Improving Dialogflow Accuracy

Current Status

Not Enrolled

Price

Free

Get Started

Please note: when you enroll in a free course, after you log in, the login button text will change to "Take this course". 

You must ALSO click on the "Take this course" button above to get access to all the lessons inside the course. [Screenshots]

This course explains a systematic process for improving the accuracy of Dialogflow chatbots.

Course Content

Download ES conversation logs

Download history JSON using Google Chrome

Convert history JSON to CSV for easier analytics Sample Lesson

Step by step guide to improving Dialogflow ES accuracy

Export your agent for versioning Sample Lesson

Convert History JSON to CSV Sample Lesson

Annotate user utterances Sample Lesson

Calculate the accuracy Sample Lesson

Update your bot Sample Lesson

How do you measure your Dialogflow bot's accuracy

Chatbase UMM Sample Lesson

The confusion matrix

Sample size

Chatbase Alternative

Maintain versions easily Sample Lesson

Group by sessionID to visualize conversations Sample Lesson

Group by intent names to see the most popular intents Sample Lesson

Filter by fallback intents to design new intents Sample Lesson

Calculate bot accuracy Sample Lesson

Create auto-updating summary views for your team Sample Lesson

Dialogflow Conversation Analytics Tips

You cannot do analytics if you cannot calculate accuracy Sample Lesson

Don’t use slot filling (required parameters) Sample Lesson

Minimize candidate intents

Follow good conversation design principles

Use Versioning

Design a Flowchart

Use Airtable for calculating accuracy

Use a conversation-friendly integration channel if possible

Don’t use the Training feature (blindly)

Set up automated tests

Follow a systematic process

Dialogflow ES Unit Testing Framework

Introduction Sample Lesson

Verbatim vs non-verbatim test phrases Sample Lesson

Generating test scripts

How to automatically generate the test script

Running Test Scripts