Dialogflow

Dialog flow is a service offered by Google to automate customer interactions using predefined scenarios, machine learning and AI.

The technology can be broken down to three main segments on how this works.

1. Intent Tracking.

With this Dialogflow will try to identify the intent of request. e.g. if user is trying to book an appointment or he/she is looking for a information. Based on the intent, Dialogflow will vary the response required. You can consider this as predefined scenarios or context

2. Entity Extraction

From the information given by the customer, Dialogflow will match entities to get the idea about the details embedded in the conversation. e.g. if the customer says "I would like to book an appointment today", Dialogflow can related today as an entity of system date. There could be two types of entity defined.

  • System defined entity - Entities not specific to a business scenario and mainly applicable for any customer / business.
  • Developer defined entity - Specific to a business.

3. Dialog Control

Dialog control is the core functionality of Dialogflow. Using Dialog control, Dialogflow can be truly conversational - with customer asking questions and Dialogflow providing contextual answers.
Conversations in real life can be simple or complex.  There are two broad types of conversation

  • Linear conversation 
One information is passed after another. Once all information are passed, following action/response can be provided to customer. In this type of control, information can be marked as required. Until customer provides all the information, Dialogflow will ask follow-up question to get the relevant information.           
  • Non-Linear Conversation
Non-linear conversations are not based on just on information but also based on context. In this type of control input and output context can be defined and linked. Based on the response from customer, it will be linked to right context to extract meaning of the response and follow-up actions.

More details of Dialogflow can be found in  https://dialogflow.com/

Comments