Working Principles
It uses the FastText: an open-source, free, lightweight library that allows users to learn text representations and text classifiers. It works on standard, generic hardware. Models can later be reduce
The services eligibility criteria (text) is passed through a data processing pipeline in which the following take place
Each word meaning lookup
Text striping and cleaning
Tokenisation
lemmatisation
The vocabulary is then built upon the lemmas upon which a model is trained for 10 epochs
The eligibility is calculated using ratio of highest similarity of user information corpus to the built vocabulary.
The eligibility threshold is set for 0.45 by default which can be altered between 0 and 1 by the organisation.
Last updated
Was this helpful?