| --- |
| license: apache-2.0 |
| task_categories: |
| - text-classification |
| - feature-extraction |
| language: |
| - en |
| pretty_name: >- |
| This dataset contains GitHub issues (and their comments) collected from |
| the [huggingface/datasets](https://github.com/huggingface/datasets) public |
| repository. It was created as part of an NLP pipeline exercise to |
| demonstrate data collection, augmentation, semantic search, and transformers |
| inference. |
| tags: |
| - github |
| - issues |
| - nlp |
| - text-classification |
| - semantic-search |
| - huggingface |
| size_categories: |
| - 1K<n<10K |
| --- |
| --- |
|
|
| - |
|
|
|
|
| ---- |
| version: v1.0<p> |
|
|
| <u>Data Collection Process</u><br> |
| 1. Data Source: Publicly available records fetched from the GitHub REST API.<br> |
| 2. Collection Method: Using Python-based "requests" library to fetch issue titles, descriptions and metadata.<br> |
| 3. Augmentation: The raw data was transformed into a structured format and saved locally as a .jsonl file before being uploaded to Hugging Face.<br> |
| 4. Tools Used: Use <b>Pandas</b> to manipulate the dataset, drop irrelevant data rows, and use <b>datasets</b> for final repository hosting.<p> |
|
|
| <u>Compliance and policy alignment</u><p> |
| 1. Data Access and Classification Policy: The dataset is classified as "<b>Public</b>". This data is approved for public release as its disclosure causes no harm to the organization.<br> |
| 2. AI Use Policy: The use of the sentence-transformers model for indexing is "<b>Low-Risk</b>" AI use cases. It focuses on information retrieval rather than autlmated decision-making. |
|
|
| <u>Risks</u><p> |
| 1. Data Breach - Low: There aren't personal information in the issues, also only few relevant columns are kept during data cleaning stage. <br> |
| 2. Model Bias - Medium: There is only one repo used in this dataset, which is very limited. Should have included multiple repo resources in the future. |
