README.md

---
license: mit
task_categories:
- text-classification
language:
- ru
size_categories:
- 10K<n<100K
tags:
- psyhology
- text classification
- suicide
pretty_name: Dataset for presuicidal signal detection
dataset_info:
  features:
  - name: text
    dtype: string
  - name: label
    dtype: int64
  splits:
  - name: train
    num_bytes: 4006893
    num_examples: 22787
  - name: test
    num_bytes: 1721497
    num_examples: 9767
  download_size: 3145819
  dataset_size: 5728390
---
# Dataset Card for Dataset for presuicidal signal detection

<!-- Provide a quick summary of the dataset. -->

This dataset dedicated to find texts that contain information that helps to diagnosis person's suicide rating.

## Dataset Details

### Dataset Description

<!-- Provide a longer summary of what this dataset is. -->

- **Curated by:** Igor Buyanov ([email protected])
- **Language(s) (NLP):** Russian
- **License:** MIT

### Dataset Sources

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

- **Repository:** [link](https://data.mendeley.com/datasets/86v3z38dc7/1)
- **Paper:** [link](https://astromis.github.io/assets/pdf/buyanoviplussochenkovi046.pdf)

## Uses

<!-- Address questions around how the dataset is intended to be used. -->

The dataset is intended to use to train the model that can help the psychologists to analyze the potential suicidal person accounts faster in order to find clues and facts that helps them in threatment.

## Dataset Structure

The dataset has two categories: the normal text (0) and text with potential useful information about person's suicide signals (1). These signals are:
* Texts describing negative events that occurred with the subject in the past or in the present - messages that are factual, describing negative moments that can happen to a person, such as attempts and facts of rape, problems with parents, the fact of being in a psychiatric hospital, facts of self-harm, etc.
* Current negative emotional state - messages containing a display of subjective negative attitude towards oneself and others, including a desire to die, a feeling of pressure from the past, self-hatred, aggressiveness, rage directed at oneself or others.

Note that source dataset that was pointed in **Repository** contains five categories. Due to unrepresentation of some categories and extremeimbalance, the dataset were transformed to have only two categories. See the paper for more details.

The dataset is splitted to train and test parts. Current count distribution is as follows:
```
DatasetDict({
    train: Dataset({
        features: ['text', 'label'],
        num_rows: 22787
    })
    test: Dataset({
        features: ['text', 'label'],
        num_rows: 9767
    })
})
```

## Dataset Creation

### Source Data

Accounts of Russian persons on Twitter that were marked as having tendency to suicide.

### Annotations

<!-- If the dataset contains annotations which are not part of the initial data collection, use this section to describe them. -->

See the paper.

#### Personal and Sensitive Information

<!-- State whether the dataset contains data that might be considered personal, sensitive, or private (e.g., data that reveals addresses, uniquely identifiable names or aliases, racial or ethnic origins, sexual orientations, religious beliefs, political opinions, financial or health data, etc.). If efforts were made to anonymize the data, describe the anonymization process. -->

The dataset may contain some personal information that was shared by Twitter users themselves.

## Citation

<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->

**BibTeX:**
```bibtex
@article{Buyanov2022TheDF,
  title={The dataset for presuicidal signals detection in text and its analysis},
  author={Igor Buyanov and Ilya Sochenkov},
  journal={Computational Linguistics and Intellectual Technologies},
  year={2022},
  month={June},
  number={21},
  pages={81--92},
  url={https://api.semanticscholar.org/CorpusID:253195162},
}
```

## Dataset Card Authors

Igor Buyanov

## Dataset Card Contact

[email protected]