I hope you are all well. You see I was curious how to fine-tune an llm for fine-grained sentiment analysis. For those of you who dont know, fine-grained sentiment analysis means having multiple output sentiment classifications(not just positive,negative,neutral). E.g It can be angry,happy,sad,depressed,neutral,surprised. Or it can be negative,little negative, neutral, little positive, positive.
I tried coding different stuff, I thought of including fine-grained audio sentiment analysis in my multimodal sentiment analysis fyp.