|
--- |
|
language: |
|
- tl |
|
license: mit |
|
task_categories: |
|
- text-classification |
|
dataset-info: |
|
- config-name: paragraph-level_10auth |
|
features: |
|
- name: author |
|
dtype: string |
|
- name: sentence |
|
dtype: string |
|
tags: |
|
- literature |
|
- author attribution |
|
size_categories: |
|
- 10K<n<100K |
|
--- |
|
# Panitikan-10 Corpus |
|
Panitikan-10 contains Filipino literary texts from 10 Filipino authors from the late 19th to early 20th century. The data was gathered from the Tagalog books category of [Project Gutenberg](https://www.gutenberg.org/browse/languages/tl) using the [scrapy](https://scrapy.org/) library. Various pre-processing techniques were also applied to the dataset which can also be adopted in other languages as discussed in the paper. Dictionaries, thesauruses, and works translated from other languages were excluded to solely focus on literary texts and the original author's writing style. Given that this is a subset of [Panitikan-34](https://huggingface.co/datasets/Galahallt/Panitikan-34), only the top 10 authors with the most literary works in the website were used. This would lessen the confusion of the models since there are significantly less writing styles to be compared. It also observed that the full dataset is imbalanced in terms of the number of written works per author; thus, this smaller dataset may reduce the bias toward predicting authors with more entries. |
|
|
|
# Dataset Specifications |
|
| Items | Count | |
|
| :---: | :---: | |
|
| No. of tokens | 458,254 | |
|
| Vocabulary Size | 41,210 | |
|
| No. of literary works | 19 | |
|
| No. of authors | 10 | |
