A probabilistic inference model for information retrieval

摘要

Based on the epistemological view of probability, a probabilistic inference model is proposed in this paper. It is argued that some of the problems presented in the conventional probabilistic models may be resolved if one takes the epistemological view instead of the aleatory view of probability. The new model also recognizes the close connections between probabilistic and vector-based approaches. It is explicitly shown that both the standard and generalized vector space models are special cases of the proposed probabilistic inference model. Our method is applicable to any basic concepts chosen from an indexing scheme. Furthermore, it provides a theoretical basis for the vector-based relevance feedback models, and has the additional advantage that no ad hoc parameters are needed. The new approach gives a plausible explanation for adopting a linear measure (discriminant function) for decision making in general, although our discussion is mainly focussed on information retrieval.