Unable to capture entity value in intent

I have created an entity of custom data type where I listed all the accepted value. Then I create an intent which accept the entity in utterance.
When I try to type some sample sentence, VF can successfully detect the intent but unable to capture the complete entity value.
My expected result is it can capture the correct value instead of cropping the word or sometimes capture other value which is not listed in entity value.

Yyeanpei I have created an entity of custom data type where I listed all the accepted val...

Tico | Voiceflow Assistant•4/23/24, 9:26 AM

Based on your description, it seems like you're encountering an issue where Voiceflow is not capturing the complete entity value in your intent. This could be due to a few reasons:

Entity Recognition: Ensure that the entity you've created is correctly defined with all the possible values you expect to capture. If the entity is not recognizing the complete value, it might be because the value is not listed in the entity's possible values.
Utterance Definition: Check the utterances you've defined in your intent. The utterances should include examples that contain the entity value in various contexts. This helps the system understand how to extract the entity from the user's input.
Entity Type: Make sure the data type of your entity matches the type of data you're trying to capture. For example, if you're trying to capture a phrase, the entity should be of a type that can capture phrases.
Entity Overlap: If you have multiple entities that could potentially match a part of the user's input, the system might get confused and capture the wrong value. Make sure your entities are distinct and specific.
Entity Synonyms: If your entity has synonyms, make sure they are correctly defined. If a synonym matches part of the user's input, the system might capture the synonym instead of the complete value.

Remember, entities are used to extract specific pieces of information from the user's utterance. If the user's input doesn't exactly match a value or synonym in the entity, the system might not be able to capture the complete value. Regularly updating your entities with new phrases and terms that users naturally use can help improve entity recognition.

If you've checked all of these points and are still having issues, it might be a more complex problem that requires a deeper look into your project. In such cases, it would be beneficial to review the specific utterances and entity values that are causing issues.

few-sapphireOP•4/23/24, 9:30 AM

I have changed the data type for entity to Key Phrases and still not working, it is still cropping the word

few-sapphireOP•4/25/24, 2:00 AM

Can anyone help with this?

colossal-harlequin•4/25/24, 6:56 AM

I think you want to extract AB right? I guess the problem is in your examples. Entity recognition in VF works currently under the hood with NLU. In my experience it needs therefore „many“ example special when you cases is not very common. Maybe you can provide VF more examples?

Ccolossal-harlequin I think you want to extract AB right? I guess the problem is in your examples. E...

few-sapphireOP•4/25/24, 8:31 AM

I have doubled it to 50+ examples but still not working well. Do you know what is the minimum number of examples we need to provide for the intent?

colossal-harlequin•4/25/24, 2:32 PM

NLU is not as advanced as LLM. Therefore, the number of examples needed depends on the specific use case. It might be too complex to provide further advice remotely. You could also consider creating a prompt to extract your subjects with ChatGPT or Claude model.