a broader context. This paper is an attempt to find a theoretical tool suitable to the broader context. The Boolean model and the probabilistic model are reviewed with respect to some situational factors, showing their inability for the task. Conditional logic is then suggested as a potential alternative. Relevance is interpreted as a causal relation between a document and a query in a given context.