IDOL can analyze the structured fields and elements of a set of documents to find patterns in that data for the identification of trends or anomalous behavior. The analysis can take a number of forms and the implementation largely depends on the nature of the problem that you want to solve.
In one aspect of Parametric Refinement, you can successively improve retrieval by selecting parametric values that allow a user to narrow the results to the exact record they want.
When searching for a wine, choosing Riesling as the grape, immediately rules out all red wines, and eliminates Spain, and Italy as producing countries.
But what if the user then asked for a Riesling wine from Alsace for under $10 only to find that there were none fitting those criteria?
Ordinarily the search would end there, but Proximate Analysis can automatically identify other wines that it believes might fit the requirements. In proximate analysis, the parametric selections are viewed as fuzzy rather than rigid filters.
IDOL can determine that the wines of the Alsace are strongly similar to those from Germany, and indeed some Australian regions also use the Riesling grape. Proximate Analysis can then return results that match a selection of Rieslings from Germany and Australia, as well as locating Alsace Rieslings at $12.
|
|