Exam AI-102 topic 3 question 20 discussion

I would say is D) as per the following: https://docs.microsoft.com/en-us/azure/cognitive-services/language-service/conversational-language-understanding/concepts/none-intent#adding-examples-to-the-none-intent

False positive means => The model needs examples of what it should not classify as "GetContactDetails," which is the role of the "None" intent. Therefore, the most effective approach is to add a diverse range of examples to the "None" intent, covering various phrases and queries that are outside the scope of "GetContactDetails." This helps create a clear boundary for the model, reducing the likelihood of it mistakenly classifying unrelated inputs as belonging to the "GetContactDetails" intent.

I say this answer is D.

D is answer.

For me and ChatGPT: d

D. Add examples to the None intent.

It must be D.

The correct option to decrease the likelihood of a false positive in the Language Understanding model is to add additional None intent examples. Option D is correct. By adding more varied examples that do not map to a valid intent to the None intent, the model can better learn when an utterance does not apply and avoid falsely matching invalid queries to a valid intent. Options A, B, and C may improve the model in certain ways, but they do not directly address reducing false positives. Only adding additional out-of-scope examples to the None intent will help the model better distinguish when new utterances do not match any existing intent's patterns. So out of the options, adding examples to the None intent is the way to decrease the likelihood of false positives.

to me the answer is D. Non intents have the purpose to reduce false positive too. https://docs.microsoft.com/en-us/azure/cognitive-services/language-service/conversational-language-understanding/concepts/none-intent#adding-examples-to-the-none-intent

200 sample data. --> much false positive. --> Increase number of training data. --> Add example to the None intent, not active learning in this context.

D is the answer. https://learn.microsoft.com/en-us/azure/cognitive-services/language-service/conversational-language-understanding/concepts/none-intent#adding-examples-to-the-none-intent The None intent is also treated like any other intent in your project. If there are utterances that you want predicted as None, consider adding similar examples to them in your training data. For example, if you would like to categorize utterances that are not important to your project as None, such as greetings, yes and no answers, responses to questions such as providing a number, then add those utterances to your intent. You should also consider adding false positive examples to the None intent. For example, in a flight booking project it is likely that the utterance "I want to buy a book" could be confused with a Book Flight intent. Adding "I want to buy a book" or "I love reading books" as None training utterances helps alter the predictions of those types of utterances towards the None intent instead of Book Flight.

A. Enable active learning. By enabling active learning, the model can actively request feedback from users when it encounters uncertain or ambiguous queries. This feedback loop helps improve the model's understanding and reduces false positives by incorporating user input into its training process. Option A (Enable active learning) is the correct choice to decrease the likelihood of false positives.

You should also consider adding false positive examples to the None intent. For example, in a flight booking project it is likely that the utterance "I want to buy a book" could be confused with a Book Flight intent. Adding "I want to buy a book" or "I love reading books" as None training utterances helps alter the predictions of those types of utterances towards the None intent instead of Book Flight. https://learn.microsoft.com/en-us/azure/cognitive-services/language-service/conversational-language-understanding/concepts/none-intent#adding-examples-to-the-none-intent

You should also consider adding false positive examples to the None intent. For example, in a flight booking project it is likely that the utterance "I want to buy a book" could be confused with a Book Flight intent. Adding "I want to buy a book" or "I love reading books" as None training utterances helps alter the predictions of those types of utterances towards the None intent instead of Book Flight. https://learn.microsoft.com/en-us/azure/cognitive-services/language-service/conversational-language-understanding/concepts/none-intent#adding-examples-to-the-none-intent

To decrease the likelihood of a false positive, you can add additional examples to the GetContactDetails intent. This will help the model to better understand the intent and reduce the likelihood of false positive predictions.

As explained in MS Document "false positive" = None intent

Add examples to "None" intent