--- task_categories: - text-classification language: - en - af - xh - zu tags: - history - human rights - oral testimony --- # Dataset Card for the Truth and Reconciliation Commission Testimonies Dataset ## Dataset Summary This dataset contains the transcriptions of testimonies from the Truth and Reconciliation Commission (TRC) hearings in South Africa. It includes a wide range of testimonies from different types of hearings such as Amnesty Hearings, Human Rights Violation Hearings, and Special Hearings. We obtained these testimonies from the [South African Historical Archive](https://sabctrc.saha.org.za/). ## Languages English ## Data Instances A typical data point in this dataset comprises a transcript excerpt from a TRC hearing. The data includes metadata about the hearing, such as the type, location, date, and more. An example from the dataset is as follows: ```json { "speaker": "MR PRIOR", "dialogue": "May I proceed Mr Chairman?", "file": "../data/data_saha/amnesty_hearings/aliwal_north/54583.json", "file_index": 0, "saha_page": "https://sabctrc.saha.org.za/documents/amntrans/aliwal_north/54583.htm", "saha_loc": 2, "hearing_type": "Amnesty Hearings", "location": "Aliwal North", "file_num": 54583, "date": "1998-04-28", "umap_x": 4.24962, "umap_y": 8.829064, "hdbscan_label": -1 } ``` ## Data Fields: 1. **`speaker`**: - **Type**: String - **Description**: The name or title of the person speaking. - **Example**: "MR PRIOR" 2. **`dialogue`**: - **Type**: String - **Description**: The content of what the speaker said. - **Example**: "May I proceed Mr Chairman?" 3. **`file`**: - **Type**: String - **Description**: The path to the JSON file containing the full dialogue. - **Example**: "../data/data_saha/amnesty_hearings/aliwal_north/54583.json" 4. **`file_index`**: - **Type**: Integer - **Description**: The index of the testimony in the file. - **Example**: 0 5. **`saha_page`**: - **Type**: URL - **Description**: A link to the SAHA page where the full transcript can be found. - **Example**: "https://sabctrc.saha.org.za/documents/amntrans/aliwal_north/54583.htm" 6. **`saha_loc`**: - **Type**: Integer - **Description**: A numerical identifier for the location of the hearing. - **Example**: 2 7. **`hearing_type`**: - **Type**: String - **Description**: The category of the hearing (e.g., Amnesty Hearings, Human Rights Violation Hearings, Special Hearings). - **Example**: "Amnesty Hearings" 8. **`location`**: - **Type**: String - **Description**: The physical location where the hearing took place. - **Example**: "Aliwal North" 9. **`file_num`**: - **Type**: Integer - **Description**: A unique identifier for the file. - **Example**: 54583 10. **`date`**: - **Type**: Date - **Description**: The date when the hearing took place. - **Example**: "1998-04-28" 11. **`umap_x`, `umap_y`**: - **Type**: Float - **Description**: Coordinates in a UMAP dimensionality reduction, useful for visualization and analysis. - **Example**: 4.24962, 8.829064 12. **`hdbscan_label`**: - **Type**: Integer - **Description**: Label from HDBSCAN clustering, potentially useful for grouping similar testimonies or dialogue themes. - **Example**: -1 ## Curation Rationale This dataset is curated to provide researchers, historians, and the public with access to the testimonies recorded during the Truth and Reconciliation Commission hearings en masse. These testimonies are crucial for understanding the extent of human rights violations and the process of reconciliation in South Africa post-apartheid. ## Initial Data Collection and Normalization The testimonies were initially collected during various TRC hearings across South Africa. The data was then digitized, with care taken to accurately transcribe the spoken words and to preserve the integrity of the original testimonies. It was put online by the [South African Historical Archive](https://sabctrc.saha.org.za/documents.htm) (SAHA). While the data on the SAHA website is accessible to all and while the HTML data some structure to it, much of the metadata is not consistently structured across all testimonies. Further, some of the data is not encoded in UTF-8. Finally, important aspects of the dialogue within testimonies is not preserved. For example, there is currently no way to separate the speaker from their corresponding dialogue. For this dataset, we created custom Python scripts to process each testimony from the original HTML files on the SAHA website. For each section of dialogue in a transcript, users have access to the original SAHA webpage. For this project, we also embedded each segment of testimony, processed those embeddings with [UMAP](https://github.com/lmcinnes/umap) and identified clusters with [HDBScan](). ## Who are the source language producers? The language producers include the victims, witnesses, and other participants in the TRC hearings, who provided their testimonies during the commission's operation. ## Annotations No additional annotations are provided in this dataset beyond the metadata that accompanies each testimony excerpt. ## Annotation process Not applicable. ## Who are the annotators? Not applicable. ## Personal and Sensitive Information This dataset includes sensitive information as it contains real testimonies from individuals involved in or affected by human rights violations. Care should be taken to treat this information with respect and sensitivity. ## Social Impact of Dataset The dataset is a valuable resource for understanding the historical context of human rights violations in South Africa. It provides insights into the impact of apartheid and the subsequent process of reconciliation. It is a vital tool for education and awareness about human rights and social justice. ## Discussion of Biases There may be inherent biases in the dataset due to the nature of the TRC hearings. Some testimonies might not be included, and the dataset may not fully represent all perspectives or experiences during the apartheid era. ## Other Known Limitations The dataset is based on transcriptions and may contain transcription errors. As the data was processed via Python scripts, there may be some inconsistencies in the separation of speaker and dialogue. ## Dataset Curators Steve Davis, Associate Professor of History, University of Kentucky William J.B. Mattingly, Postdoc Fellow, Smithsonian Institution Jennifer Hootman, Coordinator of Digital Scholarship, William T. Young Library, University of Kentucky Robert Vaughan, Research Assistant, University of Kentucky Jamari Turner, Graduate Assistant, University of Kentucky ## Licensing Information MIT License. ## Citation Information ## Contributions Thanks to the Truth and Reconciliation Commission, the individuals who provided their testimonies, and the team involved in the transcription and curation of these testimonies.