metadata

task_categories:
  - text-classification
language:
  - th
tags:
  - not-for-all-audiences
size_categories:
  - 1K<n<10K

dataset_info: features: - name: accusation dtype: string - name: category dtype: string splits: - name: train num_bytes: [estimated_training_size_in_bytes] num_examples: 1214 - name: test num_bytes: [estimated_testing_size_in_bytes] num_examples: 300 download_size: [total_compressed_download_size_in_bytes] dataset_size: [total_uncompressed_dataset_size_in_bytes] configs: - config_name: default data_files: - split: train path: "dataset.parquet"

Dataset Card for NACC Categorization Dataset

This dataset is designed to train models for categorizing citizen-submitted accusations according to the National Anti-Corruption Commission (NACC) policies. It supports developing machine learning models to automate and streamline the categorization process, enhancing both efficiency and consistency.

Dataset Details

Dataset Description

Purpose: To support a model specialized in categorizing accusations per the NACC's classification policy, enabling accurate and consistent categorization.
Content Type: Text.
Language(s): Primarily in the language used by the NACC (specify language if possible).
Domain: Anti-corruption and public policy classification.
Curated by: National Anti-Corruption Commission (NACC).
License: King Prajadhipok's Institute (KPI).

Dataset Sources

Repository: Currently hosted and utilized for model training on Google Colab using the Unsloth library.
Paper: Pending publication by NACC.
Provided by: NACC.

Uses

Direct Use

This dataset is suitable for developing and training models that categorize public accusations submitted to the NACC by citizens. Such a model could help expedite and standardize the categorization process, improving operational efficiency.

Out-of-Scope Use

This dataset should not be used for purposes outside of categorizing citizen accusations for the NACC. Using it in other domains, such as unrelated text classification tasks, could yield inaccurate results or misinterpretations, as it is specific to NACC's categorization policy.

Dataset Structure

Fields:
- accusation: The text content of accusations submitted by citizens.
- category: The category assigned to each accusation according to NACC policy.
Splits: With 1710 test samples, the dataset split is as follows: Training samples: 1,410 Testing samples: 300

Dataset Creation

Curation Rationale

Motivation: The dataset was created to improve consistency and speed in the categorization of accusations, making the process more efficient for NACC.

Source Data

Data Collection and Processing: Collected, processed, and filtered by the NACC research team to ensure relevance and quality for categorization model training.
Who are the source data producers? The data was provided directly by the NACC.

Annotations

Annotation Process: Information on whether the dataset contains additional annotations beyond the category label is currently unavailable.
Who are the annotators? Information on who performed the categorization labeling, if applicable, is currently unavailable.

Personal and Sensitive Information

This dataset does not contain any private, sensitive, or personally identifiable information.

Bias, Risks, and Limitations

This dataset may contain certain biases inherent to the NACC’s existing categorization framework, though these are currently unexamined. Further analysis of possible biases or limitations would be beneficial for responsible use.

Recommendations

Users should be cautious of possible biases and limitations in categorization, which may arise from the initial data collection and categorization practices by NACC.

Citation

The recommended citation format is APA. Details for citing a pending NACC publication related to this dataset will be added once available.

Dataset Card Authors

VAP Solution

Dataset Card Contact

info@vapsolution.app