2016 ends under the sign of Chatbots . Like major brands such as PizzaHut, Lufthansa or Bank of America who launched their conversational robots this year, the phenomenon seems to be growing and the prospects are numerous. We offer you a brief overview of a Chatbot that we set up for the 2017 presidential elections.
Why a Chatbot?
Out of curiosity first. Some even predict that they will replace mobile applications, so it was worth making up our own minds.
Understanding human language by a machine is based on the interpretation of the intention and associated parameters (entities). NLU (Natural Language Understanding) has made enormous progress in recent years (particularly with the democratization of Machine Learning) and makes possible highly ergonomic experiences.
Regarding our presidential election chatbot, which allows costa rica mobile database you to vote anonymously (in our system, not in real life!), expressions like "Go ahead Fifi" or "Flat behind you, Alain" are correctly interpreted: {Intention: Vote, Candidate: François Fillon} (for example for the first one).
The basic bricks
The GAFA have not been mistaken, and have invested in this sector. Facebook with the acquisition of Wit.ai and Google with that of Api.ai intend to be suppliers of basic components for the development of a Chatbot. IBM's Watson or Microsoft's Luis are, with their specificities, in this same movement.
Developing a Chatbot starts with listing what it will be able to do. In our example, we decided that these areas of expertise would be:
Give your opinion for the candidates
View the survey results
Indicate the polling stations
To make the experience interesting, we determined the sequence of actions associated with each area of expertise, which the user can trigger using our suggestions or in natural language.