The Use of Computers in Research on Machine Translation

Work on machine translation has been carried on in various countries for several years. During this time, a number of research groups have carried out machine translation of more or less extensive texts, from one language into another. In some cases, these were experimental translations of a relatively small number of sentences, in order to test the proposed algorithm; in other cases, long texts have been translated, but as a rule, with simplified algorithms and with reduced requirements as to the quality of translation. At present it is clear that the construction of algorithms by which computers can produce high-quality translation is quite feasible; but in order that these algorithms should have the necessary potentialities, and the computers be effectively used, it is necessary to solve a number of different theoretical problems. In order to achieve success in machine translation, it is essential to formulate the problems accurately, and to systematize both the problems and the difficulties arising in their solution. This is especially important if we consider not only the achievement of practical results in machine translation, but also the advance towards the wider use of computers to assist man in the various problems which involve the handling of verbal information, and which include MT as a special case.