Information Retrieval can be regarded an inference or evidential reasoning process in which we estimate the probability that a user’s information need, expressed as a query, is met by a document as “evidence”.25 The techniques required to support these kinds of inference are similar to those used in expert systems that must reason with uncertain information. Some of the inference models developed for expert systems can be adapted to the document retrieval task. The Bayesian inference network has been used to:
- Support multiple document representation schemes. Research has shown that, even when retrieval against each individual representation has similar performance, documents that are retrieved using multiple representations have higher relevance;
- Allow the results from different queries and query types to be combined. Given a single natural language description of an information need, different searchers will formulate different queries to represent that need. The same searcher may also formulate multiple queries for the same need, each based on a different strategy. Even when average performance is similar for each query, documents retrieved by multiple queries are more likely to be relevant; and
- Facilitate flexible matching between the terms or concepts mentioned in queries and those assigned to documents. The poor match between vocabulary used to express queries and the vocabulary used to represent documents appears to be a major cause of poor recall. Recall can be improved by using domain knowledge to match query and representation concepts without significantly degrading precision.
25Turtle H and Croft W B, Evaluation of an inference network-based retrieval model, ACM Transactions on Information Systems. Volume 9 , Issue 3 (July 1991) Special issue on research and development in information retrieval Pages: 187 – 222 ISSN:1046-8188