“OK, Google, book a hotel room this Saturday for two days in a three-star hotel in Sydney.” Will you get an answer? No, your request will be implemented and a room will be booked for you. How does the system actually works?
There are two types of bots. One is master bot like the Google Assistant, Amazon Alexa and Samsung’s Viv. The master bot accepts your command and figures out subordinate bot that can handle the request. In this case, it will find a travel agent bot like Booking.com, Alterra, and Expedia. The request will be forwarded and the sale will be made if the subordinate bot is intelligent enough to understand the language.
Similarly, if you are chatting with the Lonely Planet bot about Colorado attractions, you can ask like “What will be the weather in Colorado this Saturday.” The master bot of Lonely Planet will pass the request to bots like Weather.com and if it is capable of understanding it will work.
The standardization of bot is English now. The language has become inter-bot communication protocol.
However, there are complaints several bots feel dumb too many times and we wonder how it happens. There are two broad types of Natural Language Processing (NLP) algorithms. These are categorized as traditional and deep learning.
The first NLP algo requires a lot of coding and requires anticipating of words and phrases that one could say. A lot of rules work in it. All these takes lot of time to write and debug the programs, but the bot becomes dumb if user deviates from the expected path.
The deep learning type gives a hope. Constructing of an artificial neural network is required in it instead of writing explicit rules. It is similar to the child and understands the language.
This is the reason chatbots need deep learning.
What do you say?