JPH03218566A - Machine translation device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] Industrial applications The present invention relates to a machine translation device equipped with a pre-editing function.

2. Description of the Related Art In recent years, there has been remarkable progress in devices related to language processing, one of which is machine translation devices.

Here, if you try to translate the translation target sentence as it is in the original text,
Sometimes parsing fails, and even if it succeeds, the translation may not be valid. In order to prevent these problems, the operator performs pre-editing such as spell checking, but the pre-editing function includes the following means and methods. For example, if the grammatical function of a word (or phrase) in the original text is specified,
JP-A-62-203270 and JP-A-62-203
There are those shown in Publication No. 271 and the like. Further, a user dictionary registration method is disclosed in Japanese Patent Laid-Open No. 62-219165. According to this, when you want a specific translated word, you can register the word in the source language and information such as the word's origin, co-occurrence pattern, and translated word in a user dictionary, and then specify the use of that dictionary before executing the translation. I can do it.

Problems to be Solved by the Invention However, the specification of grammatical functions is mainly used to determine the sentence structure of the source language, and may not be sufficient to select an appropriate translation from multiple translations.

Furthermore, even with the user dictionary registration method, this method cannot deal with words that have multiple meanings in the source language and whose meanings change depending on the context.

For example, when translating from English to Japanese, the translation of "right" in the following example sentence may change depending on the context.

Example sentence: He was on the right
ide. He was on the right side.

He was on the right.

That is, there are two possible translations for ``Correct'' and ``Right'' for ``Correct'' and ``Right'', and it may not be possible to specify either of them, and it may not be possible to obtain an accurate translation as a whole.

Means for Solving the Problem In a machine translation device having a pre-editing function for pre-editing a sentence to be translated, a means for specifying a word in the sentence to be translated is provided during the pre-editing, and the specified word is A means for looking up the dictionary and displaying the contents of the dictionary is provided, a means for selecting a translated word from the displayed contents, and a means for newly specifying and inputting an arbitrary translated word other than the displayed contents.

Furthermore, when a new arbitrary translated word is specified and input, a syntactic analysis unit is provided that analyzes the sentence to be translated using information about the specified translated word and the words and phrases in the source language corresponding to the translated word.

At the pre-editing stage before execution of functional translation, a valid translation of a word with multiple meanings is selected and specified, or a new translation is input and specified for translation, making it possible to obtain a more accurate translation of the sentence to be translated. becomes possible.

In particular, when an arbitrary translation word is newly specified and input, a special translation can be temporarily assigned to that word using this specified translation word and the word can be translated, even if the words are the same. Even when the meaning differs depending on the context, a more accurate translation can be obtained.

Example An embodiment of the present invention will be described based on the drawings.

FIG. 2 is a block diagram showing the configuration of a machine translation device to which the present invention is applied.
A printing section 2 such as a printer, a display section 3 such as a CRT, a translation section 4, and a storage section 5 are connected to a control section 6.

FIG. 3 is an algorithm diagram showing an outline of processing in the case of an English-Japanese machine translation system using such a device in a flowchart. Input text in English (sentence to be translated)
To obtain the output text (translation) in Japanese from
The pre-editing section 7, the morphological analysis section 8, the syntax analysis section 9, the English structure identification section 10, the conversion section 11, the syntax generation section 12, and the morpheme generation section 13 are processed in this order. In these processes, a dictionary 14 that stores information such as word forms, inflections, parts of speech, translations, etc.
However, inflectional word form file 15, information file 1 16,
It is used in the form of information file 2 17, information file 3 18, information file 4 19, etc. as appropriate. Also, the grammar rules 20 are parsing rules 1 21, 2 2
It is used in the form of 2.

To give an overview of the processing in each section, first, the input English text is spell-checked in the pre-editing section 7, and spelling errors and
Words not registered in the dictionary are detected.

You can also specify the part of speech of words in English sentences and set the range of phrases and clauses so that they can be used during syntax analysis. After pre-editing, to begin translation, the morphological analysis unit 8 first recognizes each word as a single word, and special character strings (for example, Mr. Smith) are grouped together as phrases. Next, the structure of the input English sentence is analyzed by the syntactic analysis unit 9. That is, the sentence structure is determined by applying dictionary descriptions and grammatical rules from right to left, starting from the end of the sentence. For example, "verb" + "adverb" is analyzed as "verb phrase J.

The English structure identification unit 10 creates a tree structure with verb nodes as parents in order to clarify dependency relationships and the like. This tree structure is converted by the conversion unit 11 into an intermediate structure that easily represents Japanese phrases. Furthermore, after the syntactic converter 12 generates Japanese phrases in units of clauses, the morpheme generator 13 determines the arrangement of particles and arranges the conjugations of predicates.

Therefore, this embodiment is characterized by the processing by the pre-editing section 7. First, in pre-editing, it is possible to specify a translation. When the operator selects the translated word specification mode during pre-editing, the cursor on the screen of the display unit 3 can be moved to any word using a space or the like as a delimiter. Furthermore, a phrase can be specified by specifying the first word position and the last word position. Selection judgment a in Figure 1
The process indicated by "translation specification mode?" and selection decision b "translation specification?" is a means for specifying a word in an English sentence.

When a word is specified, the word is searched for in the inflected word form file 15 (for example, shown in Fig. 4) in the system dictionary 14, and if a matching word form is found, the word form and the dictionary description contents such as the original form, part of speech, Display part 3 for the translated word, part of speech of the translated word, and conjugation type.
display on the screen. In Figure 1, Process C “Dictionary Search”,
The process indicated by process e [Dictionary content display] corresponds to the dictionary content display means.

If the operator finds a valid word among the displayed translations, he or she can move the cursor to that translation and select that translation. The process indicated by the selection judgment f "translation word selection?" in FIG. 1 corresponds to the translation word selection means. When a translated word is selected, information necessary for later processing, such as the translated word's part of speech, conjugation type, and English part-of-speech subclassification corresponding to the translated word, is taken and stored in a temporary storage area (processing g ).

On the other hand, if there is no valid translation among the displayed translations, the operator is allowed to input a new arbitrary translation. The process indicated by process h "translation, etc. manually" in FIG. 1 corresponds to the translation specifying means. When inputting this specification,
At the same time, a list of subclassifications of English parts of speech, parts of speech and conjugations of the translated word are displayed on the display screen (for example, Figure 5 is an example of displaying English parts of speech, and Figure 6 is an example of displaying conjugations of Japanese parts of speech). ). This information is also stored in a temporary storage area for use in post-processing (
Processing i).

Even if the word in the sentence to be translated specified by the operator is not registered in the dictionary 14, the necessary information such as the translated word, part of speech of the translated word, conjugation type, and English part of speech is inputted and processed according to process h. 1, it can also be stored in the time storage area. If necessary, this information can be registered in the user dictionary and the same information can be used from now on.

In this way, for words for which the operator has temporarily designated a translation word or necessary grammatical information, a sign indicating that information has been designated is erected (process j), and information indicating the position of the word in the sentence to be translated is added. It is also sent to the morphological analysis unit 8.

When the pre-editing is finished and the operator gives an instruction to start translation processing, the pre-edited English sentence is sent to the morphological analysis unit 8. The morphological analysis unit 8 cuts out each word and performs translation. At this time, as shown in FIG. 7, the system dictionary 1
4 or the user dictionary, refer to the temporary storage area,
After comparing the stored word with the position information of the word in the sentence, information about the word is obtained and used in the next processing and subsequent processing.

A specific example will be explained.

Specific example I See that you take the rig
httrain. When this English sentence is translated, it is possible to obtain multiple translations such as the following.

l. Make sure you get on the train on the right.

2. Make sure you get on the right train.

3. ...... In the pre-editing section 7 before executing the translation, the operator selects the target word specification function (selection judgment a), moves the cursor on the display screen, and selects the translation target sentence, for example, as shown in Fig. 8. I of
I right tI+ is specified (process b). r
When "ight"' is selected, the inflectional word form file 15
is searched (processing C), and II right I+ is registered in the dictionary (see judgment d and Figure 4).
, the part of speech, translation word, part of speech of the translation word, and conjugation type, which are the description contents of "right", are taken and displayed as shown in FIG. 9 (process e). Here, if the operator selects "correct" as a valid translation for example sentence 1 (selection judgment f), the part of speech of "correct", conjugation type, 11 right
The part of speech of t l+ is stored in a temporary storage area. Further, the position information in the sentence "right" is also stored (processing g). When the operator finishes the translation specification process,
An indicator is attached to "right" indicating that a translation, etc., has been specified (processing J).

When the operator specifies a translation (selection judgment k), start from the period at the end of the sentence and separate each word with a space, refer to the system dictionary and the user dictionary specified for use, if any, and convert the dictionary. The data is retrieved and stored (processing 0, p). In this case, “right” in the sentence
Since there is a sign indicating that the translated word has been specified, we refer to the temporary storage area without referring to the dictionary, check the word position information, and find the translation of ``'right'', the part of speech of the translated word,
The conjugation type is taken and stored (processing n+p). Continuing,”
After all dictionary lookups are completed, the results are sent to the next step, the syntactic analysis unit 9. After this, syntactic analysis, English structure identification, conversion, and generation processing are performed. Eventually, a translated sentence such as ``Make sure you get on the correct train.'' is obtained using the translated words specified in the previous editing.

Specific example 2 He is the right person for
r the job. When this sentence is translated, it is possible to obtain multiple translations such as the following:

1. He is the right person for the job.

2. He's the right person for the job.

3. ...... When the operator specifies ``right'' in the translation word specification function (processing b) as in the case of specific example 1, the inflectional word form file 15 is searched and the dictionary contents are displayed (see Figure 9). ).In this case, there is no valid translation of "suitable" registered in the dictionary, so the operator inputs "suitable" as a translation using the character keys (process h).
. Further, the part of speech of "right" is selected from the list of parts of speech of the source language (see FIG. 5). List of parts of speech and conjugations of translated words (
(see FIG. 6), these pieces of information are stored in a temporary storage area. At the same time, the position information in the sentence "right" is also stored (processes h, i). Also,
An indicator indicating that a translation, etc., has been specified is attached to "right" (processing J). When the operator finishes pre-editing and instructs to start translation (processing k), a dictionary lookup is performed as in the case of specific example l, but if rlgh in the sentence
Since there is an indicator at t ++ indicating that a translated word, etc. has been specified (judgment m), the word position information is collated by referring to the temporary storage area without referring to the system dictionary or user dictionary (
Processing n), translation of 'right' [suitable]
, the part of speech and conjugation type of the translated word are taken and memorized (processing 0). Next, proceed to the dictionary lookup for “the”, and after all the dictionary lookups are completed, the result is memorized (processing p
), and sends it to the syntax analysis unit 9 in the next step. After this, we went through syntactic analysis, English structure identification, conversion, and generation processing, and finally used the translated words specified in the pre-editing process [He is the right person for the job. You can get a translation like ``.

Effects of the Invention Since the present invention is configured as described above, it is possible to select and designate a valid translation word for polysemic words in the pre-editing stage before translation, or input and designate a new translation word to provide for translation. This makes it possible to obtain a more accurate translation of the target sentence. In particular, when a new arbitrary translation word is specified, a special translation using the specified translation word is temporarily applied to that word. Therefore, even if the same word has a different meaning depending on the context, a more accurate translation can be obtained.

[Brief explanation of drawings]

The drawings show one embodiment of the present invention, and FIG. 1 is a flowchart showing the translation word specification process, FIG. 2 is a block diagram showing the overall configuration, FIG. 3 is an algorithm diagram of the English-Japanese machine translation system, and FIG. Figure 5 is an explanatory diagram showing the contents of an inflectional word file, Figure 5 is an explanatory diagram of an example of English part-of-speech display, Figure 6 is an explanatory diagram of an example of Japanese part-of-speech/inflection type display, and Figure 7 is a flowchart showing morphological analysis processing. , FIG. 8 is an explanatory diagram showing a display example, and FIG. 9 is an explanatory diagram showing an example of dictionary content display. Applicant: Rikowosato Co., Ltd. Revise 3 Z Ward S Ichiatsu, 3 illustrations 7.

Claims (1)

[Scope of Claims] 1. In a machine translation device having a pre-editing function of pre-editing a sentence to be translated, a means for specifying a word in the sentence to be translated is provided during the pre-editing, and a means for specifying a word in the sentence to be translated is provided. Machine translation characterized by providing a means for looking up a dictionary and displaying the contents of the dictionary, a means for selecting a translated word from the displayed contents, and a means for newly specifying and inputting an arbitrary translated word other than the displayed contents. Device. 2. It is characterized by the provision of a syntactic analysis unit that analyzes the target sentence for translation using information regarding the specified translation word and the word/phrase in the source language corresponding to the translation word when a new translation word is input. The machine translation device according to claim 1, wherein: