README.md

---
task_categories:
- text-classification
language:
- ru
pretty_name: GreenRu
size_categories:
- n<1K
---

### Dataset Description

A Russian dataset for detecting mentions of green practices in social media texts.

The dataset has a sentence-level markup and contains 1,326 texts including 3,513 annotated sentences collected in Russian online communities. GreenRu contains the mentions of the following practices: waste sorting, studying the product labeling, waste recycling, signing petitions, refusing purchases, exchanging, sharing, participating in actions to promote responsible consumption, repairing. 

- **Language:** Russian

### Load Dataset

To load the dataset, install the library datasets with ```pip install datasets```. Then,

```python
from datasets import load_dataset

dataset = load_dataset("beogradjanka/GreenRu")
```

to download train and test sets.

### Dataset Sources

<!-- Provide the basic links for the dataset. -->

- **Repository:** <a href="https://github.com/green-solutions-lab/GreenRu">GreenRu</a>
- **Paper:** Zakharova, O.; Glazkova, A. GreenRu: A Russian Dataset for Detecting Mentions of Green Practices in Social Media Posts. Appl. Sci. 2024, 14, 4466. https://doi.org/10.3390/app14114466 


## Dataset Structure

The dataset is presented in a form of two csv files containing training and test subsets. 
Each entry of the files contains the practice ID (column "id_practice", the practice ID from the table below), the name of the green practice ("name\_practice"), the sentence containing the mention of the practice ("span"), the full text of the post ("text"), and the post ID ("id_post"). 

<table>
    <tr>
        <td>Characteristic</td>
        <td>Training subset</td>
        <td>Test subset</td>
    </tr>
    <tr>
        <td>Number of posts</td>
        <td>913</td>
        <td>413</td>
    </tr>
    <tr>
        <td>Symbols</td>
        <td>880.05</td>
        <td>908.53</td>
    </tr>
    <tr>
        <td>Tokens</td>
        <td>154.91</td>
        <td>162.33</td>
    </tr>
    <tr>
        <td>Symbols</td>
        <td>111.35</td>
        <td>114.99</td>
    </tr>
    <tr>
        <td>Tokens</td>
        <td>18.85</td>
        <td>19.86</td>
    </tr>
    <tr>
        <td>Number of posts # Waste sorting (ID = 1)</td>
        <td>1275</td>
        <td>560</td>
    </tr>
    <tr>
        <td>Number of posts # Studying the product labeling (ID = 2)</td>
        <td>55</td>
        <td>17</td>
    </tr>
    <tr>
        <td>Number of posts # Waste recycling (ID = 3)</td>
        <td>272</td>
        <td>121</td>
    </tr>
    <tr>
        <td>Number of posts # Signing petitions (ID = 4)</td>
        <td>22</td>
        <td>31</td>
    </tr>
    <tr>
        <td>Number of posts # Refusing purchases (ID = 5)</td>
        <td>236</td>
        <td>75</td>
    </tr>
      <tr>
        <td>Number of posts # Exchanging (ID = 6)</td>
        <td>146</td>
        <td>52</td>
    </tr>
    <tr>
        <td>Number of posts # Sharing (ID = 7)</td>
        <td>109</td>
        <td>62</td>
    </tr>
    <tr>
        <td>Number of posts # Participating in actions to promote responsible consumption (ID = 8)</td>
        <td>510</td>
        <td>209</td>
    </tr>
    <tr>
        <td>Number of posts # Repairing (ID = 9)</td>
        <td>10</td>
        <td>3</td>
    </tr>
</table>

## Dataset Creation

### Source Data

Social media posts from the VKontakte social network.

The research centers around the green practices of Tyumen, a sizable town in the Russian Federation with a population of 850,000. The GreenRu dataset consists of social media posts from the VKontakte social network. A social graph was employed to identify VKontakte communities discussing waste sorting. This graph was built to group VKontakte green communities of Tyumen covering significant user topics such as animals, eco-food, eco-markets, and separate waste collection <a href="https://www.researchgate.net/publication/351339333_Green_Practices_Ways_to_Investigation">[more details]</a>. Another essential selection factor was the availability of posts created from January 2021 to June 2021 during data collection. Thus, six VKontakte communities for text collection were selected. 

#### Data Collection and Processing

To collect posts, the VK API tool was utilized. GreenRu includes only the posts that contain textual information and does not include duplicate posts.

#### Annotation process

Two experts on green practices from the University of Tyumen, Russia, annotated the sentences. Initially, they labeled each sentence with the relevant types of green practices, allowing for multiple labels per sentence. Both experts worked independently on this first stage. In the second stage, they reviewed and adjusted their labels. If there were any discrepancies, they resolved them through discussion.

## Citation

```
@Article{app14114466,
AUTHOR = {Zakharova, Olga and Glazkova, Anna},
TITLE = {GreenRu: A Russian Dataset for Detecting Mentions of Green Practices in Social Media Posts},
JOURNAL = {Applied Sciences},
VOLUME = {14},
YEAR = {2024},
NUMBER = {11},
ARTICLE-NUMBER = {4466},
URL = {https://www.mdpi.com/2076-3417/14/11/4466},
ISSN = {2076-3417},
DOI = {10.3390/app14114466}
}
```