Preparing data for the Chatbot

The article elaborates on the process of data preparation for training the NLU model (natural language understanding functionality) of the Agent.

It is easier and faster to develop a dialogue design when there is a message history between the company and future Bot Users. For example, when creating an Agent for providing technical support, recordings of telephone conversations with technical support operators can be useful. Voice recordings are transcribed and clustered with the help of the intent mining tool. As a result, the messages will be grouped into topics, and for each topic, there will be examples of clients’ messages and Human Agents’ answers. Examples of clients’ messages will be included in the training set for creating intents, and the Human Agents’ answers will be used in the Agent’s answers in the Dialogue Script. If you are interested in intent mining, feel free to email us: at For more information, see Query for Dialog Clustering.

In case there is no such data, to create content for the Bot you can:

  • Describe the business process in detail and compile the data set yourself;
  • Find publicly available chatbots designed to automate similar business processes and adapt the dialogue script so that you could use it for your case;
  • Contact GMS experts for the consultation. We can develop a simple test Agent right away if you provide some information – see the Request for a Test Bot article for details.