FAQs

A.

Graph Analysis consist in creating a synthetic profile for each participant in a connected graph, based on the connections (directed or not), and it's connections' connections, and its connections' connections' connections...  you get the idea.

Graph Analysis allows you to answer a large variety of questions :

• In the existing Graph, what connections are the most important ?
• With which elements could we create new connections that are not in the graph (i.e. Who should I follow on twitter, what URL could be meaningful for this blog article, given a set of URLs I know,...)
• On which blogs should I try to get a backlink ?
• And the classical classification question.

And probably other questions you may have that we haven't yet thought of ! With eXenGine, you can perform this analysis on a really huge amount of data, and with unprecedented relevance and speed. Our technology relevance reaches the levels obtained with very expensive methods, like Alternating Least Squares or Gradient Descent approaches, while being much faster. Additionnaly, our methods allow to inject external knowledge during the learning phase. It can be knowledge from details semantic analysis, knowledge from cross-linguistic resources, or knowledge from Text Mining or Behavioural Analysis.

A.

Test mining allows you to answer a large variety of questions :

• What terms are relevant for a given document ?
• What documents are similar to this one ?
• What documents are relevant for these terms ?
• To what class does this document belong ?

And probably other questions you may have that we haven't yet thought of ! With eXenGine, you can perform this analysis on a really huge amount of data, and with unprecedented relevance and speed. Our technology relevance reaches the levels obtained with very expensive methods, like latest deep learning "word2vec" approaches, while being much faster. Additionnaly, our methods allow to inject external knowledge during the learning phase. It can be knowledge from details semantic analysis, knowledge from cross-linguistic resources, or knowledge from Graph Analysis or Behavioural Analysis.

A.

NCISC is a Stochastic Matrix Factorization algorithm of the RandNLA family. It has been discovered circa 2009 by Guillaume Pitel, the founder of eXenSa. We haven't published anything about it yet, in the secret hope that its amazing capabilities would allow us to have a huge advantage in the field of recommender systems and text analysis.