Which AI models available on Hugging Face are most useful for detecting emotions or opinions in text for real applications?
For English, I guess I should share my own Orthogonal Model of Emotions (OME) built with Transformers and PyTorch using my own curated and MIT-licensed dataset. The latest is here. There is no other model that has anything near the accuracy on even some Shakespearean insults as well as contemporary slang. Also, there is no publicly available dataset with as much coverage as OME, including a null set and three levels of clarity. Finally, the model and the dataset reflect the theoretical basis of OME, in particular, the dimensionality, but in addition, the old adage for guessing how we feel: “Use your words!”
Thanks in advance for checking it out (and sorry about the self-promotion, but ten years of effort have been invested in this project—so, I have to share).