Concept

DualNAVI manual

Concept

DualNAVI Interaction Model

DualNAVI's most distinctive features are summarized by two types of duality. One is the dual view of retrieval results, in which topic word graphs provide summary information about the retrieved results (see technical points for details). Another one is the different query types DualNAVI accepts, search by key documents together with search by key words (see technical points for details).
The following figure illustrates how these features are closely related.

You can start with either of search by words or search by key documents. If you find interesting documents in the title list, you can proceed to next search with these interesting documents.
If you find interesting words in the topic word graph, you can proceed to next search with these interesting words.

Another advantageous point of dual-view interface is the cross checking function. By selecting interesting documents, you can easily find what topic words are related to them, and vice versa.

Recovering Order through Metrics

The objective of DualNAVI is not only to support information retrieval but to recover order in the huge amount of information by introducing associative metrics both in the document space and in the word space. The associative similarity between two documents is measured by the weighted amount of shared words and vice versa. Under this consideration of duality between documents and words, we can re-consider the meaning of each operation. For example, the topic word selection is considered as the dual operation of document retrieval.

References

"Interactive Document Search with DualNAVI" (1999) in Proceedings of NTCIR Workshop 1 organized by NACSIS (now National Institute of Informatics). (downloadable)

DualNAVI was introduced as Concept albums in "Souped-up search engines" by Declan Butler, a news feature article of Nature May-11-2000.

DualNAVI was selected as Innovative Product by RIAO'2000.

Concept