---
license: mit
task_categories:
- question-answering
- summarization
language:
- en
tags:
- legal
pretty_name: JFK TELL
size_categories:
- 100M<n<1B
---
# Dataset Card for JFK TELL

This dataset contains text that was extracted from the John F. Kennedy assassination records released
on [archives.org](https://www.archives.gov/research/jfk)
by the US government.

## Dataset Details

### Dataset Description
The data in this dataset is generated by processing another HF dataset, the [JFK-Archives](https://huggingface.co/datasets/farhanhubble/jfk-archives) dataset, 
which consists of PDF files originally released on archives.org. The PDF files have been proessed using
the Google Gemini API to generate Markdown text. 

Since the release of JFK records is ongoing, this and [JFK-Archives](https://huggingface.co/datasets/farhanhubble/jfk-archives)
datasets are both snapshots of the archives.org data as of April 2025.


- **Curated by:** Farhan Ahmad
- **Funded by [optional]:** Farhan Ahmad
- **Language(s) (NLP):** English
- **License:** MIT

### Dataset Sources [optional]


- **Repository:** https://github.com/farhanhubble/jfk-tell


## Uses
The dataset can be used to index the JFK assassination records for search, summarization and Q&A.

## Dataset Structure
The data is a single parquet file. The `file_path` column in the parquet file can be used to 
identify the year(s) of release for each data point. The `text` column is the text extracted
from the corresponding PDF file, in Markdown format.

|file_path|text|
|---------|-----|
|2017-2018/104-10180-10145.md|CLASSIFIED MESSAGE 104-10180-10145 ...|
|2022/docid-32331372.md| JFK Assassination System Identification Form Date ...|

The years(s) are: [2017-2018, 2021, 2022, 2023, 2025]


## Dataset Creation

### Curation Rationale
The JFK assassination has been an unassailable mystery even after decades of investigations by
premier agencies, the media and ordinary people. A large-scale analysis of the assassination
records may offer new clues, and help substantiate or refute some of the theories. There are about
six millions files related to the event that are being made public through archives.org over time.

This HF dataset makes it easier for anyone interested in performing such an analysis to access
the text data, build search indexes and use AI tools to navigate and summarize the huge volume
of information. Hopefully, writers, journalists, computational linguists and data scientists
can try their hands on the breadth and variety of this data.

### Source Data

https://www.archives.gov/research/jfk

#### Data Collection and Processing

Markdown text files were generated by processing the [JFK-Archives](https://huggingface.co/datasets/farhanhubble/jfk-archives) dataset,
with Google Gemini API using a simple prompt. The code and prompts for the processing are available through
[this](https://github.com/farhanhubble/jfk-tell) repo.
The download and text-processing are lossy processes so some files could be missing and some
others may be empty.

#### Who are the source data producers?

[Archives.org](https://www.archives.gov/research/jfk)


#### Personal and Sensitive Information

Some of the records may contain personal information. Please see the relevant archives.org page
for details. For example: https://www.archives.gov/research/jfk/release-2025


## Dataset Card Authors [optional]

Farhan Ahmad

## Dataset Card Contact

https://huggingface.co/farhanhubble