Total Ranking: g-RAK


Principle

This component maximises its available functionality when used with the component g-REQ, Requests Generator.

The ranking consists of determining the distance, in terms of meaning, between a wide request (i.e. a request and its graph of assumptions) and a group of texts.

These texts can be in the form of complex documents or the results from a search directed towards one or more engines. Ranking makes it possible to calculate the relevance of the results of a search compared to a request.

The compared texts can be complex objects, consisting of several fields (for example: title, description, author, URL). g-RAK makes it possible to give an importance weighting to each one of these fields.


Applications


g-RAK is used to:

Create a list ordered by document “descriptors”
From the results of a search, presented in the order in which they were found (random, by date, other…), it is possible to add sorting by relevance. In addition results can be generated in ways others than that of entered text search, for example, by navigation, classification, etc…

Create applications for Federated Search
The Score (indication of relevance) generally provided by a search engine is information which is not standardised. The relevance of this value, being no more than the measuring unit employed, cannot be compared with the results from another search engine. A popular application for g-RAK is post-ranking, which makes it possible to reconcile heterogeneous lists of results.


Example

Request: “biography of Pablo Picasso”