I was wondering if it would be possible to use this library as an explanation, but in the context of music. Let me explain:
If I use a tokeniser from the Miditok library:
https://github.com/Natooz/MidiTok/blob/main/miditok/midi_tokenizer.py
and a trained model like BertForSequenceClassification in that domain, could I use SequenceClassificationExplainer to give me an attribution for each of the tokens?
I understand that this is not prepared for this context and will give "token attributions". Am I right?
Is this feasible? Could you give me some advice on how to do this?
Congratulations on the work!
I was wondering if it would be possible to use this library as an explanation, but in the context of music. Let me explain:
If I use a tokeniser from the Miditok library: https://github.com/Natooz/MidiTok/blob/main/miditok/midi_tokenizer.py and a trained model like BertForSequenceClassification in that domain, could I use SequenceClassificationExplainer to give me an attribution for each of the tokens?
I understand that this is not prepared for this context and will give "token attributions". Am I right?
Is this feasible? Could you give me some advice on how to do this?
Thank you very much!