Our Apex NER

Dynamic Knowledge Graphs

 

Google’s mT5 Transformer is currently considered the most advanced neural network architecture. Our technology team have ‘extended’ mT5 with our own neural network ensemble, to now have a new Apex Named Entity Recognition (NER) solution that is unrivaled in the market on accuracy (F1 score 0.95) and able to perform NER tasks on 100+ languages.

One of the top Data & Analytics trends in business is the visualization of data via so called ‘Knowledge Graphs’ (KGs). We have started building our own KG (focus on the Financial Markets domain initially; we will connect all 41.000 listed entities globally), where our NER will provide the ’nodes’ (see the bubbles) and the output of our other AI microservices inside our FalconV platform such as Automated Text Classification, Duplicate Recognition and Related Content Detection will provide the ‘edges’ (see lines in visual).

Given also our platform’s ability to continually and organically capture new (web) text data the HIPSTO KG will become Dynamic.

There is a new NER deck out with more commercial use cases. Drop us a line if you would like to see it.