Feel It Italian Emotion

🚀 FEEL-IT: Emotion and Sentiment Classification for the Italian Language

FEEL-IT is a novel benchmark corpus for Italian Twitter posts, enabling both emotion and sentiment classification with competitive results.

🚀 Quick Start

You can find the Python package that uses this model for emotion and sentiment classification here. It provides a very simple interface over HuggingFace models.

✨ Features

• Dual Classification: FEEL-IT allows for both emotion classification (anger, fear, joy, sadness) and sentiment analysis on Italian text.
• State-of-the-Art Performance: The feel-it-italian-emotion model, fine - tuned on the UmBERTo model, achieves state - of - the - art results on different benchmark corpora.
• Open - Source Library: An open - source Python library is released for researchers to use the model for inference.

📦 Installation

The installation details are not provided in the original README. If you want to use the package, you can refer to the package repository for installation instructions.

💻 Usage Examples

Basic Usage

from transformers import pipeline
classifier = pipeline("text-classification",model='MilaNLProc/feel-it-italian-emotion',top_k=2)
prediction = classifier("Oggi sono proprio contento!")
print(prediction)

📚 Documentation

Abstract

Sentiment analysis is a common task to understand people's reactions online. However, more nuanced information is often needed, such as whether a negative post is due to anger or sadness.

An abundance of approaches has been introduced for both emotion and sentiment classification. But for Italian, most of them handle only one task at a time. We introduce FEEL-IT, a novel benchmark corpus of Italian Twitter posts annotated with four basic emotions: anger, fear, joy, sadness. By collapsing these emotions, we can also perform sentiment analysis. We evaluate our corpus on benchmark datasets for both tasks and obtain competitive results.

Model

The feel-it-italian-emotion model performs emotion classification (joy, fear, anger, sadness) on Italian. We fine - tuned the UmBERTo model on our new dataset (FEEL - IT) and achieved state - of - the - art performances on different benchmark corpora.

Data

Our data has been collected by annotating tweets from a broad range of topics. In total, we have 2037 tweets annotated with an emotion label. More details can be found in our paper (https://aclanthology.org/2021.wassa-1.8/).

Performance

We evaluate our performance using MultiEmotions-It. This dataset differs from FEEL - IT in terms of topic variety and the considered social media (YouTube and Facebook). We considered only the subset of emotions present in FEEL - IT. To provide a reference, we also show the Most Frequent Class (MFC) baseline results. The results show that training on FEEL - IT brings stable performance even on datasets from different contexts.

Model Download

🔧 Technical Details

The technical details about the fine - tuning process and the construction of the FEEL - IT dataset can be found in our paper (https://aclanthology.org/2021.wassa-1.8/).

📄 License

Users should refer to the following license

Citation

Please use the following bibtex entry if you use this model in your project:

@inproceedings{bianchi2021feel,
    title = {{"FEEL-IT: Emotion and Sentiment Classification for the Italian Language"}},
    author = "Bianchi, Federico and Nozza, Debora and Hovy, Dirk",
    booktitle = "Proceedings of the 11th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis",
    year = "2021",
    publisher = "Association for Computational Linguistics",
}