JPH03222069A - Machine translation device - Google Patents

Abstract

PURPOSE:To output a translation sentence adjusted to a target by providing a dictionary where classification codes (CB) are added to translation words and respective means which designate CB at the time of translation and preferentially output the translation words with designation CB which is the same as designated CB when there are plural translation word candidates for a word. CONSTITUTION:The type and the style of a document to be translated are selected and designated by a prescribed classification code (CB) table and they are stored in the CB designation buffer (BU) 21 in a main memory (MM) 2. Then, the sentence is inputted and a dictionary consulting morpheme analysis part 51 fetches information from a dictionary/grammar rule/tree structure conversion rule memory (ME) 6, introduces part of speech information to respective words of word BU 22 of MM 2 and introduces the translation word and CB added to respective translation words to translation word BU 23 of MM 2. Then, a construction analysis part 52 executes analysis, applies the grammar rule of ME 6, obtains tree structure, condenses the parts or speech candidates. Then, it refers the translation word of BUS 23 to CB of BU 21, introduces the translation word having CB with first priority to BU 22 and outputs the translation sentence by a generation part 54. Thus, the translation sentence adjusted to the target of a user can be generated.

Description

DETAILED DESCRIPTION OF THE INVENTION <Industrial Application Field> The present invention relates to a machine translation device that performs processing such as analysis, conversion, and generation on an input sentence and outputs a translated sentence.

<Prior art> In conventional machine translation devices, when there are multiple translation candidates for each word or phrase in an input sentence, which translation to output is determined based on grammatical conditions and semantic connections with surrounding words. In the case of interactive machine translation devices, the user selects which translated word to output each time.

<Problem to be solved by the invention> However, in the above-mentioned machine translation device, when the most suitable translation word is determined depending on the field of expertise or writing style of the input text, it is difficult to understand the grammatical conditions and semantic connections with surrounding words. I couldn't narrow down the translation candidates just by looking at the following words.

For example, 'book' is generally translated as 'book', but in the field of economics, 'book' is preferred. Matago performance
' to a formal expression such as 'to carry out' or 'to do'
The choice of soft expressions is a matter of the user's preference and writing style, and conventional machine translation devices have not been able to differentiate between them.

On the other hand, if the user specifies a translated word each time, the desired translated word can be obtained, but there is a problem in that the operating efficiency decreases.

<Means for Solving the Problems> In view of the above-mentioned problems, the present invention provides a machine in which a user specifies fields and writing styles in advance using a classification hood, and outputs translations that are preferentially suited to the user's purpose. The purpose of this machine translation device is to provide a translation device that uses a dictionary to perform syntactic analysis, syntactic conversion, and translation word generation for an input sentence, and then outputs a translated sentence. Translation that adds a classification code to each translated word. means for specifying a classification code at the time of translation execution, and when there are multiple translation candidates for a word in an input sentence, priority is given to outputting a translated word to which the same classification code as the specified classification code is added. The invention is characterized by having means for.

In addition, there is a means for specifying multiple classification codes with priorities at the time of translation execution, and when there are multiple translation candidates for a word in an input sentence, a method for specifying a higher priority among the plurality of specified classification codes is provided. The present invention is characterized by having means for outputting translated words having the same classification code as the classification code with priority.

<Operation> According to the present invention, each translated word in the translation dictionary is attached with a classification code such as the field of the translated word, writing style, etc.
The user sets a desired classification code in advance, and if there are multiple translation candidates for a word in the input sentence during translation, the same classification code as the specified classification code is added. Outputs the available translation candidates first.

In addition, the user can set multiple desired classification codes in advance,
In addition, priorities are assigned to the multiple classification codes, and when there are multiple translation candidates for a word in the input sentence during translation, the translation candidates to which the classification code with the highest priority is added are selected. Output with priority.

<Example> Hereinafter, an example of the present invention will be described in detail based on the drawings.

Figure 1 shows the countries in which the translation device according to the present invention is made up of.
) 1. Programs for processing, etc. are stored, as well as a classification code designation buffer 21, a word buffer 22, which will be described later.
A main memory 2 provided with a translation buffer 23, etc., a CRT display device 3 for displaying translated sentences, a cow hoard 4 for inputting input sentences, and translation of input sentences (original sentences) into output sentences (translated sentences). It consists of a translation module 5 that executes a translation module 5, a memory 6 that stores translation dictionaries, grammar rules, tree structure conversion rules, and the like.

A specific configuration of the translation module 5 is shown in FIG. 2.

The translation module 5 includes a dictionary lookup morphological analysis unit 51, which manually inputs the source language. It consists of four parts: a parsing unit 52, a converting unit 53, and a generating unit 54 that generates the target language.

The dictionary lookup morphological analysis unit 51 converts the input sentence into each morpheme sequence (
Dictionary search is performed to obtain grammatical information such as parts of speech and translations for each word, and further information on tense, person, and
Analyze numbers, etc.

The syntactic analysis unit 52 determines the structure of the sentence (structural analysis tree), such as dependencies between words, as will be described later.

The conversion unit 53 converts the sentence structure obtained by the syntax analysis unit 52 into the sentence structure of the target language.

The generation unit 54 generates a translated sentence according to the sentence structure obtained by the conversion unit 53.

In the translation dictionary of the memory 6 mentioned above, a classification code indicating the field of the translated word, writing style, etc. is added to each translated word according to a predetermined classification food table shown in FIG.

The field is, for example, economics, machinery, information, etc., and the writing style is katakana, Japanese, Chinese, etc.

A predetermined classification code for these classifications is added to each translated word in the translation dictionary.

Hereinafter, specific processing operations when performing translation in the apparatus of the present invention will be explained using FIGS. 3 to 10.

FIG. 10 is a diagram illustrating the processing operation of the translation module of this embodiment.

First, in step Sl, the user selects and specifies the classification hood desired by the user according to the type of document to be translated and the style of text desired as a result of translation, according to the classification code table shown in FIG.

The classification code OOO in the classification code table of FIG. 3 is a code when no classification code is selected. Classification code 001 represents various fields, such as the economic field, classification code 002 the mechanical field, and classification code 003 the information field.

Also, classification code 100 is a translation of katakana, classification code 1
01 represents the Japanese translation, classification code 102 represents the Chinese translation, and so on, indicating the style of the translation.

Classification codes 201, 202, . . . are classification codes for users to define their meanings and add them to their preferred translations.

The classification code designated by the user is stored in the classification code designation buffer 21 of the main memory 2. For example, the first
If the classification code 003 representing the information field is left in the second category, and the classification code 100 representing the katakana translation style is selected, the classification code designation buffer 21 contains the classification code 003 representing the information field, as shown in Figure 4. The code 003 is placed in the second buffer, and the classification hood 100 is placed in the second buffer, and OOO is placed in the third buffer to indicate that there is no other classification code other than the specified one.

Here, the input sentence is The display shows a docum
ent.

Assume that .

The input sentence is taken into the device in step S2, and the input sentence is read into the device in step S2.
In step 3, dictionary lookup is performed by the dictionary lookup morphological analysis unit 51. The dictionary lookup information is taken out from the translation dictionary 6, and part-of-speech information for each word is introduced into the word buffer 22 of the main memory 2 as shown in FIG. 5(a). In addition, the translation for each part of speech of each word and the classification code added to each translation are introduced into the translation word buffer 23 of the main memory 2 as shown in FIGS. 8(a), (b), (c) and (d). be done.

Next, in step S4, the syntax analysis unit 52 performs syntax analysis.

Figure 6 is a part of the analysis grammar rules used in this example, and is stored in the memory 6 (grammar rules).
') means that the - sentence is made up of a noun phrase and a verb phrase, and (b) means that the noun phrase is made up of an article and a noun.

Also, (C) indicates that a verb phrase is made up of a verb and a noun phrase.

For example, the input sentences “The display” and “a d
The ``ocument'' part is grouped into a noun phrase by applying the grammar rule (b): noun phrase → article ten nouns.

Also, the part "shows a document" is combined into a verb phrase by applying the grammar rule (C) verb phrase → verb ten noun phrase.

When the syntax analysis is completed in this manner, a tree structure of the analysis result shown in FIG. 7 is obtained.

As a result, the multiple part-of-speech candidates for each English word in the word buffer of FIG. 5(a) are narrowed down to one, as shown in FIG. 5(b).

That is, "display" is a noun, "shows" is a verb, and "1ocument" is only a noun.

Next, in step S5, the translation for each English word is shown in Figure 8 (a
), (b), (c), and (d) are transferred from the translated word buffer 23 to the word buffer 22 in FIG. be done.

That is, the translated words with the same classification code as OO3, which is the first priority classification code in Fig. 4, are (a), (b), (C),
If it is in the translation buffer of (d), first select it in Figure 5 (b).
) into the translation part of the word buffer. Here, “d
Translations for “display” and “show”
Since the translated word for ``display'' is classification code 003, these are transferred first.

Next is the second priority classification code in Figure 4. Translated words with the same classification code as OO are Figure 8 (a), (b), (C),
Translation of (d)2Go It is checked whether it is in the buffer.

Here, the translated word "display" for "display" and the translated word "document" for "document" are moved to the translated word portion of the word buffer 22.

Since the third priority classification code in FIG. 4 is OOO, the remaining translated words in FIGS. 8(a), (b), (c) and (d) are directly transferred to the translated word portion of the word buffer 22.

As a result, the translation part of the word buffer in Figure 5(b) is
For "display", "display device", "display", "line" in this order, "shows", "display", "show" in that order, and "docume"
nt”, the order is “document” and then “document”.

Next, in step S6, the generation unit 54 shown in FIG. 2 generates a translated sentence, or the translated sentences generated at this time are output in priority order of the translated words stored in the translated word part of the word buffer 22 shown in FIG. 5(b). Therefore, the translated text is output as ``The display device displays the document.'' as shown in FIG. 9(d).

In this way, the classification code specified by the user in advance from among multiple translated words, in this case, Information OO4, Katakana 100,
The translated words that have been added are output preferentially, and a translated sentence is generated.

As another example, if the user selects only 100 classification codes representing katakana, "Display displays document" is output as shown in FIG. 9(C).

If the user selects only the classification code 004 representing the information field, the message "The display device displays the document" is output as shown in FIG. 9(b).

Furthermore, if the user does not specify any classification code, "Display indicates document" is output as shown in FIG. 9(a).

In this way, by allowing the user to specify an arbitrary classification code from the predetermined classification food table with priority, it is possible to generate a translated text that meets the user's purpose.

Furthermore, in addition to the classification codes added in advance to translated words in the translation dictionary 6, a user-defined classification code can be added to any translated word of any word.

For example, the user may use the classification code 200
(User 1), 201 (User 2) codes are added to the third translation of "computer" and the fourth translation of "computer" in Figure 8 (cl). can do.

In step S1 of FIG. 10, the user sets the classification code to 2.
If you specify 00, priority will be given to translated words with a classification code of 200, just like other classification codes.
ter'" can be translated and output as "computer."

Effect> As described above, according to the translation device of the present invention, the user can select and specify a classification code from the classification code table of predetermined fields and translation styles, thereby achieving the purpose of the user. It is possible to output a translation that matches the text.

In addition, a classification code that allows users to define their own usage can be added to any translation of any word in the translation dictionary, making it possible to generate a translation based on the user's preferred translation. become.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a translation device according to the present invention, and FIG.
The figure is a detailed diagram of the translation module, Figure 3 is the classification code table, Figure 4 is the classification code specification buffer, and Figure 5 (a).
and (b) are word buffers, Figure 6 is part of the grammar rules used in syntactic analysis, Figure 7 is a diagram showing the tree structure of the syntactic analysis results, and Figures 8 (a), (b), (C ), (d) are the translated word buffers, Figure 9 (a), (b), (c) and (d) are the contents of the classification code designation buffer and the translations generated in response to each designation, and Figure 1O is the translated text. It is a figure explaining operation of translation processing.

Claims (1)

[Claims] 1. Using a dictionary, perform syntactic analysis, syntactic conversion, and
A machine translation device that generates translated words and outputs translated sentences requires a translation dictionary that adds a classification code to each translated word, a means for specifying the classification code when executing translation, and a means for specifying multiple translated words for each word in an input sentence. A machine translation device comprising: means for preferentially outputting translated words to which the same classification code as the specified classification code is added when there are candidates; 2. In the machine translation device described in paragraph 1, means for specifying a plurality of classification codes with priorities when performing translation; and when there are plural translation candidates for a word in an input sentence, the plurality of specified A machine translation device comprising: means for preferentially outputting translated words having the same classification code as a classification code with a higher priority among the classification codes;