---
license: cc-by-sa-4.0
dataset_info:
  features:
  - name: header
    dtype: string
  - name: abstract
    dtype: string
  - name: keys
    dtype: string
  - name: text
    sequence: string
  splits:
  - name: train
    num_bytes: 11239555
    num_examples: 1160
  download_size: 5133568
  dataset_size: 11239555
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
language:
- ru
tags:
- nlp
- segmentation
size_categories:
- 1K<n<10K
---

# Dataset Card for Dataset Name

Russian science corpus with paragraph annotation and more

This dataset card aims to be a base template for new datasets. It has been generated using [this raw template](https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/templates/datasetcard_template.md?plain=1).

## Dataset Details

### Dataset Description

The small Russian corpus was constructed from scientific papers written by high school and university entry course students. Be aware that texts are not cleaned of redundant punctuation and noise. Some tasks that this corpus might be suitable for:

Some dataset statistics: 

* 925165 of tokens in total
* 11.5 Mb of text
* 12659 paragraphs with mean sentenc count is 3.2 and std is 2.5
* 1160 papers with mean token length is 798 and std is 532

- **Language(s) (NLP):** Russian
- **License:** CC-BY-SA-4.0

### Dataset Sources

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

- **Repository:** https://github.com/Astromis/Small-Student-Science-Corpus

## Uses

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

### Direct Use

* Paragraph segmentation
* Keyword extraction
* Title generation
* Summurization
* Static word vector construction

## Dataset Structure

<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->

Each data example has four fields:
* header - header of the paper
* abstract
* keys - keywords
* text - a list of texts in which every element is a paragraph

## Dataset Creation

### Curation Rationale

<!-- Motivation for the creation of this dataset. -->

The main objective is to create a small dataset that allows to test an automatic text segmentation method.

### Annotations 

As the paper texts come from the PDF, the annotators carefully reviewed each plain text to be sure that its paragraphs go along with the source PDF text version.