20 Mar ,2020
The development of machine translation is not smooth
At present, machine translation is undoubtedly another technology direction that is widely recognized as having a broad application prospect. However, the history of the technology has not been smooth, and its feasibility and practicality have been continuously questioned.
However, this historical history of machine translation technology is not smooth sailing. It was even once regarded as a nightmare, and its possibility and practicality have been continuously questioned. Throughout the history of the development of machine translation technology, it can be basically divided into five stages: creation, sprout, silence, recovery, and development.
The research history of machine translation can be traced back to the 1930s and 1940s. In the early 1930s, French scientist G.B.Archuni put forward the idea of using machines to translate.
Soon, the pioneer of information theory, American scientist Warren Weaver, proposed the idea of using a computer for automatic language translation in 1947. In 1949, Warren Weaver published the Translation Memorandum, which officially proposed the idea of machine translation.
In 1954, in collaboration with IBM, Georgetown University in the United States completed the first English-Russian machine translation experiment with the IBM-701 computer, demonstrating the feasibility of machine translation to the public and the scientific community, thereby opening the prelude to machine translation research.
The United States, the former Soviet Union, and some European countries have given considerable understanding to machine translation research.
However, while everything is progressing in an orderly way, the machine translation research that is still in its infancy is holding its head.
In 1964, the American Academy of Sciences established the Automatic Language Processing Advisory Committee. After two years of research, the committee announced a report entitled Language and Machine in 1966.
The report completely denied the feasibility of machine translation, and at the same time there is no hope for the development of a practical machine translation system in the near future or foreseeable future. Affected by this report, various types of machine translation projects have plummeted, and research on machine translation has experienced an unprecedented depression.
In the mid-to-late 1970s, through the development of computer technology and linguistics and the demand for social information services, machine translation began to recover and flourish. Developing countries have developed many translation systems, such as Weinder, EUROPOTRAA, TAUM-METEO, etc.
Among them, the TAUM-METEO system, which was jointly developed by the University of Montreal in Canada and the Federal Government of Canada in 1976, is a sign in the history of machine translation and marks the recovery of machine translation from prosperity.
The word-alignment-based translation model proposed by IBM's Brown and Della Pietra in 1993 marked the birth of modern statistical machine translation methods.
In 2003, Koehn of the University of Edinburgh proposed a transformation model, which significantly improved the effect of machine translation. At the same time, Franz Och proposed a logarithmic linear model and its weight training method. As a result, translation models began to be widely used in industry.
In 2005, David Chang further proposed a hierarchical model. At the same time, many universities and research institutes have also made great progress in the research of syntax tree-based translation models.