Outline of the Research By

The history of machine translation (MT) research at the University of Washington in Seattle from its beginnings in November, 1949 until June 30, 1958 has already been outlined in our previous Technical Report.1 This research, which was inspired by Dr. Warren Weaver's memorandum Translation2 of July 15, 1949, and which was in 1952 and 1953 supported by grants from the Rockefeller Foundation, has since May, 1956 been sponsored by RADC, ARDC, USAF, by grants totaling $235,500. Under the USAF contract, the University of Washington undertook to study the lexicographical, linguistic and engineering prerequisites for the automatic translation of scientific Russian into English. The University was specifically charged with the preparation of a Russian-English lexicon to be used in an automatic system designed by Dr. Gilbert W. King, then with the International Telemeter Corporation of Los Angeles and now with IBM. This automatic system includes a photoscopic memory device which has, or soon will have, sufficient storage capacity for all lexicographical requirements of machine translation. Since Dr. King's automatic system did not yet include logical equipment for linguistic purposes, it was decided to attempt to solve as many bilingual linguistic problems as possible by purely lexicographical means. It was, however, clear from the outset that such a lexicographical solution of linguistic problems would be possible only in the case of certain types of divergent structure and multiple meaning. Therefore, investigations were initiated very early for the purpose of developing logical procedures and machine programs for the solution of the remaining linguistic problems. During the initial, lexicographical, phase of the research, the project compiled a lexicon of about 14,000 Russian-English entries whose Russian "semantic units"3 were taken from 111 Russian texts covering 40 fields of science. They did, of course, in these texts not occur in all their possible paradigmatic transformations. It was therefore necessary to supplement all those paradigmatic forms which did not happen to occur in the 111 texts, but were likely to occur in other scientific texts. This supplementation work had already been initiated during the previous report period. At the end of that period, in June, 1958, it was estimated that the supplementation would increase the number of entries to about 200,000. The final count at the conclusion of the lexicographical phase of the project in June, 1959, showed that our estimate had been too generous. By that time the sponsors of the project had received from us a RussianEnglish MT lexicon of 170,563 entries on 556,141 IBM punch cards.