The ALPAC report

A National Academy of Sciences committee concluded that machine translation was slower, costlier and worse than human translation, and recommended ending its funding.

Why it matters

The first official finding that a result had not been achieved cut US machine translation funding for two decades.

The report did not dismiss computational linguistics; it recommended basic research in it. But its verdict on the applied programme was unambiguous, and US funding of machine translation was cut for a long time. The historian Hutchins says the effect of the report can be exaggerated: several publicly funded projects had ended before it (Washington and Michigan 1962, Harvard 1964, Georgetown without funds from 1963), while Wayne State continued to 1972 and Texas to 1975. Together with the 1973 Lighthill report it defined the period of reduced funding later called a winter.

Event record

Event date

November 1966

Timeline date

Event date

Verification

Sources gathered automatically · September 17, 2026

Lines

ID

evt-0106

The report was published in November 1966.

Sources

primary Language and Machines: Computers in Translation and Linguistics
National Academy of Sciences · Published November 1966
secondary ALPAC: the (in)famous report, John Hutchins
MT News International, no. 14, June 1996 (copy at the MT Archive) · Published June 1996

Related events

Earlier

Contrasts Georgetown-IBM experiment 1954
The report assessed the very programme the 1954 demonstration set off.
- Language and Machines: Computers in Translation and Linguistics · National Academy of Sciences

Later

Related Section 203: research must serve a military purpose 1969links to this record
Both are decisions of US public bodies about what the budget should pay for: ALPAC advised cutting the funding of translation research, the 1969 law demanded a military relationship of research.
Contrasts SYSTRAN in daily use 1970links to this record
The 1976 seminar opens with its editor's words 'More than a decade after the ALPAC report', asking which systems were 'in regular use'; by Toma's biography, SYSTRAN was on the drawing board in 1964 while the ALPAC hearings were in progress.
- FBIS Seminar on Machine Translation (American Journal of Computational Linguistics, Microfiche 46) · Association for Computational Linguistics (ACL Anthology)
Related The ARPA Speech Understanding Research programme 1971links to this record
The agency returned to language five years after the ALPAC report stopped machine translation.
Contrasts METEO translates weather forecasts 1975links to this record
METEO appears in the same 1976 proceedings, which open 'More than a decade after the ALPAC report', in the section on available systems.
- FBIS Seminar on Machine Translation (American Journal of Computational Linguistics, Microfiche 46) · Association for Computational Linguistics (ACL Anthology)
Related Statistical machine translation 1988links to this record
It revives the machine translation ALPAC stopped, now on statistics instead of rules.