The best-known event in the history of machine translation is, without a doubt, the publication of the ALPAC report (Automatic Language Processing Advisory Committee, 1966) in November 1966. Its effect was to bring to an end the substantial funding of MT research in the United States for some twenty years. More significantly, perhaps, was the clear message to the general public and the rest of the scientific community that MT was hopeless. To this day, the "failure" of MT is still repeated by many as an indisputable fact. ALPAC's impact has been undeniable. [caption id="attachment_6740" align="aligncenter" width="604"]
John Robinson Pierce, research director at AT&T, and head of ALPAC.
The first half of the report investigated the translation needs of US scientists and government officials, and overall demand and supply for Russian to English translations. ALPAC began by asking whether, with the overwhelming predominance of English as the language of scientific literature, it might be simpler and more economical for frequent users of Russian to English translations to learn to read the documents in the original language; this could be achieved in 200 hours or less, and an increasing amount of American scientists and engineers possessed the skill. Next, it established that only some 20 to 30% of Russian articles in some fields would have been accepted for publication in American journals. So there was no urgent need for translation.
There were, however, several crucial problems with translation. These concerned quality, speed, and cost. Quality had to be appropriate for the requesters' needs, since flawless and polished translation for a user-limited readership was a waste of both time and money. When it came to speed, ALPAC saw much room for improvement: the most rapid service (from JPRS) was 15 days for 50 pages; documents sent to external contractors by the US Foreign Technology Division were taking a minimum of 65 days; and when processed by the FTD's MT system, they were taking 109 days (due to long post-editing and production processes).
Concerning cost, ALPAC considered what government agencies were paying to human translators, which varied from $9 to $66 per 1000 words. Calculations were made of cost per reader of the different forms of translation, including unedited output from the FTD system. These costs included the time invested by readers. Assuming that the average reader took twice as long to read unedited MT documents, as opposed to good quality human translation, it concluded that if documents are to be read by more than 20 people, traditional human translation was cheaper than MT.
Next, the report turned to the current state of machine translation. It began with a definition: MT "presumably means going by algorithm from machine-readable source text to useful target text, without recourse to human translation or editing and in this context, there has been no machine translation of general scientific text, and none is in immediate prospect." The alternative it saw was post-edited MT. However, it admitted that it could not "assess the difficulty and cost of post-editing."
In some respects, the impact of the ALPAC report is exaggerated. MT research in the US did not come to a complete and sudden halt in 1966. Some projects continued, notably at Wayne State University, under Harry Josselson until 1972, and at the University of Texas, under Winfred Lehmann and Rolf Stachowitz until 1975. Furthermore, in hindsight it can, of course, be agreed that the ALPAC report was quite right to be skeptical about MT, because the quality was undoubtedly poor, and did not appear to justify the level of financial support it had been receiving. It was also correct to identify the need to develop machine assistance for translators, and to emphasize the need for more basic research in computational linguistics.
However, it can be faulted for concentrating too exclusively on the translation needs of US scientists and agencies and not recognizing the broader commercial and industrial needs in an already expanding global economy. Secondly, the report concentrated exclusively on US government and military needs when analyzing and scanning Russian-language documents. It was not concerned in any way with other potential uses or users of MT systems, or with any other languages.
In the chapter on "Automatic language processing and computational linguistics" there is a consideration of the contribution of MT research to NLP advances in general. Its effect on computer hardware had been insignificant; it had contributed to advances in computer software, but that, by far, the most important outcome was its effect on linguistics. Here they highlighted insights into syntax and formal grammar, bringing subtler theories into confrontation with richer bodies of data, and concluding that although "the revolution in linguistics has not been solely the result of attempts at machine translation and parsing, it is unlikely that the revolution would have been extensive or significant without these attempts."
However, despite this favorable influence, the ALPAC report underlined once again that "we do not have useful machine translation and there is no immediate or predictable prospect of useful machine translation." It repeated the potential opportunities to improve translation quality, particularly in machine assistance: "Machine-aided translation may be an important avenue toward better, quicker, and cheaper translation." But ALPAC did not recommend basic research: "What machine-aided translation needs most is good engineering." ALPAC recommended that research should be supported as a means for speeding up the human translation process, the evaluation of the speed and costs of various sorts of machine-aided translation, the adaptation of existing mechanized editing and production processes in translation. For further information about the ALPAC report, please click here.