README.md

---
dataset_info:
  features:
  - name: data_id
    dtype: string
  - name: text
    dtype: string
  - name: exact
    dtype: string
  - name: group
    dtype: string
  - name: ternary
    dtype: string
  - name: binary
    dtype: string
  - name: label
    dtype: string
  splits:
  - name: train
    num_bytes: 14338723
    num_examples: 31318
  - name: val
    num_bytes: 1545579
    num_examples: 3355
  - name: test
    num_bytes: 3285868
    num_examples: 7282
  download_size: 4686461
  dataset_size: 19170170
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: val
    path: data/val-*
  - split: test
    path: data/test-*
license: mit
task_categories:
- text-classification
language:
- ru
size_categories:
- 10K<n<100K
---

This is a derivative dataset from `psytechlab/presuisidal_antisuisidal_dataset-master` that used in the paper experiments "The methodology of constructing the large-scale dataset for detecting presuicidal and anti-suicidal signals in social media texts in Russian".

The preprocessing list:

* removed all classes that have less than 100 examples
* the `re.sub("[^А-Яа-я ]+", "", x)` has been applied to all texts
* all multilabel text was removed

The dataset has all class granularity levels:

* exact - the full class name with upper group and specialization
* group - the group class only (applicable to a presuisidal part)
* ternary - presuisidal, antisuisidal and irrelevant
* binary - relevant and irrelevant

In the paper experiments the presuisidal and antisuisidal parts share the irrelevant texts. The texts from presuisidal part are not marked as irrelevant for antisuisidal dataset and vice versa, only natural `irrelevant` texts are used.