Conceptual Space Dimensions

Analytics indexes in Relativity are based on concept spaces. Documents in a training space are mapped onto a multi-dimensional space. Documents that are closer together in the conceptual space will be considered to be more conceptual similar. Every document and every term will have a spatial relationship to every other document and term. A conceptual rank (or score) is assigned to measure the distance between two terms in keyword expansion or two documents in categorization. The higher the score, the closer to the queried for term or document.

When creating a new analytics index, in the advanced settings an admin will be able to change the number of dimensions in the conceptual space. The default setting is 100. Increasing the number of dimensions, will allow for more exact conceptual relationships to be made between documents and terms. Conceptual similarities can be more precisely defined. However, increasing the conceptual value has diminishing returns after a certain point. An increase in the number of dimensions will take up more RAM on the server.

Relativity recommends not increasing the number of dimensions to more than 300.