Artificial Intelligence By Example

Chatbots

Most NLP algorithms use probabilities to predict the next words. Or they use preset sentences to answer questions without analyzing the real meaning. 

This means building systems that guess, sometimes effectively, sometimes with errors. In any case, they do not think. Many applications produce good results with statistical approaches. But for some projects, customizing a chatbot comes in handy.

For example, imagine that somebody is speaking to you in a language you do not know at all. You have the inbuilt ability of present NLP algorithms to predict representations through statistics on your smartphone.

Although you do not understand a single word, you produce words, statistically, that could fit the situation. For isolated words ...