How to use this dataset with financial sentiment analyse models
Hello,
Thanks a lot for this super dataset. I'm trying to a financial news sentiment analyse based on this dataset. However, with the models such as FinBERT, it can give positive, negative or neutral label with scores. But it's difficult to know the "entity" of the answer.
For exemple, the headline: "ING maintains a bearish outlook on the Canadian dollar".
Obviously, it's negative for Canadian dollar (CAD), but there is the wold "dollar" inside. The response will be confusing, it's positive for CAD or USD ?
There are some ABSA (aspect based sentiment analyse) models, but we still need to parse the headline and pass the good entity to it. So the tough task here is how to extract the good "aspect" or "entity".
I also tried some financial-specific NER (named entity recognition) models to get the good entity then pass to the ABSA model. However, the NER models also split the sentence to several parts and each has a score. It's difficult to get the good entity to pass to ABSA model.
Do you have any idea ? Thanks.