---
language:
- it
license: cc-by-nc-sa-4.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
  - split: test_ood
    path: data/test_ood-*
dataset_info:
  features:
  - name: id
    dtype: string
  - name: text
    dtype: string
  - name: emotion_labels
    sequence:
      class_label:
        names:
          '0': Anger
          '1': Anticipation
          '2': Disgust
          '3': Fear
          '4': Joy
          '5': Love
          '6': Neutral
          '7': Sadness
          '8': Surprise
          '9': Trust
  - name: target_labels
    sequence:
      class_label:
        names:
          '0': Direction
          '1': Topic
  splits:
  - name: train
    num_bytes: 1010988
    num_examples: 5966
  - name: test
    num_bytes: 169792
    num_examples: 1000
  - name: test_ood
    num_bytes: 137719
    num_examples: 1000
  download_size: 844581
  dataset_size: 1318499
---

### EMit
**Disclaimer: This dataset is not the official EMit repository from EVALITA. For the official repository and more information, please visit the [EVALITA EMit page](http://www.di.unito.it/~tutreeb/emit23/index.html) or the [EMit repository](https://github.com/oaraque/emit).**

#### Overview

The EMit dataset is a comprehensive resource for the detection of emotions in Italian social media texts. This dataset was created for the EMit shared task, organized as part of Evalita 2023. The EMit dataset consists of social media messages about TV shows, TV series, music videos, and advertisements. Each message is annotated with one or more of the 8 primary emotions defined by Plutchik (anger, anticipation, disgust, fear, joy, sadness, surprise, trust), as well as an additional label “love.” 

#### Annotations

The dataset includes multilabel annotations for each text, indicating the presence of specific emotions. An additional auxiliary subtask involves identifying the target of the affective comments, whether they are directed at the topic of the content or at issues under the control of the direction (e.g., production quality or artistic direction).

#### Structure

The dataset is composed of the following fields:
- `id`: Identifier for the entry.
- `text`: Social media messages related to TV shows, TV series, music videos, and advertisements.
- `emotion_labels`: Anger, anticipation, disgust, fear, joy, sadness, surprise, trust, and love.
- `target_labels`: Topic, direction, both, or neither.