HistoryLab/foiarchive

Dataset Summary

Our multidisciplinary team of researchers has gathered nearly 5 million documents, comprising over 18 million pages, to create the Freedom of Information Archive (FOIArchive), the world's largest database of declassified government records. The FOIA Archive data contains the full text and selected metadata from History Lab's collection of declassified government documents.

Dataset structure

The data are in JSONL format with 9 different fields.

Data fields

• doc_id - Unique document identifier
• classification - Original classification status of document
• authored - Original author of document (if known)
• title - Title of document
• body - Text of document
• source - Collection document is from
• pg_cnt - Number of pages in the original document
• char_cnt - Number of characters in the document
• word_cnt - Number of words in the document

Data summary

source	collection	start date	end data	document count	page count	word count
briefing	Presidential Daily Briefings	1946-02-15	1977-01-20	9680	81633	10851841
cabinet	UK Cabinet Papers	1907-10-19	1990-12-13	42539		110622458
cfpf	State Department Central Foreign Policy Files	1973-01-01	1979-12-31	3214293	3764465	443965365
cia	CIA CREST Collection	1941-01-01	2005-08-14	935716	12391471	1655689803
clinton	Clinton E-Mail	2009-03-09	2013-07-07	54149		6537986
cpdoc	Azeredo da Silveira Papers	1973-11-15	1979-11-24	10279		11943789
frus	Foreign Relations of the United States	1620-11-03	1989-01-17	311866	750389	190038744
kissinger	Kissinger Telephone Conversations	1973-01-02	1976-12-24	4552		1722000
nato	NATO Archives	1949-09-17	2013-03-07	46002	281963	107319669
un	United Nations Archives	1997-01-01	2016-12-31	192541	943484	258542774
worldbank	World Bank Archives	1942-04-01	2020-09-16	128254	128254	29004048

Note: Some of the documents did not come as separate pages, so a page count is not possible.

Languages

The majority of documents across all collections are in English. The UN, NATO, and World Bank collections also contain some documents written in French. There are a smattering of documents in the other collections written in a language other than English.

Collections

Detailed information about our collections is available on our website. Here, we briefly describe each of the collections in the file.

• briefings - This collection contains the daily briefings a U.S. president receives from intelligence agencies from 1951 to 1961. The briefing was originally named Daily Summary, a name which it held from 1946 to 1951. The CIA has released the entirety of the Daily Summaries and the Current/Central Intelligence Briefings (1951-1961) as well as all of the President's Daily Briefs between 1961 and 1977.

• cabinet - The UK Cabinet Papers collection are the official records of the most important decision-making body within the United Kingdom. The collection is predominantly divided into conclusions and memoranda from 1907 to 1990.

• cfpf - Our processed collection of these cables comes from the Central Foreign Policy Files at the National Archives, and includes cables, records of cables withdrawn for classification reasons, and so-called “P-Reel” records corresponding to communications sent physically via diplomatic pouch. We currently hold all available digital records from this collection, which spans 1973-1979.

• cia - The FOIArchive provides access to the CIA records made available via the automatic declassification provisions of Executive Order 13526, which requires declassifying nonexempt historically valuable records 25 years or older.

• clinton - The Clinton E-Mail collection consists of 54,149 individual messages covering her tenure as Secretary of State from 2009 to 2012.

• cpdoc - The Azeredo da Silveira collection is comprised of the personal papers of the Brazilian foreign minister during the presidency of Ernesto Geisel, with documents spanning 1974 to 1979. The papers are provided to us by The Center for Research and Documentation of the Contemporary History of Brazil (CPDOC) at <a href="https://portal.fgv.br/en"Fundação Getulio Vargas (FGV), where the original paper documents are housed.

• frus - Foreign Relations of the United States, colloquially known as FRUS, is published by the Office of the Historian of the State Department and represents the official diplomatic history of the United States.

• kissinger - The Kissinger collection is a series that comes to us from the State Department’s FOIA Reading Room. It is made up of transcriptions of recorded phone conversations involving Henry Kissinger, from 1973 to 1976.

• nato - In 2020, we partnered with NATO's digital archives to obtain documents that have been declassified through the NATO’s Public Disclosure Program. The digital collection provides a documentary record of the first thirty years of the alliance system, with particularly extensive records on the Alliance’s formation and earliest operations.

• un - The United Nations collection contains about 193,000 documents for the years 1997 to 2016. The collection is made up of two main source: 1) the digital records from Kofi Annan's tenure as Secretary General and 2) the digital records from Ban Ki-Moon's time as Secretary-General of the United Nations.

• worldbank - We have worked with the World Bank Group Archives to make some of their collections available on our website. The largest subcollection consists of the records of Robert Strange McNamara. Other subcollections include the papers of other World Bank Presidents as well as Records of the Office of the Treasurer and Central Files.

Dataset Curators

Numerous people have worked on FOIArchive over time. Principal Investigator Matthew Connelly has been a constant, overseeing efforts since History Lab started. Eric Gade, Daniel Krasner, Thomas Nyberg, and Rohan Shah wrote most of the scripts that made the early collections possible. Marii Nyrop rewrote the FRUS ingestion scripts to make them more efficient and replicable. Raymond Hicks and Ben Lis were instrumental in ingesting the later collections and in making the final product available.

About History Lab

The History Lab's mission is to use data science to recover and repair the fabric of the past. We are beginning with declassified documents, which include some of the earliest examples of electronic records. By bringing together fragmented collections in a common database, we can use natural language processing and machine learning tools to explore them. The ultimate goal is to develop history as a data science so that citizens can keep the government accountable in the age of big data and AI.