Dataset Card for the TRC Volume 7 Final Report Victim List

Dataset Summary

Languages

English

Data Instances

A typical data point comprises the metadata for a victim in Volume 7 of the TRC's Final Report.

An example from the dataset looks as follows:

{
    "full_name": "Thabo Simon Aaron",
    "lastname": "Aaron",
    "firstname": "Thabo Simon",
    "description": "An ANCYL member who was shot and severely injured by SAP members at Lephoi, Bethulie, Orange Free State (OFS) on 17 April 1991. Police opened fire on a gathering at an ANC supporter's house following a dispute between two neighbours, one of whom was linked to the ANC and the other to the SAP and a councillor.",
    "place": [
        "Bethulie"
    ],
    "homeland": [],
    "province": [
        "Orange Free State"
    ],
    "hrv": [
        "shoot",
        "injure"
    ],
    "org": [
        "ANC (African National Congress)",
        "ANCYL (African National Congress Youth League)",
        "Police",
        "SAP (South African Police)"
    ],
    "date": [
        datetime.date(1991, 4, 17)
    ],
    "coordinates": [
        25.97552,
        -30.50329
    ],
    "age": 22,
    "gender": "Unknown"
}

Data Fields:

full_name:
- Type: String
- Description: The complete name of the individual. It contains the first name(s) and last name combined.
- Example: "Thabo Simon Aaron"
lastname:
- Type: String
- Description: The last name or surname of the individual.
- Example: "Aaron"
firstname:
- Type: String
- Description: The first name(s) or given name(s) of the individual. May contain middle names if applicable.
- Example: "Thabo Simon"
description:
- Type: String
- Description: A textual description providing contextual information about the individual. This may include details about significant events, affiliations, or incidents related to the person.
- Example: "An ANCYL member who was shot and severely injured..."
place:
- Type: List of Strings
- Description: The location(s) associated with the individual or significant events related to them. Each element in the list represents a different place.
- Example: ["Bethulie"]
homeland:
- Type: List of Strings (empty lists allowed)
- Description: The homeland or origin of the individual. May be empty if this information is not available or not applicable.
province:
- Type: List of Strings
- Description: The province(s) associated with the individual or significant events related to them. Each element in the list represents a different province.
- Example: ["Orange Free State"]
hrv:
- Type: List of Strings
- Description: Codes or keywords representing human rights violations or significant events that the individual experienced or was involved in.
- Example: ["shoot", "injure"]
org:
- Type: List of Strings
- Description: Organizations or groups that the individual is affiliated with or related to. Each element in the list represents a different organization.
- Example: ["ANC (African National Congress)", "ANCYL (African National Congress Youth League)", "Police", "SAP (South African Police)"]
date:
- Type: List of Date Objects
- Description: Date(s) associated with significant events related to the individual. Each element in the list represents a different date.
- Example: [datetime.date(1991, 4, 17)]
coordinates:
- Type: List of Floats
- Description: Geographic coordinates related to the individual or significant events. Typically, the list contains two elements representing latitude and longitude.
- Example: [25.97552, -30.50329]
age:
- Type: Integer
- Description: The age of the individual.
- Example: 22
gender:
- Type: String
- Description: The gender of the individual. Can be "Male", "Female", or "Unknown" if the gender information is not available or not specified.
- Example: "Unknown"

Each field in the data provides specific information about an individual, their affiliations, and significant events or attributes related to them. The data is structured to facilitate easy analysis and retrieval of information about individuals and their experiences or affiliations.

Curation Rationale

This data was made available as a PDF in the early 2000s at the conclusion of the TRC. The original data had only three fields for each victim: name, description, and age. While useful for dissemination, it made accessing the data systematically challenging. At some point SAHA-SABC converted the data into a CSV file. They ensured that each field was represented by a unique column. Unfortunately, the data was not encoded in UTF-8. Further, additional vital metadata, such as the types of human rights violations, organizations, dates, etc. were not extracted.

In 2019, we began working with this data to extract this vital metadata using a combination of heuristics and machine learning. This dataset is unique in both is size and specificity for human rights violations. By making this data available, researchers will be able to understand the scale of violence in South Africa during the 20th century across time and space. Further, it will allow researchers to understand the impact of powerful organizations, such as the South African Police.

Initial Data Collection and Normalization

In 2019, we encoded the data into UTF-8. Next, we normalized the text so that there were no trailing or leading white spaces. We also ensured that certain peculiarities, such as double white spaces in the middle of a text were removed. Finally, we provided proper None values

Who are the source language producers?

The initial testimonies were collected by administrators and interviewers for the TRC. These interviews were then interrogated and researchers for the TRC determined which individuals met the threshold to be including in this the Volume 7 Final Report.

Annotations

Our additional annotations include the fields: hrv, places, homeland, provinces, gender, and dates.

Annotation process

To annotate the data, we leveraged Python and spaCy to automate the identification of categories of human rights violations, places, homelands, provinces, gender, and dates. To identify coordinates, we automated the mapping of identifiable places with no toponyms in South Africa, such as Cape Town. For more challenging places (approximately 2,000 locations), we performed manual validation. In some cases, specifically places with many toponyms, we provided the most likely candidate.

Who are the annotators?

The annotators for places were Steve Davis (PI) Robert Vaughan (Geographer).

Personal and Sensitive Information

The dataset contains the personal details of victims of human rights violations in South Africa. This data is already publicly available.

Social Impact of Dataset

Discussion of Biases

This dataset was cultivated originally by a group of researchers at the TRC who were responsible for only including the victims of provable human rights violations. In addition to this, many victims did not come forward to record their accounts. This is especially true for women and victims of sexual assault. As a result, this dataset should not be viewed as a comprehensive list of victims of human rights violations.

Other Known Limitations

In nearly ever instance of a place's coordinates, we are using the central location for that place. In some rare cases, we were able to identify a specific geolocation because it was a well-studied and named attack, such as the Trojan Horse Incident or named bombings of particular buildings. In many cases, the coordinates are rough estimates for the location.

Dataset Curators

Steve Davis, Associate Professor of History at the University of Kentucky

William J.B. Mattingly, Postdoc Fellow at the Smithsonian Institution

Robert Vaughan

Licensing Information

Citation Information

Contributions

Thanks to the TRC for cultivating the original data. Thanks to SAHA-SABC for making the initial dataset available as a CSV file.