I am trying to develop a custom sentiment analysis model on Google AutoML. I have achieved the same, however, I have the following query:

Scenario1: I have excluded records from my training data where Classic Language API is already giving accurate sentiment labels and I have trained the AutoML model only on fallouts.

F1-Score in this scenario pretty less when tested on benchmarking data

Secnario2: I have trained the AutoML model on the whole of training data.

F1-Score in this scenario has shown significant improvement than Classic Language API.

Hence,wanted to check with you all on what could have wrong on AutoML when I trained the model only on the fallout from Classic Language API


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.