The Problems of Machine Translation

The problem of automatic (machine) translation of texts appeared more than two decades ago. Its essence is reduced to the construction of an automaton, the input of which is a text in one natural language (for example, English), and the output is generated in another language (for example, Russian). Now in a role of the mentioned automatic machine electronic computers are used, for which algorithms of translation and dictionaries are developed.

The present stage of development of machine translation can be characterized as a stage of syntactic word-by-word translation. The basic unit of meaning is a word; the grammatical forms and order of words in a generated text are determined on the basis of syntactic links between words in an original text. By analogy with the established traditions, dictionaries for automatic translation are built primarily as dictionaries of words (the proportion of word combinations in them is small).

Meanwhile, in natural languages, a word is not the only unit of meaning. They use several levels of semantic units at the same time (the level of morphemes, level of words, level of phrases, level of sentences, etc.), and the meaning of units of a higher level can’t always be determined on the basis of the meaning of units of a lower level (this is most clearly manifested in relations between levels of morphemes and words). Therefore, dictionaries for automatic translation must contain units of different levels.

The centuries-old experience of communication between people speaking and writing in various languages shows that a good translator must not only understand the meaning of a translated text, but also own phraseological richness of a language to which translation is made. Thus, manual translation of texts is more phraseological than word-for-word translation. This circumstance should be taken into account in automatic translation.

Qualitative machine translation of scientific, technical and business texts can be built on the basis of the use of complexes of dictionaries that include semantic units of various levels of words, phrases, sentences. The leading place should be occupied by word combinations, as in the above texts, the names of concepts are most often expressed by word combinations and much less often by single words. In the process of translation, the preference rule should be followed, according to which, first of all, it’s necessary to isolate and translate the semantic units of higher levels, and to units of lower levels to apply in the second turn and only if there is no other possibility to perform translation.

An important place in the machine translation systems should be occupied by an automated vocabulary, the task of which is to create and maintain (addition, adjustment). The role of a person in the automated vocabulary will be to mark up foreign texts in order to separate the elements of dictionaries from them and assign the corresponding equivalent in the target language to the latter. The markup of texts can also be transferred to the machine.

At present, the problem of machine translation in full isn’t yet solved. To solve it, not only profound theoretical studies are needed, but also a large amount of experimental work.

We need to create large, well-organized scientific teams; we need a modern powerful electronic computer with sufficient mathematical (software) support for its work.

Research in the field of machine translation should be attributed to the category of fundamental research. They need to be conducted not only and not so much because it will allow to save money on translations of foreign texts and to facilitate the exchange of scientific and technical and business information between different countries in the future. They’re necessary as one of the most effective ways to find technical solutions for a number of problems of automatic information processing, and the level of development of means and methods of automatic information processing directly affects pace of scientific and technological progress.

Sarah Allen, England

You can safely choose Russian Translation Services. I got qualitative translation of my very difficult text.

Laura Butler, Ireland

RTS was helpful when I almost lost a hope to get a visa. Simultaneously, the price-list is enough democratic one!