Back to timeline

Availability · December 9, 1975

METEO translates weather forecasts

On 9 December 1975 METEO, from the TAUM group at the University of Montreal, began translating Canada's public weather forecasts from English into French on an experimental basis, with full operation planned for 15 May 1976. Sentences the system accepted went to radio stations and newspapers without human revision.

Why it matters

For the first time, by its builders' account, the output of a machine translation system was to go straight to the public. The closed world of weather forecasts, with a small lexicon and telegraphic syntax, proved narrow enough for fully automatic translation to pay, which general text did not then allow.

John Chandioux, head of the METEO team, at the seminar of March 1976: sentences the system rejected were under 20 per cent of the input, for misspellings, characters blurred in transmission, words not in the dictionary and poor English; a human translator at the Translation Bureau handled those. Estimated load 30,000 words a day at over 1,000 words a minute, all-inclusive cost about a third of human translation. The program is a sequence of rewriting grammars in Q-Systems, run in FORTRAN on a CDC 7600; METEO derives from the TAUM 73 model. Chandioux's article in Meta was not read: érudit.org showed a bot check. The record does not claim 7,000-8,000 words a day, a share of edits or a date in 1981: the document read does not have them. What came after the plan. Benoît Thouin, writing after 1981, gives other dates and figures: the Translation Bureau entrusted TAUM with the system on 15 May 1975; a first version came a year later; the operational version was put in use on 25 May 1977, at first only for the offices at Halifax and Toronto, and for the whole country in June 1978. Of the 18,000 words a day the Meteorological Sub-section translated, about 11,000 went to METEO, which translated about eighty percent of them without human intervention. So 15 May 1976 and 30,000 words a day are the plan and the estimate of the seminar of March 1976, not the later reported operation; the record does not claim that METEO was in regular operation in 1976.

Event record

Event date
December 9, 1975
Timeline date
Event date
Verification
Sources gathered automatically · September 24, 2026
Lines
ID
evt-0690

The talk at the seminar of 8-9 March 1976 says the system had been operating experimentally 'since last 9 December' and was due to be fully operational on 15 May 1976. The year is derived from the seminar's date; the document, written before, cannot say whether full operation began that day. The request's brief gave 1981; no document read supports it.

Sources

Related events

Earlier