الترجمة الآلية - صنع الحاسبات .. الحاسبات تعمل

Machine Translation - Making Computers .. Computers Work

Machine Translation

The machine translation of languages ​​​​appeared by the computer in the early fifties and is still in continuous development parallel to the development of the capabilities of computer systems. Machine translation is considered one of the most complex applications, as there is no point in translating each word separately, but rather the rules of grammar and word order in the two languages ​​​​between which are being translated must be taken into account. Since the computer does not understand it, machine translation requires providing the computer with fixed rules for the rules of grammar in the two languages ​​​​and a dictionary to convert words between them.

However, several factors make the matter more complicated. A single word in the original text may have several meanings in the language into which it is translated, so the true meaning of a word cannot be known except from the context of the text. While it is easy for a human translator to choose the appropriate meaning, the computer is unable to understand the meaning and thus choose the appropriate word. Or a word in the text may not have a synonym in the language into which it is translated. While the human translator relies on his ability to formulate the sentence to convey the desired meaning, the computer is unable to do so.

Due to these difficulties, machine translation may not be 100% correct. The accuracy of the translation is estimated by the percentage of errors in a hundred sentences, and the accuracy rate may reach ninety or ninety-five percent at the present time. However, a correct translation does not necessarily mean a good translation, as the translation may be correct from a linguistic point of view, but its formulation may be poor, and since it is difficult to estimate the quality of the translation by a specific percentage, such sentences are considered correct. It is clear from all this that the machine translated text must be reviewed by human editors to revise it in most cases and put it in its final form.

The simplified context diagram on the opposite page illustrates the main steps in translating a text from French to English. Sentences are entered using a touchpad appropriate to the language of the text to be translated, and each word is paired with a number indicating its sequence in the sentence. The sentence is stored temporarily while a sorting program sorts the words alphabetically. Alphabetical order is necessary because the dictionary words are stored alphabetically on magnetic tape. If you search the words according to their original order in the sentence, this may require moving the tape back and forth between each word, which wastes a lot of time. However, if you search the words alphabetically, the tape only moves forward from one word to the next, which can save a lot of time. This is, of course, what someone who wants to search a large number of words in a regular dictionary does. The words are rearranged after being translated based on the numbers associated with them. Next comes the grammatical analysis stage, which is the most complex of all the translation stages. The original text is analyzed word by word, and the relationship of each word to other words is examined, based on the rules of

(Contextual statement of the translation process by computer. It may be difficult to reach a specific decision regarding the translation of some words despite the use of a complex and advanced program).

The morphology and syntax of that language, taking into account all exceptions to these rules and special expressions and phrases. This information may help the computer choose the appropriate meaning for words in the text that have more than one meaning, provided that these meanings are stored in the context in which each of them appears. In some cases, as in the example shown on page 106, the computer cannot choose the appropriate meaning except by referring to the sentences preceding or following the relevant sentence, but this complicates matters greatly and only giant systems can do it.

(Center for Machine Translation)