JPH03191472A - Sentence analyzer and machine translation device using the same - Google Patents

Abstract

PURPOSE:To quickly and accurately attain analytical processing by specifying the grammatical categories of words and sentences in an inputted original and executing the analytical processing based upon the specification. CONSTITUTION:At the time of operating a keyboard in an input part 1, character codes corresponding to depressed character keys are successively set up in an input buffer of a control part 8. The inputted original is displayed on a display part 6. Then, which word or word group out of the displayed contents is applied to the specification of a category is determined. The category is selected by a cursor moving key to specify the category. An analyzation part 5 executes analytical processing based upon the specified grammatical category by using an analytical dictionary part 4. In the analytical control, only the candidates of a part of speach based upon the category specification are sent to the analyzation part 5. Since the syntax analysis of other categories is not executed, the analytical processing can quickly be executed without generating fuzzines.

Description

[Detailed Description of the Invention] [Objective of the Invention] (Industrial Application Field) The present invention provides a sentence analysis device that grammatically analyzes natural language sentences, and a sentence analysis device that translates data in a first language into data in a second language. Regarding machine translation.

(Prior Art) First, a conventional sentence analysis device will be described. When words and sentences expressed in natural language (hereinafter referred to as original text) are input, conventional sentence analysis devices perform analysis processing on the original text using an analysis dictionary, and display the analysis results.

The analysis dictionary includes a conjugation/change dictionary, a word/idiom dictionary,
and an analysis grammar, etc., and the analysis unit performs analysis processing by referring to these.

For example, a sentence analysis device K having the above-mentioned configuration is “Th
e v-al-u6 is convH-ted 1n
The following sentence is input: to l; he form of (u, v),' and analysis processing is performed. In this case, conventional su analysis equipment analyzes the active arc as an inserted phrase.

However, if "(u,,v)" is treated as a noun phrase, it is difficult to analyze it even with conventional sentence analysis devices. In other words, it is extremely difficult for conventional sentence analysis devices to distinguish between syntactic usages, such as whether a word or sentence within an active arc is an inserted phrase or an element of a sentence.

Furthermore, in a conventional sentence analysis device, for example, when j=100
゜the 5ubroutine will 5top
”. Here, 1=100
# should be parsed as one sentence in the original text, such as "When i is 100, the subroutine will stop." However, with conventional sentence analysis devices, it is extremely difficult to analyze the sentence "1=100#" as a sentence (one noun clause).

Note that a noun clause indicates a structure of ``subject and ten predicates,'' and words such as @(u, v) # and the like are noun phrases.

Furthermore, some words have multiple parts of speech. For example, the word "1ike" has four parts of speech: verb, noun, adjective, and adverb.

In addition, there are words that have different parts of speech, such as prepositions and conjunctions, and the number of words is not limited to one. For this reason, the number of combinations of parts of speech increases in proportion to the number of words, and the analysis processing time increases accordingly. Further, even if an analysis result is obtained, the result is one of many part-of-speech candidate sequences and contains ambiguity, so there is a problem that it lacks accuracy.

Furthermore, the above can also be said about conventional machine translation devices. In other words, “When 1=100゜t
he 5ubroutine will 5top,'
It was extremely difficult to translate the data in the first language mentioned above. Translation processing proceeds through the processes of analysis processing, conversion processing, and generation processing, but as mentioned above, there was a problem when performing analysis processing as described above. Therefore, when performing translation processing, there are problems in that it takes a lot of time and accurate translation results cannot be obtained.

(Problems to be Solved by the Invention) As mentioned above, conventional sentence analysis devices cannot analyze K when metalinguistic expressions are incorporated into the input original text, or when the analysis results are The problem was that it took a lot of time to obtain it.

For this reason, similar problems exist in machine translation devices that incorporate conventional sentence analysis devices as part of the translation section.

Here, metalinguistic expressions will be explained. First, a metalanguage is a language used to talk about language itself. Based on this, a metalinguistic expression is an expression that explains the nominative case in a noun clause, for example. This is an explanation of "1#", such as "i is 100", and can be said to be a meta-1! expression.

In addition to this, it also includes f-word external phenomena that are different from normal linguistic phenomena such as the above-mentioned "(u, v)". This is “(U,V)
'' alone is a set of meanings. Since this explains the active arc P within the active arc, it can be called a metalanguage l?g' expression.

The present invention was made to solve the above-mentioned problems, and allows an operator to easily specify a grammatical category for any word or group of words in the original text, and performs analysis processing based on this. The purpose of the present invention is to provide a sentence analysis device that is characterized by the following functions.

Furthermore, it is an object of the present invention to provide a machine translation device characterized in that an operator can easily specify a grammatical category for arbitrary data in a first language, and translation processing is performed based on this. . Note that the grammatical categories refer to noun phrases, verb phrases, adverb phrases, adverb clauses, adjective phrases, adjective clauses, and the like.

(Means for Solving Problem a) In order to achieve the above object, the sentence analysis device of the present invention:
an input unit for inputting words and sentences expressed in a natural language to be analyzed; an analysis dictionary unit that stores knowledge used in analysis processing; and an analysis dictionary unit that uses the analysis dictionary unit to input the original text input from the input unit. It includes an analysis section that performs analysis processing, and a display section that displays the analysis results of the original text obtained by the analysis section. The input section is not only a means for inputting the original text, but also for inputting a grammatical category for any word or word group of the KN sentence.

Based on this designation, the analysis section performs analysis processing.

In addition, in order to achieve the other object mentioned above, a machine translation device incorporating the sentence analysis device of the present invention as a part of the translation section includes an input section for inputting data in the first language and an input section for inputting data in the first language. a dictionary unit that records necessary dictionary data; a translation unit that translates data in a first language input from the input unit into data in a second language with reference to the dictionary unit; and a display section for displaying data in the second language. The input unit is not only a means for inputting data in the first language, but also inputs a grammatical category for arbitrary data in the first language. Based on this input, the translation section performs translation processing.

(Operation) According to the sentence analysis device of the present invention configured as described above,
The operator inputs the grammatical category IJ- for any word or word group in the original text from the input section. Based on this input, the analysis section performs analysis processing. This allows the analysis unit to perform analysis processing only on the specified grammatical category. Therefore, the burden on the analysis section is reduced, and analysis processing can be performed at high speed even if the same analysis section as the conventional one is used. Furthermore, by specifying a grammatical category by the operator, the part-of-speech candidates for a word or word group that has multiple part-of-speech candidates in the original text are narrowed down, and ambiguity can be eliminated.

This also reduces the burden on the analysis section and allows analysis processing to be performed quickly and accurately.

Furthermore, according to a machine translation device incorporating the sentence analysis device of the present invention configured as described above as a part of the translation section,
Specify a grammatical category for arbitrary data in the first language from the input section. Based on this input, the translation section performs translation processing. As a result, the analysis process within the translation process can be performed more quickly and accurately than before. As a result, translation processing can be performed more quickly and accurately than before.

(Example) Hereinafter, the present invention will be described in detail with reference to the drawings.

FIG. 1 shows a block diagram of one embodiment of the invention.

In FIG. 1, an input unit 1 is used to input words and sentences expressed in a natural language to be analyzed, and to input a grammatical category for any word or group of words in the original text. For this reason, a keyboard is usually used. The original text storage unit 2 is for storing the original text input from the input unit 1. The analysis result storage unit 3 stores the analysis results of the input original text. In addition,
For example, a hard disk or a RAM disk is used as the original text storage section 2 and the analysis result storage section 3. Analysis dictionary part 4,
It stores knowledge used in analysis processing, and consists of three dictionaries: a conjugation-translation dictionary 4a, a word-idiom dictionary 4b, and an analysis grammar 4C.

This conjugation/transformation dictionary 4a stores, for example, basic forms of words, corresponding conjugation/transformation rules, and modified examples for each (t5).

Further, the word/idiom dictionary 4b stores the part of speech, concept, etc. of each simple compound word. The parsing grammar 4C stores parsing grammars.

The analysis section 5 analyzes the original text input from the input section 1 or the original text stored in the original text storage section 2 using the analysis dictionary 4 . The display section 6 displays the original text input from the input section 1, the results of analysis processing, grammatical categories, etc. (a 1% T display, etc. is used. This is used to control the screen when displaying the original text and analysis results.

The control unit 8 controls the entire device.

FIG. 2 is a diagram showing an example of the key layout of the keyboard of the input unit 1. The input unit 1 includes the following various keys in addition to the character keys for inputting the original text.

Analysis instruction key = 100 Edit key: 101-104 Kanle movement key: 110-113 Cursor enlargement key: 114 Cursor reduction key = 115 Category specification key: 120 Confirmation key: 130 Cancel key: 131 Analysis process end key = 140 Fig. 3 is a flowchart for explaining the operation of the control section 8. Step 8301.8304゜8306.
At step 8308, the key input from the input unit 1 is determined. In step 5301, it is determined whether a character key on the keyboard of the input unit 1 shown in FIG. 2 has been pressed. Here, if a character key is pressed, it is determined that the operator is inputting the original text. Then, in step 5302, the character codes corresponding to the pressed character keys are sequentially set in the input capacitors in the control unit 8. This input original text is sent to the display unit 6 via the display control unit 7.
(Step 8303).

In addition, if input editing such as correction, insertion, deletion, etc. is required during the original writing, use the cursor reduction keys: 110 to 113.
After moving the cursor to the desired editing location using the keys, these operations can be performed using editing keys such as insert key: 10L, delete key: 102, etc. Next, in step 5304, it is determined whether the analysis instruction key 100 has been pressed from the input unit 1. Since the analysis process cannot be performed unless there is at least one sentence, in that case, proceed to the following step. Here, if one or more sentences have been input and the analysis instruction key: 100 is pressed, step 830
In step 5, analysis processing is performed. This analysis process will be explained in detail later. Next, in step 8306, it is determined whether the category designation key 120 has been pressed from the input unit 1. If there is key personnel here, step 53
In step 07, processing for specifying a category is performed. This process will be explained in detail later.

If other key personnel are performed, step 8308
Then, the process corresponding to the key human power is performed. This refers to processing such as deleting characters on the screen when the delete key=102 is input. In addition, the other keys are the keys mentioned above, and the analysis process ends with key = 1.
Judgment will be made based on the presence or absence of key human resources based on 40. If the analysis processing end key: 140 is pressed here, step 5311
In step 5301, the original text inputted and the analysis result of the original text in step 5305 are stored in the original text storage unit 2 and analysis result storage unit 3, respectively, and the analysis process is ended. Note that this analysis result includes the word or word group for which a category has been specified and which category has been specified for the word or word group. Further, if the analysis processing end key: 140 is not pressed, step 8301
Repeat the process from.

FIG. 4 shows step S305 in FIG. That is, it is a flowchart of analysis processing in the analysis section 5. When it is determined in step 5304 in FIG. 3 that the analysis instruction key: 100 has been input, the analysis section 5 starts its operation. When the key input is received, first in step 5401, morphological analysis is performed using the conjugation/change dictionary 4a. This morphological analysis involves distinguishing between words in the original text, which is a list of words, to obtain combinations of word strings. In order to obtain this word string, it is necessary to use the conjugation/inflection dictionary 4a to convert words with changes in their endings, etc. to their basic forms. Next, in step 5402, a dictionary search is performed. This is to search the word/idiom dictionary 4b for each word that became the basic form in step 5401.

Through this search, the part of speech for each word and additional information unique to the word are obtained. Then, in step 5403, analysis control is performed. This is to send the part-of-speech candidates for the word or word group obtained in step 8402 to the next step. Furthermore, in step 8404, parsing is performed using the parsing grammar 4c. This is to analyze the sentence structure based on the part-of-speech candidates sent from step 8403. In step 8405, the step 5404
Determine the result of parsing. Here step 540
In step 4, if analysis cannot be achieved even after using all the grammars in the analysis grammar 4c, the process returns to step 8403 and analysis is performed again based on another part of speech candidate for the word or word group.

If the analysis is successful, the analysis process at step 8305 in the llX3 diagram ends.

FIG. 5 is a flowchart of category designation in step 8307 of FIG. This process starts when there is an input from the category designation key=120 in step 8306 in FIG. First, step 55
At step 01, a check is made to see if the cancel key=131 has been input. If there is an input here, category designation is not performed, so the category designation process ends in step 5502. If there is no input, proceed to the next step.

Next, it must be determined which word or group of words should be assigned a category.

Therefore, in step 5503, it is determined which word or word group is to be designated as a category. Once the word or word group has been determined, it is determined in step 8504 whether or not the confirmation key=130 has been input from the input unit 1. Here, if a confirmation key other than 130 is input, the process returns to step 5503 and again it is determined which word or word group is to be designated as a category. If the confirmation key=130 is input, the process advances to step 5505.

Next, in step 5505, cursor movement key=110
-113, it is checked whether or not the up and down keys have been input. This is because the category shown in FIG. 6 is selected by using the up and down keys of the cursor. If the operator moves the cursor up or down, the category changes accordingly. As a result, a category is selected (step 8506).

If the category currently under the cursor matches the operator's selection, proceed to the next step. In step 5507, it is determined whether the confirmation key 130 has been input. If you have the key skills here, specify the category to perform analysis based on the category where the cursor is currently located (step 5sos).
, return to step 8501. Note that how the category designation is used in the analysis process will be described in detail later.

If the key is not manually operated as described above, the process returns to step 5505 to select the category again.

Here, we will describe how the grammatical categories specified in step 8508 are used in the analysis process.

In the analysis control in step 5403 of FIG. 4,
Normally, all part-of-speech candidates are sent to syntactic analysis in step 8404. As a result, in step 8404, syntax analysis is performed sequentially based on the nine parts of speech candidates sent. However, when the grammatical category is specified in step 8508,
In step 5403, only part-of-speech candidates based on the category specification are sent. For this reason, syntax analysis is not performed for other categories, so the analysis process can be performed faster than in the past. Also, step 840
Even if all part-of-speech candidates are sent in step 3 as in the past, a range of words or word groups is specified in conjunction with category specification. Therefore, the number of combinations of part-of-speech candidates is naturally decreasing. This means that the vertigo in the original text has also been reduced. This allows analysis to be performed more quickly and accurately than before.

In the above processing, a grammatical category can be specified for any word or word group in the sentence,
Analysis processing can be performed based on this designation.

Next, Figure 7 (a) The value is conv.
erted int.

A case will be described in which the original sentence form of (u, v) is analyzed by the sentence analysis device provided by the present invention.

In step 5301 of FIG. 3, the original text shown in FIG. 7(a) is input. In step 5302, the original text is set in the input buffer as a character code.

In step 5303, the original text is displayed on the display section 2.

Then, in step 5304, it is determined whether the analysis instruction key: 100 has been input. If the category designation key: 120 is input here, the process advances to step 8306, and in step 5307, processing for category designation is performed. Note that step 8301, step 5304, step 8306, step 8308 . In step 5310, the process enters a normal input waiting state and performs processing corresponding to the input. In step 5307, a category is specified. The following explanation will be based on FIG.

If there is no input of the cancel key 131 from the input unit 1 in step 5501, then in step 5503 the word or word group to be designated as a category is determined. Here the seventh
When specifying a category for "(u, v)" in the original text in FIG. 7(a), the cursor is moved to the position shown in FIG. 7(b).

Next, when the cursor enlargement key=114 is pressed 4@, the cursor is enlarged to ")#" (Fig. 7(C)).Here, if the confirmation key=130 is input in step 5504,
The screen display changes as shown in FIG. 6(a). Then, in step 5505, cursor enlargement keys: 110 to 11
Using the up and down movement keys 3, select the desired category (step 8506). Currently @(u,V)
” ?: We want to analyze it as a noun phrase, so move the cursor to the noun phrase. Once moved, enter confirmation key: 130 from input section 1 (step 8507
).

Then, "(u, v)" is designated as a noun phrase category (step 5so8). Along with this, a signal is sent to the display control unit 7, as shown in FIG. 7(d).
(u, v)" is shaded or otherwise displayed to attract the operator's attention. Then, the process returns to step 5501. Here, for other words or word groups in the original text,
If there is no need to specify the category, manually press the cancel key 131 from the input section 1. Then, in step 5502, the category designation process ends.

On the other hand, if it is necessary to specify a category, the processes from step 8503 onwards are performed. After performing the above processing, the system returns to the state of waiting for the input of each key as shown in FIG. Then, when the analysis instruction key: 100 is input, step 83
If the determination is made in step 04, analysis processing in step 5305 is performed. This is done according to the flowchart in FIG. Morphological analysis is performed in step 5401, and step 840
2 to perform a dictionary search. Then, in step 5403, part-of-speech candidates are sent. However, step 8307 in FIG.
So, @ (u, v) # specifies the noun phrase category. For this reason, as in the past, “(u,
v)' is not treated as an infix, and step 5
At 404, parsing can be performed using parsing grammar 4C. Then, in step 5405, it is determined that the analysis process is completed, and the process returns to the state of waiting for each key input.

Since the above-mentioned analysis result was determined by the operator, if the analysis end key=140 is input, the human power of this key is determined in step 5310. Then, in step 5311, the analysis results of the original text are stored in the original text storage unit 2, and the analysis results storage unit stores the analysis results, which words or word groups have been specified as categories, and which categories have been specified for those words or word groups. 3, memorize each.

Next, here, Fig. 8(a)" When 1=100
, thesubroutine will 5to
A case will be described in which an original sentence, p, is analyzed by the sentence analysis device provided by the present invention.

At step 8301 in FIG. 3, the original text shown in FIG. 8(a) is input. In step 5302, the original text is stored in the input buffer as a character code.

In step 5303, the original text is displayed on the display section 2.

Then, in step 5304, it is determined whether the analysis instruction key: 100 has been input. If the category designation key=120 is input here, the process proceeds to step 8306, and in step 5307, processing for category designation is performed. Note that steps 5301, 5304, 5306, 19308, and 5310 are normally in a state of waiting for input, and perform processing corresponding to the input. In step 5307, the category is specified.

The following explanation will also be made with reference to FIG.

If the cancel key 131 is not input from the input unit 1 in step 8501, when category designation is to be performed in step 8503 for determining the word or word group for which category designation is to be performed, the cursor is moved to the position shown in FIG. 8(b). move it.

If the confirmation key 130 is input in step 5504, the screen display changes as shown in ta+ in FIG.

Then, in step 5505, the cursor movement key
: The specified category is selected using the up and down keys 110 to 113 (step 8506). Currently, we want to perform analysis processing using "1=100#" as a noun clause, so move the cursor to the noun clause (Figure 6 (b)
). Once moved, input confirmation key=130 from input unit 1 (step 8507). Then, “t=xoo
" is designated as a noun clause category (step 5sos). Along with this, a signal is sent to the display control unit 7, and the part "1=100" is shaded as shown in FIG. 8(d). Then, the process returns to step 5501.Here, if there is no need to specify categories for other words or word groups in the original text, press the cancel key = 131
The key is human power. Then, in step 5502, the category designation process ends. On the other hand, if it is necessary to specify a category, the processes from step 8503 onwards are performed. After performing the above processing, the system returns to the state of waiting for the input of each key as shown in FIG. And analysis instruction key: 100
If it is determined in step 5304 that there is an input, analysis processing in step 5305 is performed. This is done according to the flowchart in FIG. Morphological analysis is performed in step 8401, and dictionary search is performed in step 5402.

Then, in step 5403, part-of-speech candidates are sent.

However, in step 5307 of FIG. 3, '1=100''
is a noun clause. Therefore, unlike in the past, 11=100# cannot be treated as a sentence (noun clause), and step 8
At 404, parsing can be performed using parsing grammar 4C.

Then, in step 8405, it is determined that the analysis process is completed, and the process returns to the state of waiting for each key input. Since the above-mentioned analysis result is what the operator desires, if the analysis end key: 140 is input, the human power of this key is determined in step 5310.

Then, in step 5311, the original text is stored in the original text storage unit 2.
Then, the analysis results, which words or word groups in the sentence were designated as categories, and which grammatical categories were designated for those words or word groups are stored in the analysis result storage unit 3, respectively.

As described above, one operator specifies the category of the original text on the screen and specifies a word or word group. Categories can be easily specified by simply selecting the grammatical category on the screen. Therefore,
Specifying a category does not increase the burden on the operator.

Let's consider a case where this is done. This is a case where there are words with many part-of-speech candidates in the original text, which causes ambiguity. The original text is generally said to have three syntactic ambiguities. Three of them are shown below.

(1) When “1 ike” and below are treated as adverbial phrases.

(2) When treating “1ike” as an adjective.

(3) When “me ke” is treated as a verb.

Of these three, the original text should be analyzed according to (1). Therefore, “1 ike (ana
rrow” is a group of words to be specified as a category, and this is used as an adverb phrase to specify a category. Then, analysis processing is performed based on this category specification. In this way, ambiguity can be resolved, and it is possible to Analysis processing can be performed more quickly and accurately.

In addition, the original F! A is not limited to the embodiments described above. For example, in the above embodiment, the analysis result storage unit 3
It is conceivable that the word or word group for which the category has been specified and the category specified for the word or word group stored in the .

This is because when the analysis section 5 performs analysis processing, it is searched to see if the word or word group for which the category is specified has already been stored in the analysis result storage section 3. As a result of this search, if the word or word group for which the category is specified has already been stored, this analysis result is used. Further, if the analysis result is not stored, the analysis result is stored in the analysis result storage section 3. As a result, when a similar word or a group of strange words appears in the original text, the analysis process can be performed quickly and accurately without the need for the operator to specify a category. Furthermore, in the above-mentioned embodiment, the grammatical categories shown in Figure 6 are listed as candidates for category designation, but in addition to these, adjective phrases, adjective clauses, etc. can also be easily used as candidates for category designation. can. Further, in the above-described embodiment, on the screen for selecting a specified category from the grammatical category candidates shown in FIG. 6, when specifying a category for the first time, the cursor is placed on any displayed category. However, after specifying a category once, you will have to specify the category again, so when selecting a category using the screen shown in Figure 6, you can also place the cursor on the previously selected category. . This allows the operator to perform fewer key operations when specifying the same category for other words or word groups in the original text.

Further, although the keyboard is used when the operator inputs from the input unit 1 which word or group of words in the original text he wishes to specify the category, he may use various instruction means such as a mouse. Although the original text and the analysis results are stored in separate storage units, they can also be stored in a single storage unit in association with each other. Although the analysis unit 4 stores three dictionaries, it is also possible to store them all in one.

Furthermore, it can be implemented not only by a sentence analysis device, but also by being incorporated as an analysis part of a machine translation system, a context analysis system, or the like.

Hereinafter, a machine translation device incorporating the sentence analysis device provided by the present invention as a part of the translation section will be explained.

FIG. 9 shows a block diagram of one embodiment of the present invention. The following is an explanation using English-Japanese translation as an example. In FIG. 9, an input unit 901 is used to input the original text to be translated (in this case, English) and to specify grammatical categories for various commands and arbitrary data in the first language. A keyboard as shown in Figure 2 is used.

A CRT display, for example, is used for the display section 902 to display English sentences input through the input section 901 and Japanese sentences translated by the translation section 905. In addition, grammatical categories are also displayed. The document recording unit 903 can independently record the data of the first language (English) and the second language (Japanese), as well as record the English data and the Japanese data as a result of its translation in association with each other. It is possible to do so. In addition, it also records which words or word groups in the English sentence in the first language were grammatically categorized, and which grammatical categories were designated for those words or word groups. The dictionary recording unit 904 stores, as knowledge information necessary for translation processing,
It records a language dictionary that records the correspondence between words and phrases in English, the first language, and translated words in Japanese, the second language, as well as headings, parts of speech, conjugations, concepts, grammatical functions, etc. This also includes records written in both English and Japanese. This includes the grammar used when translating mathematical formulas of data in the first language into data in the second language. The translation unit 905 translates English text, which is first language data, into Japanese text, which is second language data.

This is a conventional translation unit K with an added function of performing translation processing based on a grammatical category for an arbitrary word or word group of an English sentence input from the input unit 901. The translation control unit 906 controls the entire device.

FIG. 10 is a diagram showing an example of the screen layout of the display section 902. The input original text is displayed in the original text display area 10 on the left side of the screen.
01 is displayed. The translated text obtained as a result of the translation process is displayed in a translated text display area 1002 on the right side of the screen. Further, an editing area 1003 at the top of the screen can be used to display information necessary for various editing, and grammatical categories are also displayed here.

FIG. 11 is a flowchart for explaining the operation of the translation control unit 906. The translation control unit 906 specifies a document to be translated and starts its operation when a command to start translation is input from the input unit 901 . The translation control unit 906 that has started operating first reads the first language data (data A) of the designated document (step 81101).

Step 8110 to extract the first sentence of the data person
2) Transfer to the translation unit 905 and obtain a translation result while referring to the dictionary unit 904 (step 81103).

Repeat the above process (step 81103) until the last sentence of the data person is reached.Step 81104゜8110
5) When the translation has been completed up to the last sentence of data A, data A and its translation result are recorded in the document recording unit 903 (step S1106), and the process is ended.

FIG. 12 is a 70-chart illustrating step 81103 in FIG. 11, that is, the processing flow of the translation unit 905. The translation unit 905 starts its operation when data to be translated is transferred from the translation control unit 906.

Since this embodiment uses a transfluid method, analysis processing (step 81201) and conversion processing (step 3120) are performed.
2) to create the data necessary to generate the translation. Then, a generation process is performed (step 8120
3) Obtaining a Translated Text Next, referring to FIG. 13, translation processing in a machine translation device incorporating the sentence analysis device provided by the present invention as a part of the translation section will be explained. First, as shown in FIG. 13(a), "When 1=lQQ, thesubrou" is input from the input section 901 into the original text display area 1001.
Translate the sentence ``'' on page 15 of tine w.

When translating an English sentence containing such a mathematical formula, as mentioned above, as it is, "1 = 100"
It is extremely difficult to analyze as a sentence (noun clause). This means that translation processing cannot be performed. For this reason, it is determined which word or word group in the original text is to be assigned a grammatical category by processing using the sentence analysis device described above (FIG. 13(b)). And the 13th
As shown in Figure (b), it is decided which grammatical category to specify for "1=1oo". Here, a noun clause is designated as the grammatical category IJ-. Then, analysis processing, which is part of translation processing, can be performed on the English sentence written in the first language (step 81201).

Since the present embodiment uses a transfluid method, the conversion process (step 81202) and the generation process (step 81202)
203). This is done using the grammar used when performing translation processing on the mathematical expressions held in the dictionary section 904. This grammar is realized as a correspondence table between heading information 1401 and meaning/usage information 1402, as shown in FIG. 14, and is stored in an IC card, read-only memory (ROM), or magnetic disk device. There is.

In FIG. 14, the heading information 1401 is shown as symbol type data used in mathematical formulas, but it is not limited to these symbols, and data of all symbol types used in mathematical formulas are held.

Furthermore, although the meaning/usage information 1402 is written in Japanese in the device of this embodiment, it can easily be written in other languages and can be changed depending on the usage. As a result, a translated text as shown in FIG. 13(d) is obtained.

As mentioned above, if a machine translation device incorporating the sentence analysis device provided by the present invention as part of the translation section is used,
Translation processing can be performed more quickly and accurately than before. This is because the analysis process, which is one process of the translation process, allows the operator to easily specify a grammatical category for any data in the first language. This is because analysis processing can be performed more quickly and accurately than before, and subsequent processing can be performed based on this. In particular, when translating an original text containing a mathematical formula, the translation process is performed based on the grammar used when translating the mathematical formula held in the dictionary section 904. As a result, mathematical expressions can be translated not only in their original form, but also in their natural language form, resulting in an easy-to-understand translation.

Note that the operator can select whether or not to use the grammar shown in FIG. 14. This is because it is easier to understand long mathematical expressions as they are. This is because if a long mathematical formula is expressed in natural language, it becomes even longer and, conversely, difficult to understand. Also,
Although the grammatical categories are displayed in the editing area 1003, they are not limited to this, and can easily be displayed in a window at any position on the screen.

In short, the present invention is not limited to the embodiments described above, and can be implemented with various changes without departing from the spirit thereof.

〔Effect of the invention〕

As described above, according to the present invention, in the sentence analysis device, a word or word group in the input original text for which a grammatical category is to be specified is specified, and a grammatical category is selected for these words or word groups. The operator can easily specify the grammatical category by simply Since the analysis process is performed based on the designation of this grammatical category, the analysis process can be performed more quickly and accurately than in the past.

Furthermore, in a machine translation device incorporating the sentence analysis device provided by the present invention as a part of the translation section, the input first
When translation processing is performed on language data, the translation processing can be performed more quickly and accurately than in the past.

[Brief explanation of drawings]

FIGS. 1 and 9 are 70-charts for explaining the operation of the control unit 8 used by the overall professional in one embodiment of the present invention, FIG. 4 is a flowchart showing details of the analysis process in FIG. 3, and FIG. The figure is a flowchart of grammatical category specification in Figure 3, Figure 6 is a diagram showing an example of the screen display when selecting a grammatical category in an embodiment of the present invention, and Figures 7 and 8 are the FIG. 10 is a diagram showing an example of the layout of the display section shown in FIG. 9; FIG. The figure is a flowchart for explaining the processing of the translation control unit used in one embodiment of the present invention, FIG. 12 is a flowchart showing details of the translation processing in FIG. 11, and FIG. This is a conceptual diagram of the grammar. 1... Input section 2... Original text storage section 3... Analysis result storage section 4... Analysis dictionary section... Analysis section... Display section 01... Input section 02... Display section 04...Dictionary Department 05...Translation Department

Claims (7)

[Claims] (1) An input section for inputting words and sentences expressed in a natural language to be analyzed; an analysis dictionary section that stores knowledge used in analysis processing; and an input section for inputting the original text input from the input section into the analysis dictionary. an analysis section that performs an analysis process using the input section; the input section is further for inputting a grammatical category for an arbitrary word or word group in the original text; A sentence analysis device characterized by performing analysis processing based on specified grammatical categories. (2) The sentence analysis device according to claim 1, wherein any word or word group in the original text is a metalinguistic expression. (3) The sentence analysis device according to claim 1, wherein any word or word group in the original text has multiple parts of speech. (4) The sentence analysis device according to claim 1, wherein the grammatical category is a noun phrase, a noun clause, a verb phrase, an adverb phrase, an adverb clause, an adjective phrase, an adjective clause, or the like. (5) If the same word or word group for which the grammatical category has been specified exists elsewhere in the original text, the analysis unit shall assign the grammatical category of the previously specified word or word group to the same word or word group in the original text. 2. The sentence analysis device according to claim 1, wherein the sentence analysis device specifies a sentence. (6) an input section for inputting data in a first language; a dictionary section for recording dictionary data necessary for translation; The input unit further includes a translation unit for translating data in the second language, and a display unit for displaying the data in the second language obtained by the translation unit, and the input unit further includes a grammatical translation unit for any data in the first language. 1. A machine translation device for inputting a category, wherein the translation section performs translation processing based on the grammatical category specified by the input section. (7) The dictionary section further includes the first dictionary inputted from the input section.
It holds a grammar used when performing a translation process on a mathematical expression of data in a language, and the translation unit refers to the dictionary unit to translate the mathematical expression of data in a first language input from the input unit. A machine translation device according to claim 6.