Hi, i’m trying to to face this task (multi-label + sentiment of the labels) but i’m not sure about the best starting model for this purpose, Is there some model or model pipeline already defined for this kind of task?
One option that i have been exploring is to use a multi-label sequence bert based classifier for the categories and then use some other models to determine the sentiment of each label within the text that generated the categories. Maybe apply some ABSA bert over the most frequent keywords of the category within the input text?
I would like to know more about existing models or resources that can be useful for this task.