JP2880600B2 - Word string translator - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an automatic translator,
In particular, the present invention relates to an apparatus for translating a word string that can be represented in a general form, such as a idiom or a form of a number + unit.

[0002]

2. Description of the Related Art Conventionally, "as soon as possible" and "as
Idioms such as "fast as possible" or "20 yards"
Numerals with units such as, or dates (eg, Aug.
5, 1991), the expression is once decomposed into smaller units by parsing and then translated by interpreting the meaning again. One per dictionary or idiom dictionary
When a character string included in an input sentence completely matches the word or the heading of the idiom dictionary, the translation process is performed by replacing the word with another corresponding language.

[0003]

However, when each specific idiom is registered in a word dictionary or a dedicated idiom dictionary as described above, it is necessary to register all occurrence patterns of the idiom in some cases. Alternatively, it is necessary to register even a word, such as a numeral, an era, a date, etc. written with a unit, which does not make a complete meaning only with each word constituting the character string, in the dictionary. Also, when creating a idiom dictionary to provide a translation device, the result of the conversion by the idiom dictionary may be affected by the result of the syntactic analysis performed in the preceding stage. It was difficult to prepare in such a way that it could be understood.

As a result of these problems, the management efficiency is reduced, for example, the capacity of the dictionary is unnecessarily increased, or a required idiom is forgotten to be registered, and the dictionary search efficiency at the time of translation is reduced. There was a problem.

SUMMARY OF THE INVENTION It is therefore an object of the present invention to provide a word string translator capable of efficiently translating a word string with a small dictionary capacity.

[0006]

SUMMARY OF THE INVENTION A word string translator according to the present invention includes a first language word string and a second language word string selected to represent a similar meaning to the first language word string. A first dictionary storing a large number of pairs consisting of word strings of languages , and
And words in the first language with similar meanings to words in the first language
To a word in a second language pre-selected to represent
A second dictionary for storing a number of associations is included. Each word string in the first language includes a word position mark that is predetermined so that a matching relationship is established with a plurality of words in the words in the first language. Each word string in the second language includes a word replacement position mark inserted at a position corresponding to the word position mark included in the word string in the first language. Word string translation apparatus according to the present invention further, the sentence of the first language to be input,
By referencing the first dictionary and the second dictionary,
Match detection means for detecting whether there is an input word string that matches one of the word strings in the first language, and matching with any one of the word strings in the first language by the match detection means Is extracted from the second dictionary among words in the input word string in which the word position mark has been matched with the word position mark, and a second word paired with the word string in the first language is extracted from the second dictionary. And a word string replacing means for converting the input word string in the first language into a word string in the second language by replacing the word string with the word replacement position mark of the word string in the second language.

[0007]

In the word string translating apparatus according to the present invention, the first sentence in the first language is included in the first sentence in the first dictionary.
It is detected whether there is an input word string that matches one of the word strings in the language of the first language, and if so, it is detected that there is a match with the input word string. 1
The word in the second language corresponding to the word in the input word string that has a matching relationship with the word position mark included in the word string in the second language is extracted from the second dictionary. The extracted word in the second language is replaced with a word replacement position mark included in the word string in the second language paired with the word string in the first language for which a match is detected, and A word sequence is created. In this apparatus, only a word indicated by a word position mark is different, and other words are composed of common words. A plurality of word strings in the first language are stored in the first dictionary as one general form. Is stored. It is not necessary to register all individual word strings in advance.

[0008]

FIG. 1 is a block diagram of a system using a computer functioning as an automatic translation apparatus including a word string translation apparatus according to the present invention. With reference to FIG.
In order to perform a process of translating an input original sentence in a first language (English in this embodiment) into a text in a second language (Japanese in this embodiment) according to a predetermined translation process. CPU 10 and bus 20 connected to CPU 10
Connected to the CPU 10 via the bus 20 and the CPU 1
0, a translation program memory 12 for storing a translation program to be executed, and a CPU 20 connected to the CPU 10 via a bus 20. A user inputs a command or data to proceed with the process by interacting with the CPU 10. , A CRT 14 for displaying a processing result, and a language analysis rule, which is connected to the CPU 10 via the bus 20 and is referred to during translation processing executed by the CPU 10, Language knowledge, conversion rules,
A database 18 in which generation rules and the like are stored in advance.
And a buffer A22 connected to the bus 20 for storing the input text and a buffer B for storing information about each word contained in the text stored in the buffer A22.
24 and a buffer C2 for storing the Japanese meaning of a predetermined word among the words included in the buffer B24.
6, a word dictionary 28 for storing a number of pairs of English words and Japanese words corresponding to the English words, and a idiom or a number + unit connected to the bus 20 and described in a general format And a phrase dictionary 30 for storing such expressions in association with corresponding Japanese expressions. These word dictionaries 28
And the idiom dictionary 30 are stored in the memory of the computer system.
Is stored.

In the word dictionary 28, a number of English words are stored in association with their parts of speech and Japanese words preselected as translations of the English words. Word dictionary 28
In addition, a idiom in a specific format may be stored instead of a idiom in a general format described later.

Referring to FIG. 2, buffer B is provided with an area for storing each word included in the input English original sentence together with the part of speech of each word, to which respective word positions are assigned. . Buffer B has a corresponding pointer Q
Is provided, and the word (headword) stored at the word position corresponding to the value of the pointer Q is to be processed.

Referring to FIG. 3, a phrase dictionary 30 which is a feature of the word string translating apparatus according to the present invention has a plurality of items (item 1, item 2,.
…) Is stored. Each item is divided into a word sequence representing an expression such as a idiom or a numeral and a unit (hereinafter simply referred to as an idiom) described in a general form, and a word sequence after the word sequence by a heading delimiter. And a corresponding Japanese word string (character string) to be stored. Each word of the word sequence is associated with the part of speech (adverb, adjective, noun, numeral, ...) of the word. This idiom dictionary 30 differs from the conventional idiom dictionary or word dictionary only in that each item does not represent a specific individual idiom, but only those parts of the idiom having a common general form that differ from each other. Is replaced with a predetermined word position mark (“***”), thereby storing a plurality of idioms as one general form. In the word position mark “***”, the part of speech of the word to be replaced with the word position mark is stored in association with other words as well. Also, in a character string when the general form of the idiom represented by this word sequence is translated into Japanese, a word replacement position mark “***” inserted at a position corresponding to this word position mark is also included. Included in.

The storage format of each item is not limited to idioms, but is the same for expressions such as "numerals + units" (expressions such as *** yards).

As shown in FIG. 3, the idiom dictionary 30 is provided with a corresponding pointer R. Pointer R
Can be positioned at the head of each item in the case of the present embodiment, and within each item, the pointer R indicates each headword, heading delimiter, corresponding Japanese, etc. in increments of 0 to 1 It will be pointed out sequentially.

Referring to FIG. 4, when a character string that matches a character string included in the idiom dictionary excluding the word position mark is found in the input original text, The word at the position of the word position mark, the part of speech of the word, and the Japanese word assigned to the word,
This is for storing with reference to the dictionary 28.

The operation of the automatic translation apparatus including the word string translation apparatus according to one embodiment of the present invention shown in FIGS. 1 to 4 will be described below. FIG. 5 shows the entire processing flow of the automatic translation device. Referring to FIG. 5, when an English sentence to be processed is input, morphological analysis is performed on the original sentence while referring to word dictionary 28 in step S (hereinafter simply referred to as S) 001. Subsequently, in S002, word information such as information on the part of speech is fetched with reference to the word dictionary 28 for each word subjected to morphological analysis.

In S003, a idiom dictionary lookup process for converting idioms contained in the input original sentence into corresponding Japanese phrases is performed with reference to the idiom dictionary 30 having the structure according to the present invention. .

In S004, a syntax analysis process is performed in accordance with the word dictionary 28 and analysis rules based on English grammar knowledge stored in the database 18, and the input original text is converted into an internal structure such as a tree structure. Processing represented by the format is performed.

In S005 and subsequent steps, the meaning of the original sentence represented by the tree structure is converted into a Japanese sentence in accordance with the grammar knowledge, conversion rules, generation rules based on the Japanese grammar knowledge included in the database 18, and the like. Is performed. First, in S005, a process of analyzing the meaning of the internal structure represented as a tree structure using the analysis rules and the grammar knowledge of the database 18 is performed. In S006, according to the grammar knowledge and conversion rules stored in the database 18, S004 and S0
05 is converted into a tree representing a predetermined syntax according to Japanese conversion rules,
A process of assigning an appropriate translated word to each word is performed. Finally, in S007, a process of converting the converted tree structure obtained by the process of S006 into a Japanese sentence in accordance with a generation rule based on Japanese grammar knowledge included in the database 18 is performed.

FIG. 6 is a flowchart of the process performed by the word string translating apparatus according to the present invention, which is performed in S003 of FIG. First, in S310, processing for storing the input English original text in the buffer A22 is performed. Subsequently, in S320, a process of creating the contents of the buffer B24 based on the English original text stored in the buffer A22 is performed. This processing will be described later with reference to FIG.

In step S330, it is determined whether or not the data stored in the buffer B includes a word which can be one specific expression of the idiom expressed in the general form stored in the idiom dictionary 30. Is checked, and if there is, a process of moving the processing target position to the top is performed. This process corresponds to a preparatory process of the main process of the idiom dictionary lookup process performed in S340 described later. This processing will be described later with reference to FIG.

Further, at S340, the word string stored in the buffer B24 is actually stored in the idiom dictionary 30.
It is determined whether or not the idiom matches the idiom expressed in the general form, and if the idiom matches, a process of converting to a Japanese word string (character string) is performed according to a certain procedure. . This processing will be described later in detail with reference to FIG.

Referring to FIG. 7, the buffer B creation processing performed in S320 is executed as follows. First S32
At 1, the value of the pointer P indicating the word position of the buffer A22 is initialized to 0. Subsequently, in S322, the pointer indicated by the current value of the pointer P,
A process of storing the words in the buffer A22 in the buffer B is performed. Subsequently, in S323, it is determined whether or not the next word exists in the buffer A22. If the answer to the determination is YES, the control proceeds to S234, otherwise the control proceeds to S325.

In S324, a process of incrementing the pointer P of the buffer A22 by one is performed. As a result, the position of the word pointed to by the pointer P in the buffer A22 moves one position backward. The post-processing of S324 returns to S322. By repeating the processing of S322 to S324, the word strings in the buffer A22 are sequentially stored at the word positions in the buffer B as shown in FIG.

The buffer A2 as a result of the determination in S323
If it is determined that the next word does not exist in 2, that is, if it is determined that the processing has reached the end of the word string stored in the buffer A22, the processing proceeds to S325. In S325, the part-of-speech information of each headword stored in the buffer B is stored in the field of the part-of-speech information corresponding to each headword in the buffer B with reference to the result of the dictionary lookup performed in S002 of FIG. , Buffer B is completed. When the processing in S325 ends, the buffer B creation processing ends.

Referring to FIG. 8, the idiom dictionary lookup preparation process performed in S330 of FIG. 6 is executed as follows. First, in S331, the contents of the pointers Q and R indicating the word position in the buffer B and the headword position in each item of the idiom dictionary 30 as shown in FIGS. You. Subsequently, in S332, a process of comparing the headword in the buffer B indicated by the pointer Q with the headword indicated by the pointer R in a predetermined item in the idiom dictionary 30 is performed. In S333, as a result of the comparison process in S332, the headword of the buffer B pointed to by the pointer Q and
A determination is made as to whether the entry word in the idiom dictionary 30 indicated by the pointer R matches. The answer is NO
If so, the control proceeds to S334, a process of incrementing the pointer Q by 1 is performed, and the control proceeds to S335.
In S335, it is determined whether or not the headword stored in the buffer B at the word position indicated by the incremented pointer Q is an end symbol. The terminating symbol is predetermined as a symbol indicating the end of a sentence. In the case of English, three terminators are defined as a terminator: a period (.), A question mark (?), And an exclamation mark (!). When it is determined in S335 that the headword indicated by the pointer Q is a terminator, the headword reaches the final position of the word string stored in the buffer B and is stored in the idiom dictionary 30. This means that no word corresponding to the heading of each item was found, so that there was no expression corresponding to the idiom stored in the idiom dictionary 30 as a general form in the original sentence. Thus, the process of searching the idiom dictionary ends. If it is not a terminator, the process returns to step S332 again, and the processes of S332 to S335 are repeated until the headword indicated by the pointer Q matches the headword indicated by the pointer R.
Alternatively, the processing is performed until the processing target reaches the end of the headword in the buffer B. If the headword indicated by the pointer Q matches the headword indicated by the pointer R, the idiom dictionary lookup preparation processing ends, and the next idiom dictionary lookup processing is performed.

Referring to FIG. 9, the idiom dictionary lookup process performed in S340 (see FIG. 6) is executed as follows.
First, in S341, a process of incrementing the content of each of the pointers Q and R used in the process described with reference to FIG. 8 by one is performed. Subsequently, in S34, it is determined whether or not the headword in the buffer B pointed to by the incremented result pointer Q is an end symbol. If it is a terminator, it means that there is no idiom in the original sentence, as described above, so the idiom dictionary lookup processing ends. If not, the process proceeds to S343. In S343,
It is determined whether or not the headword of a predetermined item in the idiom dictionary 30 indicated by the pointer R is a headword separator. If it is not a headword delimiter, it is necessary to further compare the word string stored in the buffer B with each headword string of an item in the idiom dictionary 30 that is currently being determined. Goes to S344. If it is a headword separator, the process proceeds to S34.
The process proceeds to 9 or less, and the details will be described later.

In S344, a process is performed for comparing the headword in the buffer B24 indicated by the pointer Q with the headword of the item currently being processed in the idiom dictionary 30 indicated by the pointer R. Is performed. If the comparison results in a match, the following headword must be compared for this item, so the processing is S
341 is returned. If not, the process proceeds to S345.

When the process proceeds to S345, the headword indicated by the pointer R is a word position mark (***). In this case, since the word string stored in the buffer B is one example of the idiom in the general expression form stored in the idiom dictionary 30, the part of speech of the word position mark (***) and , The part of speech of the corresponding headword in the buffer B needs to match.
Therefore, in S345, it is determined whether or not the part of speech of the headword in the buffer B indicated by the pointer Q matches the part of speech of the headword in the idiom dictionary 30 indicated by the pointer R. If they match, the word string stored in the buffer B as a result of the processing so far may be a specific expression of an item in the idiom dictionary 30 that is currently being processed. Therefore, the process proceeds to S346, and in S346, dictionary entry processing of the headword in the buffer B pointed to by the pointer Q is performed, and this headword, its part of speech, and the Japanese language corresponding to the headword are processed. The words are stored in the buffer C26 shown in FIG. After S346, the word position mark (**
Since it is necessary to further determine whether or not the word following *) matches, the process returns to S341.

If it is determined in step S345 that the part of speech of the headword indicated by the pointer Q and the pointer R does not match, the process proceeds to step S347. In S347, since it is determined that the word string stored in the buffer B is not a specific expression of the item in the idiom dictionary 30 currently being processed, a process for shifting the process to the next item is performed. That is, in S347, it is determined whether or not the next item exists in the idiom dictionary. If the next item does not exist as a result of the determination, it means that the process has reached the last item of the idiom dictionary 30, and this idiom dictionary lookup processing ends. If it is determined that there is a next item, the process proceeds to S348, in which the pointer R is moved to the next item in the idiom dictionary, the pointer R is initialized to 0, and the pointer is moved to the first position in the item. Move R. The post-processing of S348 returns to S341.

It is assumed that the processing of S341 to S346 is repeatedly executed, and as a result, the word string stored in the buffer B and the headword string of a certain item in the idiom dictionary 30 are changed to a word position mark (***). And the part of speech of the word position mark (***) matches the part of speech in the buffer B at the position corresponding to the word position mark, the word string in the buffer B is Is determined as a specific expression of the item (that is, an expression in which the word position mark (***) is replaced with the corresponding word in the buffer B).
Therefore, if the comparison has proceeded to the headword separator in S343 and all the headwords except the word position mark match the headword string of the item, S34.
The process proceeds to 9, and the final process is performed.

At S349, the Japanese expression existing after the headword delimiter and the contents of the word stored in buffer C are merged, and as a result, the English word stored in buffer B is merged. A Japanese word string corresponding to the string (idiom) is obtained as a final result. S349
After that, among the headword strings stored in the buffer B, it is necessary to determine whether or not there is a further idiom for the succeeding headword string, so the process proceeds to S348.
Further, the processing of S341 and subsequent steps is repeatedly performed.

The above-mentioned processing will be described below in more detail with reference to the examples shown in FIGS. In buffer B,
It is assumed that “Come as soon as possible.” Is stored. It is assumed that the pointer R is currently pointing to the item 1 shown in FIG. Item 1 in FIG.
+ Adverb + as + possible ”, which is a general form of the idiom. As is apparent from a comparison between the two,“ as soon as p ”stored in buffer B
The expression “ossible” is obtained by replacing “***” in item 1 of FIG. 3 with “soon.” In this case, the idiom dictionary lookup processing is performed as follows.

First, the word "Come" in the buffer B is compared with the first entry word "as" of item 1. In this case, they do not match. So in this case,
The process of S334 in FIG. 8 is performed, and the pointer Q moves to the next “as”.

The comparison made after incrementing the pointer Q is between "as" and "as", which are clearly coincident. Therefore, the result of the determination in S333 in FIG.
Move to 0.

The processing after S341 is as follows. As a result of the pointers Q and R being respectively incremented, the next comparison is made between "soon" and "***". In this case, since the headword “***” on the idiom dictionary side is a word position mark, the two do not originally match.
However, in this case, it is determined in S345 whether or not the part of speech of the word position mark matches the part of speech of the headword to be compared in the buffer B. If they match, the words are processed as if they match. In this case, since the part of speech is an adverb in both cases, the answer to the determination in S345 is “match”
And the process proceeds to S346. And S
At 346, processing is performed on the headword of the buffer B indicated by the pointer 9, that is, the word "soon", and the buffer C as shown in FIG. 4 is created.

Hereinafter, by repeating the processing of S341 to S344 twice, it is possible that the "as" and "possible" of the headword string in the buffer B match the two following words of item 1 in FIG. It is confirmed. And the pointer R
Becomes 4 when the result of the determination in S343 is YE
In S, the process of S349 is performed. That is, S3
49, the translated word "soon" of the word "soon" stored in the buffer C (see FIG. 4) is added to the "***" portion of the word "as much as possible ***" assigned as the translation of item 1. Is substituted (substituted), and a character string "as soon as possible" is generated.

Then, as a result of the pointer Q being incremented in the subsequent processing of S341, the answer to the judgment in S342 is YE
The result is S, and the idiom dictionary lookup processing ends.

In the above description, "as adverb as possib
A idiom such as "le" is translated as "as *** as possible"
This is a method of storing in the idiom dictionary in a general expression form. However, the storage system of such a idiom dictionary and the conversion formula from English to Japanese are not limited to only idioms.
For example, if registered in a format such as 0 *** number 1 yards noun 2 entry word delimiter 4 *** yard, it is also possible to process expressions such as numbers + units.

As described above, according to the word string translator according to the present invention, not the idioms or the specific expressions such as numerical values and units but the general expressions thereof are registered in the idiom dictionary. By itself, many concrete idioms and expressions such as numerical value + unit can be translated from English to Japanese. This makes it possible to reduce the size of the dictionary,
In addition, since it is not necessary to register a large number of idioms, the amount of work for forgetting to register or performing maintenance is reduced, and management efficiency is improved. Further, since the number of items stored in the dictionary can be reduced, the search efficiency when referring to the idiom dictionary is improved.

[0040]

As described above, according to the present invention, the first language word string stored in the first dictionary is not an individual concrete expression, but only a different part. May be stored as a general expression format represented by. At the time of translation, the word replacement position mark in the corresponding word string of the second language is changed to the second word corresponding to the word whose match is detected with the word position mark in the corresponding word string of the first language.
By replacing words with words in the language, specific idioms and other special expressions can be translated. The number of pairs of word strings stored in the first dictionary is much smaller than in the case where individual concrete idioms are registered. Therefore, the capacity of the first dictionary can be reduced, and since the number of registrations is small, the management efficiency of the first dictionary is improved, and the search efficiency thereof is also improved.

[Brief description of the drawings]

FIG. 1 is a block diagram of a system for realizing an automatic translation device including a word string translation device according to the present invention.

FIG. 2 is a schematic diagram illustrating a configuration of a buffer B;

FIG. 3 is a schematic diagram illustrating a configuration of a idiom dictionary 30;

FIG. 4 is a schematic diagram showing the contents of a buffer C;

FIG. 5 is a flowchart illustrating an entire flow of an automatic translation process.

FIG. 6 is a schematic flowchart of the entire idiom dictionary lookup process;

FIG. 7 is a flowchart of a buffer B creation process.

FIG. 8 is a flowchart of a idiom dictionary lookup preparation process.

FIG. 9 is a flowchart illustrating a main process of a idiom dictionary lookup process;

[Explanation of symbols]

 10 CPU 12 Translation Program Memory 14 CRT 16 Keyboard 18 Database 22 Buffer A 24 Buffer B 26 Buffer C 28 Word Dictionary 30 Idiom Dictionary

Claims (1)

(57) [Claims] 1. A word sequence in a first language and selected as having a similar meaning to the word sequence in the first language,
A first dictionary for storing a large number of pairs of word strings in a second language , and a word in the first language,
Pre-selected as meaning similar to a word in another language
Stored in association with the words of the second language
And a memory for storing a second dictionary for each of the first languages, wherein each of the first language word strings is predetermined so that a matching relationship is established with a plurality of words among the first language words. Each of the word strings in the second language includes a word replacement position mark inserted at a position corresponding to the word position mark included in the word string in the first language. , the sentence of the first language to be input, the first
By referring to the second dictionary and the second dictionary,
A match detecting means for detecting whether or not there is an input word string matching one of the word strings in the first language; and any one of the word strings in the first language by the match detecting means. A word in the second language corresponding to a word having a matching relationship with the word position mark is extracted from the second dictionary, and the word in the first language is extracted from the input word string detected to match. A word string for converting the input word string in the first language to the word string in the second language by replacing the word replacement position mark of the word string in the second language with a string A word string translation device including a conversion unit .