Dear all, we are glad to present to you a new BERT-based model for Sentiment Analysis ( for Italian ), trained and benchmarked on multiple domains!
The model has been jointly optimized and fine-tuned on multiple domains such as product reviews, social media comments and financial news. The model has achieved better performance than fine-tuning it in isolation on every single dataset, reaching state-of-the-art results in the majority of the datasets that we used.
To get and use the model please, follow the instructions available here: https://sisl.disi.unitn.it/itfn-corpus/ https://sisl.disi.unitn.it/itfn-corpus/ or you can go directly to the official GitLab repo: https://gitlab.com/sislab/multi-source-multi-domain-sentiment-analysis-with-... https://gitlab.com/sislab/multi-source-multi-domain-sentiment-analysis-with-bert-based-models
The related paper will be presented at LREC 2022 conference. The paper is titled as "Multi-source Multi-domain Sentiment Analysis with BERT-based Models" and it is available here http://www.lrec-conf.org/proceedings/lrec2022/pdf/2022.lrec-1.62.pdf.
Best Regards ---- Prof. Dr.-Ing. Giuseppe Riccardi Founder and Director of the Signals and Interactive Systems Lab Department of the Department of Computer Science and Engineering Department University of Trento