The platform is integrated with Dialogflow, Google’s NLU engine for natural language understanding and analysis. Natural language understanding is what enables the chatbot to understand user requests without the need for all possible wording to have been previously mapped. The chatbot’s language understanding system involves supervised learning, that is, it improves its performance based on evaluating the outcome of past interactions.
Supervised Machine Learning
The NLU tool used is based on Machine Learning: based on the instances provided in the training set, the algorithm evaluates the degree of similarity between the user’s request and each of the user’s intent. This similarity is expressed by a Confidence Score. If this exceeds a certain minimum threshold, the NLU engine will output the intent with the highest score. If this threshold is not exceeded, there is a case of no match. Supervised Machine Learning refers to a machine learning approach based on a training data set provided by a supervisor. This is a non-deterministic system: it is not possible to describe or set up the system on the basis of rules such as the model “if the input contains word x associate the query with intent y”. In the setup phase, the bot is trained with a training set built according to the agreed knowledge base and tests performed. After the go-live, the maintenance service is activated. Specifically, a periodic audit is done on a significant sample of conversations to optimize the NLU component by expanding the training set.