Does the presence of slots, in any way affect the NER?

Rasa Open Source
1

The scenario I’m looking at is this.

Let’s say I’ve configured the bot to identify intents x, y and z.

A certain utterance M with a specific slot value-set leads to the bot identifying M with intent x and an entity list A.

Is it possible for -

  1. the same utterance M with an absent slot value-set leading the bot to identify it with a different intent and a different list of entities?
  2. the same utterance M with a different slot value-set overall leading the bot to identify it with a different intent and a different list of entities?

My assumption is that it does not affect the recognition of intents/entities at all (which is what I would like), but I’m trying to validate this assumption for the current version of Rasa. If this is the case, are there any reasons to not do it this way in the future?

Many thanks in advance! I’m absolutely loving the tool.

2

Slots are part of core package and to the best of my knowledge have no influence when it comes to the NLU part. Entities, regex and the likes do.

3

Thanks for replying @Mappi. Could you elaborate what you mean by -

Entities, regex and the likes do.

4

Quote from the wiki sums it up quite nicely:

The common theme here is that common examples, regex features and lookup tables merely act as cues to the final NLU model by providing additional features to the machine learning algorithm during training. Therefore, it must not be assumed that having a single example would be enough for the model to robustly identify intents and/or entities across all variants of that example.

5

Thanks @Mappi; although I’m not sure if your last post answers my question. The question isn’t about a single example, but whether the slots that exist in the conversation at the time have any impact on the intent/entity recognition of the NLU.

@Tanja would you be able to confirm?

6

@ganeshv Slots do not impact the intent/entity recognition of NLU.

1 Like
7

Thank you!