Chatbots are usually implemented in two ways:
- Click Based Approach
In this approach, people do not have freedom to type in their questions. The user inputs are restricted to the options provided to them as clickable buttons. These buttons, on click offer further options. The advantage of this approach is that, it is less error prone with the chatbot controlling the flow. The disadvantage of this approach is that it may take multiple clicks/ steps to fetch answers.
Several banking websites have such chatbots.
- Free Format Approach
In this approach, the user is free to type and ask any question. The advantage of this approach is that the user gets a much quicker response. The disadvantage is the umpteen number of possibilities that need to be supported while implementing the chatbot.
Siri, Google Assistant are examples of such chatbots.
The ideal way to implement such chatbots is via Machine learning. And while that is the preferred approach, it is also highly resource(CPU/ Memory/ HD) intensive. For a solution with a limited data subset, there is no need to support every possibility under the sun. It is not only expensive to build such models, but at times irrelevant as well. For instance, the City Chatbot doesn’t need to answer random queries like ‘Explain Big Bang theory’ nor ‘Who will win the World Cup’.
The Chatbot we are building from scratch, uses a hybrid approach between both the models above, utilizing our City database schema as guidance.
Before we roll out the release, here is a sneak peak into ‘Click based approach’ of our City Chatbot:
We hope to release our chatbot within a month after testing and including the ‘Free Format Approach’ as well.
And we will use the analytics based on user interaction to continue feeding the chatbot with further improvements.