Training a conversation

Owned by Danil Chernovalov (Unlicensed)
Last updated: Mar 17, 2021
2 min read

This page provides a quick guide on how to review and train a conversation.

Prerequisites

To efficiently review and train conversations, you must be familiar with the features of the tool. To learn about them, see the Features page.

Key points

The process of training consists of 3 main stages:

  1. Reviewing user chat sessions.

  2. Analyzing conversations and associated insights.

  3. Rating and training the insights.

Reviewing

To review a conversations and study its details:

  1. Go to Chatbot > Conversation training.

  2. Expand the pane of the conversation of your interest.

  3. Switch between the Insights, Conversation, and History tabs to study the details.

Analyzing

In the the Insights tab, see the options that Kbot uses as answers to the user’s statement. Some of them are precise and useful, some are less accurate, and some are added by default (such as “Talk to an operator”, for example).

Study the insights to identify accurate ones and decrease the rating of inappropriate ones. As the result, Kbot does not suggest the latter to users.

Rating

Once you have studied a given conversation and its insights, you can mark them as precise and helpful.

  1. To manage insights, open the Insights tab.

  2. Mark the best answer with the Star icon. This suggestion is going to be presented to users next time the similar request takes place.

  3. Train other options that might be useful with the Hat icon. Bot learns these suggestions and presents them to users as additional options next time the similar request takes place.

  4. Decrease the rating of inaccurate insights. If you think that an insight is neither accurate nor helpful, click the Cross icon.