Cyberia: Multiple Viewpoints

Information Retrieval and Exploration through Multiple Viewpoint Systems

A viewpoint is any representation of a document collection which provides a method of judging the relevance of subsets of that collection to a given query. Most information retrieval systems are examples of single viewpoints. A multiple viewpont system combines more than one viewpoint in a search to take advantage of different relevance judgements.

Multiple viewpoints are particularly well suited to addressing the problem of vocabulary mismatch. Frequently a searcher frames the description of an information need -- the query -- in a vocabulary that does not communicate the distinguishing characteristics of the information need to the system, due to misunderstanding or ignorance of how the system represents and searches within a document collection. Using more than one viewpoint can improve the situation by increasing the likelihood of vocabulary overlap between user and system, and by providing opportunities for exploring and better understanding the collection by moving among the viewpoints.

The Evoc Project is an example of multiple viewpoints used in the text-based search of a store of scientific datasets. Multi-Channel CBIR" is an example of applying multiple viewpoints to the realm of image retrieval.