AI Chatbot App Development

August 07, 2019

With the rise of messaging apps, businesses increasing need to look at ways of communicating with their customers on the tools they use, rather than change behaviour. Chatbots using machine learning are becoming a very popular method of communicating as they use artificial intelligence to understand language and not just respond to commands. As a chatbot development company, we share some of our insights.

Chatbots use Natural Language Processing (NLP), which is a type of AI that assist computers understand human language. As there are many nuances to a human language, machine learning assists with interpreting these nuances based on the analysis of millions of conversations. For example, there are different ways of asking for your account balance:

  • What is my account balance?
  • Can you please provide the balance of my account?
  • How much money have I left in my account?

When setting up a chatbot we create an agent that can respond to customer queries. Each type of query is called an Intent and we create several training phrases for a base for the machine learning system to interpret the variety of other possible phrases a customer might use to determine their intent. This means that a customer can use their own language to do the request.

Once the intent has been created we identify the various entities that may be required for the customer query to be handled. For example, if a customer wants to check their account balance, an account number would be required. There are 3 types of entities:

  • System entities – these are automatically identified by the Chatbot based on their format. For example dates, times, numbers, addresses, and email addresses.
  • Developer entities – these are entities that are required or assist with the customer query. For example if a customer would like their account balance, they are required to provide their account number.
  • User entities – these are defined per user, and help the chatbot assist a customer with a more personalised response. For example if a customer previously asked for an account balance the chatbot could ask if they wish to view the balance again.

Once the intent and all the required entities have been received, the transaction can take place. The chatbot will continue asking questions until all the required entities have been filled.

Backend code will then query a web-service (Such as obtaining account balance via an account number through an API) and return the result to the chatbot.

The chatbot will then return the result to the customer in friendly human language for example: Thomas, your account balance is R500.00. Would you like help with anything else?

Refresh develops the chatbot independent of any specific communication channel, so it can be deployed across multiple channels (Eg. WhatsApp, Facebook Messenger, SMS, Voice call) without the need to develop a chatbot separately for each.

Drop us a message if you wish to learn more about chatbot development, or would like a proposal for your company chatbot. You can also view more about ai chatbot development.

View the chatbot we created for Pick n Pay here.


Leave Your Response

* Name, Email, Comment are Required