A history of machine translation is only as good as the documents behind it, and the documents behind this one are unevenly available. The committee reports and the memoranda are held in libraries and have been reprinted; the press coverage that shaped how the field was understood at the time was scattered across trade weeklies and local papers and has mostly never been reprinted at all. Where a contemporary document and a later account disagree, the contemporary document is preferred here and the disagreement is written down rather than smoothed over.

That preference matters most for January 1954. The Georgetown–IBM demonstration is the single event in this history that everyone has heard a version of, and almost every version descends from newspaper coverage rather than from the technical write-up. Reading the press release against what appeared in print the following week shows where the exaggeration entered, and it was not where it is usually said to have entered: the vocabulary and rule counts are stated plainly in the release, and it is the coverage that drops them.

Memoranda and reports

  • Weaver, W. (1949) — Translation. Memorandum circulated July 1949; reprinted in Locke & Booth, Machine Translation of Languages (MIT Press, 1955).
  • Bar-Hillel, Y. (1960) — The present status of automatic translation of languages. Advances in Computers 1.
  • ALPAC (1966) — Languages and machines: computers in translation and linguistics. National Academy of Sciences, Publication 1416.

Contemporary press coverage

  • The New York Times, 8 January 1954 — report of the Georgetown–IBM demonstration. The gap between this coverage and the technical write-ups is itself a source.
  • IBM press release, 8 January 1954 — the vocabulary and rule counts are stated plainly here, which is why the later claim of concealment does not hold.

Technical and product documentation

  • Dostert, L. (1955) — The Georgetown–IBM experiment, in Locke & Booth.
  • Systran, Logos, METAL, ATLAS and PIVOT product literature, as catalogued in the compendium.

What the contemporary record does not settle

Three things stay open. The size of the 1954 vocabulary is stated consistently, but which sentences were selected, and by whom, is not recorded anywhere that survives. The internal reaction at the sponsoring agencies between 1962 and 1966 exists mainly as committee minutes that were never published. And the European and Soviet work of the same period is documented in languages and venues that the English-language accounts largely did not read, which is why the standard chronology is thinner after 1966 than before it.

Secondary literature

The survey work reproduced in the publications bibliography is the principal secondary source. It is used here for chronology and for the reconstruction of the early period, and it is cited rather than paraphrased silently.