NLP Configuration
Configure the NLP engine you want to use. You can also adjust the possible recognition threshold and the definite score threshold to improve the accuracy of the intent detection.
Here you can choose the AI Classifier that will be used and the language that will be managed. In V1, nAIxus support IBM Watson and SSAI Masterbert (our own classifier).
Please contact your nAIxus contact to obtain the credentials to use.
Threshold Settings
The intention detection model, as well as the entity extraction model, are based on probabilistic approaches. This means that the models are almost never 100% sure. Instead, they assign probability scores to each output. Depending on your application and the training data available, you may wish to lower or raise this threshold.
If you set the threshold too high, very few user messages will be assigned the appropriate intent. Your customers will receive many "I didn't understand that" messages (which you can configure in the bot settings "Misunderstanding"). Keep in mind that a high threshold will result in more accurate bot responses to user input, if your expressions are many and varied. Setting a threshold too low will make the bot more confident, but potentially more false as well. A low threshold increases the chances of a false link between an expression and an intent.
The possible recognition threshold is the minimum score required for an intent to be considered as a possible match. The definite score threshold is the minimum score required for an intent to be considered as a definite match.
Hesitation
We understand that words have different meanings based on the context of its usage in the sentence. If we talk about human languages, then they are ambiguous too because many words can be interpreted in multiple ways depending upon the context of their occurrence. Word sense disambiguation, in natural language processing (NLP), may be defined as the ability to determine which meaning of word is activated by the use of word in a particular context. This present scenario is called whenever the system cannot determine which is the most relevant scenario to trigger.
