Skip to content

TAL (NLU)

The Automatic Language Processing (NLU) engine is a very important brick in the assistant if it processes requests in natural language. He is in charge of understanding and interpreting what the user says. The NLU engine will analyze the incoming messages to define their meaning and extract structured information that a machine can understand and analyze.

We are convinced that there will never be THE best NLU for all use cases. Prisme.ai has several NLU engines that can be changed with a single click depending on its use case. (See Ratings for help choosing the best NLU engine for your use case).

** The NLU Switch functionality of Prisme.ai allows you to: **

  • Take advantage of natural language processing thanks to NLU engines: text analysis to extract an intention, recognition of named entities, sentiment analysis model, etc.,
  • Decorrelate your knowledge base, workflow, business rules ... from the NLU so as not to compromise your investment
  • Directly modify your NLU engine with a simple button.
  • Select the best NLU for your use case,
  • Add other NLU engines,

You have the choice between the following 9 variants:

  • Prism.NLU
  • Prism.NLU - Duckling
  • Snips-NLU
  • Snips-NLU - Duckling
  • Rasa TensorFLow
  • Rasa TensorFLow - Duckling
  • NLP.js
  • Flair
  • Flair - Duckling

NLU

The confidence rating of your NLU engine is customizable. The higher the rate (eg 0.85), the more strict the NLU engine will be in interpreting intentions. Conversely (example 0.2) the lower it is, the more flexible it will be in the interpretation; it is important to properly adjust this rate in order to optimize the bot's performance.

NLU-Switch-Mecanism