---
language:
- ar
size_categories:
- 1K<n<10K
task_categories:
- conversational
- text-generation
- text-classification
tags:
- mental health
dataset_info:
  features:
  - name: content
    dtype: string
  - name: text_size
    dtype: int64
  - name: topic
    dtype: string
  - name: prob
    dtype: float64
  splits:
  - name: train
    num_bytes: 6007437.514440433
    num_examples: 1884
  download_size: 2896563
  dataset_size: 6007437.514440433
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---
# Dataset Card for Nafsy 

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

This arabic dataset is a set of mental health articles. The original dataset was scrapped from [Nafsy.net](https://nafsy.net/).
## Dataset Details

**Language(s) (NLP):** Arabic

## Uses

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


### Direct Use

<!-- This section describes suitable use cases for the dataset. -->

Fine-tuning llm in the mental health domain.

## 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. -->

It is a CSV file with columns:
- content: the articles
- text_size: length of article
- topic: top 10 words that describe the topics of the article
- prob: topic prediction accuracy
   
## Dataset Creation

### Curation Rationale

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

Creating an arabic chatbot for mental health support.

### Source Data

<!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->

#### Data Collection and Processing

<!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->

- This dataset was originally scrapped from [Nafsy.net](https://nafsy.net/) then uploaded to Kaggle.
- An additional preprocessing was made by this repo owner:
  - Cleaning data: removing urls, extra spaces, and non words, detach punctuations, and dropping duplicates
  - Applying Topic Modeling to generate main topics for each article using bert-base-arabic model
  - Deduplicating data using sentence-transformers (paraphrase-multilingual-MiniLM-L12-v2)

#### Who are the source data producers?

<!-- This section describes the people or systems who originally created the data. It should also include self-reported demographic or identity information for the source data creators if this information is available. -->

[husamal](https://www.kaggle.com/husamal)

## Citation [optional]

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

**BibTeX:**

@misc{Husamal_2021, title={Arabic-physcology-dataset}, url={https://www.kaggle.com/datasets/husamal/arabicphyscologydataset?select=nafsy.csv}, journal={Kaggle}, author={Husamal}, year={2021}, month={May}} 


## Dataset Card Authors

Muhammad Helmy

## Dataset Card Contact

muhammadhelmymmo@gmail.com