This collection contains several basic courses that most AI Trainers should take as part of their onboarding. The courses in this collection will teach you how to build a virtual agent, how to train it, how to improve it and how to write content. It might seem like a lot, but for the best learning experience, we recommend you not take everything at once. Instead, take what you need when you need it. The intention is that you should take this while working on a virtual agent for the first time, taking the material you need when it's relevant and practicing what you've learned immediately.
Onboarding for AI Trainers
Introduction
Understand how this onboarding is structured and how to work with the content.
Building and training a virtual agent
Everything you need to build a virtual agent and train a custom NLU model. Even if you are not interested in building a model right away, we recommend the course on the intent hierarchy on its own as an introduction to creating content as well.
-
Building an intent hierarchy
The intent hierarchy is the core of the boost platform. It is here that the scope, what your virtual agent is supposed to know, is organized. This course covers how to build and navigate your own intent hierarchy.
-
Train a custom NLU model
For the virtual agent to understand end users it needs some kind of model, and the boost platform allows you to train your own custom NLU model. This course takes you through that process, including training data and test data.
-
Introduction to language processing
The language processing algorithm is a central part of how our custom NLU model understands languages. In this course, we will see how big of an impact this algorithm can have on model performance and how it works.
Improving a virtual agent
These courses are about improving the model performance and prediction accuracy. If you want to go live with a very secure and high performing NLU model, then this is a must. These course will benefit from taking it slow, following the practical assignments and learning by doing.
-
Working with clean-up reports
The clean-up reports are built-in reports that tell you where to improve your trained NLU model. In this course, we will go through some of the more important reports and give you an idea of where to start.
-
Improving your model with test results
Test results are the ultimate tool for benchmarking and improving a custom-trained NLU model. This course introduces the various kinds of issues we can find here and how to resolve them, improving prediction accuracy.
Creating content
Content, responses and conversation design is foundational to virtual agents, and you need to know the basic tools. These courses covers the basics for both pre-defined responses and generated responses.
-
Create and design basic content
A virtual agent needs to be able to respond to an end user with well-designed and informative content. This course takes you through the most basic and simple tools for creating great responses.
-
Building flexible responses with generative action
Generative action allows us to create dynamic and flexible responses that cover multiple scenarios and use cases. This course gives an introduction to how to write instructions, upload knowledge, design action and API hooks, and set up guardrails, maki...
Analytics and conversation review
This course covers conversation review and it is the gateway into analytics. This is how you turn your VA's conversation logs into actionable insights, improving you VA as you go.
