Training a model in Re:infer can be broken down into a three stage process:

 

In the first instance, it’s good practice to go through these steps in order, but this can be an iterative process. You may find in time that for different labels, you may chop and change the different steps as you become more familiar with the platform.


Discover

Discover is where similar intents, patterns and conversation themes are grouped together into ‘clusters’. This is the starting point and is used to quickly build an initial model where you analyse your data and tag each cluster with one or more label that applies.  


Explore

After reviewing clusters in Discover, Explore is used to further train your model. Most of your time will be spent here reviewing verbatims, adding labels, and improving the model’s understanding of your data. 


Refine

Refine is used to further tweak your model, find and correct potential confusions or inconsistencies, and improve the overall performance. You can also use this stage to check you have captured all the relevant concepts in your dataset.


Prune / Re-organise

This is a part of the model training process that you can do at any time - renaming, merging or deleting labels as you go through the process. The process is explained in detail here in the Explore section.


Previous: Objectives & building a Taxonomy     |     Next: Training tips and common pitfalls