Axil: a Python implementation of Contextual Network Graph

Axil (AXiom Indexing Library) is a indexing library for Divmod’s Axiom database. It is an implementation of a graph based text classification alghorithm known as CNG (Contextual Network Graphs).

Contextual Network Graphs are a modified version of what was prior refferd to as Spreading Activation Network algorithms: they are a method for implement classification strategies that are been evaluated as qualitative equivalent to LSI (Latent Semantic Indexing) or naive-baesyan classifiers, without many of their limitations (as term-document matrix recalculation as in LSI or continuos-learnig phase as naive-baesyan classifiers).

First there’ns a support for many languages (notably many european and all asian languages) in all of the indexing phases (in particular there is a stemming implementation only for EN and IT languages); this lack of support implies the lack of a chance to implement “term reduction” strategies in early stage of the process, resulting in a loss of effectivness in subsequents phases.

Furthermore document term extraction (splitting/stemming) implementations are written in pure Python and are slow comparing to C/Python based parsers;

[more in Italian]

Leave a Reply