PLEASE NOTE: UiPath Communications Mining's Knowledge Base has been fully migrated to UiPath Docs. Please navigate to equivalent articles in UiPath Docs (here) for up to date guidance, as this site will no longer be updated and maintained.

Knowledge Base

Model Training & Maintenance

Guides on how to create, improve and maintain Models in Communications Mining, using platform features such as Discover, Explore and Validation

Using 'Train'

User permissions required: ‘View Sources’ AND ‘Review and label’


TABLE OF CONTENTS

 


Overview


The new 'Train' feature provides a fully guided label training experience for users. 


The main page of Train provides useful information regarding the training done so far, the performance of the model, and a list of prioritised next best training actions to take (similar to the Validation page).



 

Train home page for a partially trained dataset with no entities enabled

 

  

When users select a training action to take, they are taken to a specific 'training batch' interface (see below), which breaks up the training into short, easy-to-consume sessions. 

 

Depending on the recommended action, the number of verbatims or clusters of verbatims in the batch will vary, but it's typically 10.




Batch training page for 'Shuffle' training

 

Once you've applied the labels (and entities) to the verbatim(s) on screen, you can click done, or move onto the next verbatim or cluster using the 'Next' button.


At the end of the batch, you're provided with a summary of the training actions taken (see below), and can then choose your next session by selecting another recommended action. 




Summary of training actions completed during a training batch

 

Please Note: The training that you complete will trigger a re-training, and the next best actions will also update as soon as possible, however, the same actions may still appear on screen if more sessions of them are required. It's perfectly normal to do multiple sessions of the same recommended action, and you don't have to wait for new actions to appear to jump into another similar session.

 

If users prefer to train without the platform's guidance, they can disable this and select themselves which sessions to complete. For more detail, see the section below.

 


How does it impact the model training process?


Train will in future become the go-to place to complete all of your model training from start to finish, but some additional features are still in development (e.g. guided entity training). At present, it's entirely additive to the existing feature set, meaning all of the functionality you're used to can be used as-is, and you can train models the way that you're used to.


We do encourage our users to use Train for a guided label training experience, and provide feedback to your UiPath contacts if encountering any issues or challenges.


Here's an overview of how you can use it as part of model training going forward:

 

Label training


If experiencing any challenges, please remember that all of the pre-existing training modes are still available as they were via the Discover and Explore pages.


Here's some key things to understand about training labels in Train:


  • It guides you right from the moment you create a dataset with the next best actions to take to advance your label training - this includes uploading a taxonomy before you begin training (see below)
  • Train will essentially guide users through the usual steps covered elsewhere in this Knowledge Base for the model training process (see the Overview here), with the exception of recommending search
    • If needing to use 'search' to seed some examples (sparingly) for particular labels, this can be done through the usual means in Discover or Explore, or by temporarily disabling the guidance in Train (see below for detail)
  • Train provides 'need to know' performance feedback in the main page and through its recommendations, but if you need detailed feedback on model performance you should still navigate to the Validation page

 

 


 

Train page for a completely new dataset

 

Entity training


Additionally, if you have entities enabled on your dataset, you can toggle between training labels and entities on the train tab. 


Toggle to switch between training labels and entities on the Train tab



Similar to training labels, if you're experiencing any challenges, please remember that all of the pre-existing training modes for entities are still available as they were via the Explore page.


Here's some key things to understand about training entities in Train:


  • It guides you right from the moment you create a dataset with the next best actions to take to advance your entity training. 
  • Train will essentially guide users through the usual steps covered elsewhere in this Knowledge Base for training entities during the model training process. 
  • Train provides 'need to know' performance feedback in the main page and through its recommendations, but if you need detailed feedback on entity performance you should still navigate to the Validation page and go to the entity validation page 
  • Please note that during the beginning of the model training process, if the platform doesn't have enough examples of entities to learn from - it will recommend shuffle by default. Once enough examples have been provided, it will recommend more targeted training for specific entities. 


Train page for entities on a trained dataset



Using Train without guidance enabled for labels


The default setting for the Train page is to have platform guidance enabled for labels, as this is our recommendation. 

 

If you're a confident model trainer, however, and you know the actions that you want to take already, you can disable the guidance using the toggle in the top right-hand of the page (see below).

 

 Toggle for enabling/disabling platform guidance

 

If guidance is disabled on the , the page will look like this:



Unguided Train interface


The platform will still highlight the phase of training it thinks is most appropriate.  Within each phase the usual training actions can be found, and you'll be able to target specific labels as needed (see below).


Teach label selector using unguided Train interface

 

This is also where you can still train in sessions using 'search'.


You can easily switch between guided and unguided training as needed.

 


Further updates


We will be adding additional functionality and improving existing functionality for the Train page frequently. As we continue to update Train, we'll also update our guidance on the best process to follow whilst training, whatever your level of experience.


Did you find it helpful? Yes No

Send feedback
Sorry we couldn't be helpful. Help us improve this article with your feedback.