Remove patterns and labels from EntityRuler

[dhruvrnaik]

I cannot find a good way to remove patterns from a label, or the entire list of patterns corresponding to a label with the EntityRuler pipe.

There is one way by using entity_ruler.matcher.remove(<LABEL>), but this is not a good solution as it doesn't remove labels from entity_ruler.token_patterns or phrase_patterns.

It would also be great if there was a way to remove a single pattern from a label, and not have to delete the entire list of patterns and the label.

[polm]

It's true we don't have a good way to do this. I think the reason is that in our experience it's typical to just build the EntityRuler and use it without modifying it over time.

I don't think it'd be complicated to add this, but can you clarify how you're using the Entity Ruler that makes this a feature you need? It'll help us understand how to best adjust the API.

[dhruvrnaik]

I would want to do entity/phrase detection for different users on a chat platform.

Users would have the flexibility to add and remove phrases they want to detect in real-time chat at any given time. While you can keep adding patterns, and ignore the results for patterns you don't need anymore, I'd prefer having the ability to add/remove patterns based on user needs.

[polm]

Thank you, that's a very easy to understand use case and I can see why you'd want to remove patterns in that situation.

No promises on when it'll get done but we'll consider adding the feature to remove patterns.

As a workaround, I don't think you can do better than directly manipulating the internals or rebuilding the EntityRuler from scratch when things are removed.