Support Centre

Model Training & Maintenance

Guides on how to create, improve and maintain Models in Re:infer, using platform features such as Discover, Explore and Validation

Entity filtering

User permissions required: 'View Sources' AND 'View entities'

 

Just as you can for labels, you can filter verbatims by whether they have entities predicted or assigned, both in Explore and Reports.


You can apply any combination of 'AND', 'OR' and 'NOT' when applying more than one entity filter. These filters can give you much greater flexibility when training and interpreting your data, and can provide much deeper insights on what's happening in your communication channels.


Here's some of the things you can now do when filtering by entity predictions:

  • Apply multiple entity filters at once, in both Explore and Reports
  • Filter to verbatims that have one of a number of selected entity predicted (i.e., Entity X OR Entity Y OR ...)
  • Filter to verbatims that have multiple different entities predicted (i.e. entity X AND entity Y AND ...)
  • Filter to verbatims that do not have certain entities predicted (i.e. NOT Entity Y)
  • Search for entities containing specific search terms, whilst having entity filters applied


All of the entities you have enabled on your dataset will appear as shown below in the filter bar. Assigning entities is covered in detail in the next article in this section here.



Applying advanced prediction filters

 

There are now two ways to apply entity filters, and they can be used in combination with each other to create the right type of query.



 

Entity filter bar

  

The default state is shown above, whereby no filter is applied and all verbatims will be shown (unless another filter is applied).


To update the entity filter, use the buttons explained below. They change colour when selected:

 

Select verbatims that have assigned entities
Select verbatims that have entities predicted

 

If you wish to filter to verbatims that have any entities assigned or predicted, use the buttons at the top (as shown above). If you want to filter to verbatims with specific entities assigned or predicted, hover over the entity in question and the same two buttons will appear to the right.


If you want to filter to an entity either assigned or predicted, just click the name of the entity, and it will show verbatims with either.


To remove your selection, simply click the button again, or to remove multiple selections, click 'All'. You can also click 'Clear All' at the top of the filter bar, but this will clear every filter you have selected, not just entity filters.



Entity Bar

 
The taxonomy of entities functions as a normal filter bar, and allows you to select multiple entities at once with a single click for each. 


Selecting multiple entities from the list creates an 'OR' type query. 

 

If you selected Entity A, Entity B and Entity C in the entity bar, this creates a 'Show me verbatims with Entity A, Entity B, or Entity C predicted' query.


When filtering to specific entities, you can make multiple selections. For instance, you could filter to see verbatims that have an address line entity assigned OR a city entity predicted (as shown below).

 

Entity filter bar with assigned address line or predicted city entities selected

  

Add Entity Filter 


The second filter option is the '+ Add Entity Filter' button above the entity bar.


This enables a dropdown entity bar that allows you to select more complex filters, including excluding certain entities from consideration.


 

From this dropdown, you can select multiple entities to include or exclude by clicking the name of the entity (for assigned and predicted), or the individual buttons (including minus for where this entity is neither assigned nor predicted). 


The result would be something like the example below, which would return verbatims predicted to have the entity 'Address Line', but that also do not have the entity 'City' assigned or predicted:

 

 

Example entity filter combination

 

You can click '+ Add Entity Filter', multiple times to add additional layers to your query. Two separate entity filters create an 'AND' type query, whilst multiple entities selected in the same entity filter create an 'OR' type query.


In the example below, multiple entity filters have been applied individually. This creates a filter that will return verbatims predicted to have any of the four entities in the first filter, but that also have the 'Value (£)' entity predicted, and do not have the 'Postcode' entity predicted or assigned.

 

 

An example of complex entity query combining 'OR', 'AND' and 'NOT' entity filters


A helpful tip is that by clicking the '&' sign in an individual filter containing multiple entities, you can automatically split them out into individual filters. This would change the query from 'OR' (i.e. any of these entities predicted) to 'AND' (i.e. all of these entities predicted).

 

 

Combining entity bar filters and added entity filters

 

It's possible to combine filters from both the entity bar, and individually added entity filters. Filters applied in the entity bar are treated as an 'AND' query with any individually applied entity filters.


For example, in the image below, this combined query would return any verbatims that had the 'Address Line' entity predicted AND either of 'City' OR 'Policy Number' predicted.

 


Combined entity filter using entity bar and individually added entity filters

 

 

Combining entity filters and sorting by entity for training

 

What these new filters also mean, is that you can now apply entity filters and sort by a specific entity for a training mode.

 

In example below, you can see verbatims where 'Postcode' may have been applied incorrectly, but that also do not have the entity 'City' predicted:



Explore page showing 'Check entity' mode for a specific entity, with an additional entity exclusion filter applied

 


Previous: Enabling, disabling, updating and creating entities     |     Next: Reviewing and applying entities

Did you find it helpful? Yes No

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

Sections

View all