Double entity extraction using DIETClassifier & RegexEntityExtractor

Rasa Open Source
1

Hello, I am facing a problem where I need to extract names using a lookup table. The table is very large ~2k entry.

Using DIETClassifier to extract the names is not working very good for examples not included in the training examples. RegexEntityExtractor works well, however, sometimes entities gets extracted twice by both extractors. I can’t set DIETClassifier entity extraction to False because I use it in extracting other entities so I need both to be running. I just need to define which extractor is used for which entity.

We are using custom actions, I understand that I can do it inside the validation function but the hassle it too much because I will need to do in a lot of entities. Any simpler solutions?

1 Like
2

you can create custom nlu component that will filter extracted entities

1 Like
3

any available examples for this component?

4

not that I know of. But you basically need to implement the same logic, that you wanted to do in your custom action

1 Like
5

Hi, I am facing the same problem were you able to fix it?