XIII Международная студенческая научная конференция Студенческий научный форум - 2021

The problem of translatability is related to the fact that reality is divided differently by different languages. It is based on this fact that many linguists have questioned the possibility of translating texts from one language to another. The most vivid idea of the irreducibility of two pictures of the world described by different languages to each other was expressed in linguistics by V. Humboldt: "Each language describes around the people. To which it belongs, a circle from which you can only get out if you enter another circle." The study of a foreign language could therefore be likened to the acquisition of a new point of view in the old worldview; just because we transfer our own worldview and our own linguistic outlook to a foreign language to a greater or lesser extent, we do not feel with full clarity the results of this process.

The development of machine translation rules from one language to another has presented linguists with a number of problems that relate both to translation from different languages and into different languages, and to the principles of text analysis in a given language. problems of ambiguity; problems of structural and lexical differences between languages; grammatical constructions consisting of several words, such as idioms and phrases.

The problem of ambiguity. In the best case for the translator, each word of the language would have one and only one meaning. But this is far from the case. When a single word has multiple meanings, it is said that there is lexical ambiguity. If the sentence can have several different structures, they say about structural ambiguity. The case where a sentence can have multiple meanings is a rule rather than an exception. Complicating the work is the fact that ambiguities can accumulate. For example, if a sentence consists of two words and each can have two meanings, in the worst case, it can be translated in four different ways. Naturally, you may need to consider all possible translation cases to choose only one of them. The only suitable solution is to make the translation process interactive. When this kind of ambiguity appears, you will have to ask the person sitting at the computer which of the translation options he considers the most meaningful.

Grammatical differences in languages The difference in the type of grammatical (and specifically morphological) structure of a language causes some differences in the nature of translation rules for a given language. The need for such differences was already discovered when developing the rules for translation from English to Russian, since the English language does not have a number of forms, as a result of which the agreement is already presented as a form of communication between the members of the phrase. The linguistic task of machine translation is to develop such a system of rules that would allow, having some text in a given language, to unambiguously determine which text of another language should be adequate in meaning to this text. Any language has a vocabulary and grammatical structure. Therefore, any text of any language contains words that are related to each other according to the norms of the grammatical structure of this language. Different grammatical relationships serve to express different relationships between concepts expressed by words. Therefore, the rules of translation from one language to another should contain both the rules of translation of words, and the rules of transfer of grammatical means of one language expressing some relations, grammatical means of another language expressing the same relations.

Complex constructions - idioms and phrases. An idiom is a set of words that has its own meaning, independent of the meaning of the constituent parts.

Idioms are similar to language spaces, and probably the solution to this problem should be the same. Idioms should be identified at one of the initial stages of translation to avoid their loss, and processed by the system as a single word. The difficulty arises only in the question of linking the idiom to the word by which it will be searched. In the previous example, you can set the search condition for an idiom by the verb " to kick "to search for its ending, or by the noun"bucket". The problem is that idioms can be written in several forms and these forms may not always be formed according to the usual grammatical rules.

In the translation machine, it is necessary to provide a program for selecting the correct equivalent. Such programs are usually based on two principles:

It is based on the principle of choosing an equivalent according to the syntactic model of the input text, most often according to the syntactic model of the sentence. Thus, for example, the automaton can distinguish between the equivalents of the verb "to book" and the noun " book»;

On the principle of choosing an equivalent according to the semantic model. According to different semantic models, the automaton, for example, can distinguish such equivalents of the word "solution" as "solution"and " solution". Both models are usually used in combination. Both the models themselves and the procedures for selecting equivalents are quite complex.

In some more complex systems, in addition to these two principles of choosing an equivalent, the principle of choosing based on non-linguistic (background) information is also used. Models for choosing equivalents that work on this principle are even more complex: they are classified as artificial intelligence models.

The first, lowest level includes the simplest models of word-by-word translation, in which the choice of equivalents is not made and the output of the translation system receives all the translated equivalents available in the dictionary.

Second-level systems, which include almost all the so-called "electronic translators" available on the modern software market, use some combination of syntactic and semantic models to select the correct equivalent and transform the structure of the input text into the structure of the translation text.

Finally, third-level models, in addition to grammar and semantics, also use background knowledge to synthesize the translation text.

A good translation of a text is not only creative, but also quite time-consuming work. As for the creative part, in the foreseeable future, the "live" translator will always win in the computer-human competition. However, MP systems can be a good tool for solving problems caused by the complexity of the translation process. In order to better understand this, we will list the advantages of machine translation programs:

Speed. Just a few seconds and you get a translation of a multi-page text. This allows you to quickly understand the meaning of the text.

Cost. If you apply to professional translators, you have to pay for each page of the translated text, or you hire a full-time translator who has to pay a salary. In the case of a machine translation system, you only pay money once - when you buy the program.

Translation of information on the Internet. In most cases, you can translate information on the Internet, unless you yourself know several languages, only with the help of translation programs. Only thanks to the online systems of the MP, it became possible to view foreign sites without having to translate them

Versatility. Any translator always has a specialization, i.e. translates texts on the topic that he knows well. The MP system is advantageous in that it is universal. You only need to correctly connect a specialized dictionary on the relevant topic.

Confidentiality. You can entrust any information to the MP system. The translator program will keep secret any texts that you entrust to it.

Conclusion

Progress does not stand still, and with the increase in the speed of processors and the increase in the amount of RAM, the quality of machine translation is gradually increasing. The ultimate goal is to create a machine that provides a sufficiently correct translation that does not need to be edited later. In the meantime, machine translation needs to be edited. The most promising direction of development of machine translation systems is to improve the subsystems of grammatical analysis and synthesis, as well as to increase the volume of contextual coverage of the text and improve semantic chains in order to more accurately select the meanings of words. In conclusion, we emphasize that the translator program is, first of all, a tool that allows you to solve translation problems or increase the efficiency of the translator's work only if it is used correctly.

References

1.Kuznetsov P. S., Lyapunov A. A., Reformatsky A. A. Basic problems of machine translation. Questions of Linguistics, 1956

2. Marchuk Y. N. Problems of machine translation. Moscow: Nauka, 1983. - 232 p.

3. Marchuk Y. N. The text-to-text model and translation correspondences in the theory of machine translation. Problems of computational linguistics // Minsk state linguistic University. 1997. - 202 p.

4. Nelyubin L. L. Computer linguistics and machine translation. Moscow: vtsp, 1991. - 198 p.

5. Frolov S. V., Pankova D. A. Problems of machine translation construction 2008. -130 p.

Код для цитирования: Скопировать