JPH0887506A - Pattern collation type translation processor - Google Patents

Abstract

PURPOSE: To provide a pattern collation type translation processor which can increase the freedom for production of the bilingual patterns, increase the application range of a single bilingual pattern, suppress the wrong applications of bilingual patterns, reduce the labor for production of the bilingual patterns, and also can improve the accuracy of translation processing results for the sentences of similar structures. CONSTITUTION: A pattern translation control part 3 includes a pattern collation means 8 which can perform the collation of both routine patterns when the routine parts of these patterns are coincident with each other even if the part corresponding to the free part of the routine pattern is not included in an input sentence, a syntactic category limitation means 14 which can limit a subject syntactic category of the free part, a syntactic category discrimination means 9 which discriminates the syntactic category of an input subject, a suffix word designation means 10 which designates the type of the suffix word positioned at the end of a subject of the free part of the routine pattern, a preposition designation means 11 which designates the types of a preposition and a preposition phrase or a conjunction positioned at the head of the free part of a collation pattern, a semantic category limitation means 12 which limits the semantic category of the subject core word of the free part of the routine pattern, and a semantic category collation means 13 which collates the semantic categories with each other for the input subject.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pattern matching combined type translation processing device, and more particularly to a pattern matching combined type translation processing device for converting a Japanese input sentence text into a translated output sentence of a target language.

[0002]

2. Description of the Related Art Generally, a translation processing device is composed of grammatical rules for analysis and conversion and word dictionary data of a source language and a target language, and is written in advance with a grammatical expression assumed at the time of creating each data. Good translation results can be provided for sentences with simple syntactic structure, but it is difficult to parse the source language sentence correctly for sentences with complicated syntactic structure or special phrases. Met.
For this reason, conventionally, special phrase patterns that frequently appear are registered in advance in a bilingual format, and by comparing the entered source language sentence with the registered bilingual pattern, analysis of complicated syntactic structure is avoided. Meanwhile, a translation processing apparatus has been proposed which also uses pattern matching to obtain an appropriate translation processing result. Japanese Patent Laid-Open No. 3-11 related to this type of translation processing device
The "machine translation system having a sentence pattern dictionary" of 6272 is proposed.

[0003]

In the conventional pattern matching combined type translation processing device, a portion to which an arbitrary phrase can be applied is set for each pair of a fixed pattern and a translation pattern registered as bilingual pattern data. There is. For example, if you want to translate the Japanese sentence "Refer to chapter 5 for the details." Into "Refer to chapter 5 for the details."
Please refer to "→" Refer to #A for the details. "
By registering as
Even if the "chapter" portion is any other phrase, it can be translated by the same bilingual pattern data.
However, in the conventional method, the input sentence that can be collated with this parallel translation pattern data is an input sentence that consists of only three components, "For details,""A (arbitrary phrase)" and "Please refer to." Section that does not apply to any of the above components, such as "For the details, please refer to Chapter 5 as it is omitted in this chapter". Therefore, the input sentence with "" added cannot be matched with the parallel translation pattern. The symbols A and #A in the bilingual pattern represent a free portion to which an arbitrary phrase can be applied in the standard pattern and the translated pattern, respectively.

In addition, the Japanese sentence "I think he will come"
If you want to translate "I'm sure he comes." Into "I'm sure he comes.", The translation pattern data is "I think A will come" because of the restriction that the free part of the fixed pattern must be a phrase.
→ If you register like "I'm sure #A comes.", There is no extension to other verbs for the verb "come", and the translation "come"
It is difficult to register such a bilingual pattern for s "because the subject does not match unless the subject is the third person singular. Furthermore, even if it becomes possible to use a clause as a free part, the bilingual pattern data I think A "→"I'm sur
It is not preferable that the phrase is matched with the free part A of the fixed pattern of "e #A.".

Further, an example of a parallel translation pattern "I think A"
→ If you want to make the free part A in "I'm sure #A." Optional, the rest of the standard pattern will be "probably" and "think," but the case particle attached to the section of free part A When there is no A part, it is necessary to delete "to" along with it. Therefore, it must be possible to add an accessory word attached to the free part together with the free part. .

In addition, as a standard pattern, "A to [B]"
If you register "Please give me the book to the boy", the Japanese sentence "Please give the book to the boy" is registered. ,
Only when it is known that "to boy" is related to the verb "pass", the prepositional translation "to" is determined. Therefore, only "to boy" is independently translated and processed to "to a boy". It is impossible to get the translation result.

In addition, in the conventional pattern matching combined type translation processing apparatus, it is determined whether or not the parallel translation pattern is applied by the superficial character string matching, so that the parallel translation pattern "A is passed to [B]" is passed. Even if it was realized by adding the preposition "to" as a translation of "B ni" to "Please", if the bilingual pattern is applied by matching only the superficial character strings, for example, "the book is handed over today. If the portion of "B to" is not human like "Please, please" there is a possibility that an incorrect translation result is output.

The present invention has been made to solve the above problems, and expands the degree of freedom in creating a bilingual pattern, expands the range of application of one bilingual pattern, and suppresses the misapplication of the bilingual pattern. However, it is an object of the present invention to provide a pattern matching combined type translation processing device capable of reducing the labor of creating a parallel translation pattern and improving the accuracy of the translation processing result of a sentence having a similar structure.

[0009]

According to the present invention, the aforementioned object is a translation processing device for converting a Japanese input sentence text into a translation output sentence of a target language, wherein a fixed word is included in the input sentence text. When there is a predetermined pattern composed of a fixed part corresponding to a part and a free part corresponding to any other character string, it has a pattern translation means for translating a translation pattern of the target language sentence corresponding to this. Even when the fixed part does not exist in the input sentence so that the pattern translation means can support one parallel translation pattern for a plurality of input sentences, the part corresponding to the free part of the predetermined pattern does not occur. Is achieved by the pattern matching combined type translation processing apparatus according to claim 1, which is provided with a pattern matching means that enables matching.

According to the present invention, the above-mentioned object is to extend the object of the free portion of the predetermined pattern or the translated pattern to not only a phrase but also a clause by the pattern translation means. The pattern matching according to claim 2, further comprising limiting means capable of limiting to a specific syntactic category such as a phrase or noun, and discriminating means capable of determining to which syntactic category the inputted object belongs. This is achieved by the combined translation processor.

According to the present invention, the above-mentioned object is that the pattern translation means selects any kind of postpositional word located at the end of a clause, phrase or word which is a target of the free part of the predetermined pattern. This is achieved by the pattern matching combined type translation processing apparatus according to claim 3, further comprising a designating means for designating.

According to the present invention, the above-mentioned object further comprises a designating means for designating the preposition located at the head of the free part of the translated pattern and the type of the prepositional phrase or the connective as desired. This is achieved by the translation processing device with combined pattern matching according to claim 4.

According to the present invention, the aforementioned object is to limit the meaning category of the central word to the clause, phrase or word targeted by the free portion of the predetermined pattern by the pattern translation means. The translation processing device according to claim 5, further comprising: a means and a matching means for matching the semantic category of the input object.

[0014]

In the pattern matching combined type translation processing device according to claim 1, the pattern translation means handles the fixed part corresponding to the fixed word part in the input Japanese sentence and any other character string. When there is a predetermined pattern composed of the free part, a pattern translation process for translating a corresponding translation pattern of the target language sentence is performed, and one parallel translation pattern can correspond to a plurality of input sentences by the pattern matching means. Therefore, even if the portion corresponding to the free portion of the predetermined pattern does not exist in the input sentence, the collation is possible if the fixed portions match. As a result, the application range of one bilingual pattern becomes wide. For example, the following two Japanese sentences, "please refer to Chapter 5 for details" and "Please refer to Chapter 5 for the details because this chapter omits the explanation". However, by allowing the setting of any free part that exists, the following parallel translation pattern "* A For details, see B" → "Refer t
By registering one "#B for the details # A.", it is possible to include the above two sentences in one bilingual pattern.
The symbol "*" of A represents an existing free part, and allows the character string corresponding to this part to be empty.

In the pattern matching combined type translation processing device according to the second aspect, the object applicable to the predetermined pattern of the parallel translation pattern or the free part in the translation pattern is expanded not only to the phrase but also to the clause by the syntactic category limiting means. Further, the object is limited to a specific syntactic category such as a clause, a phrase or a noun, and the syntactic category discriminating means discriminates to which syntactic category the inputted object belongs. As a result, it is possible to expand the degree of freedom in creating the bilingual translation pattern and suppress its misapplication.
By making it possible to apply to a free part of a predetermined pattern, for example, "* A surely B% node% think] →"
By registering a bilingual pattern such as "I'm sure #B #A.", You can translate a Japanese sentence in which the section corresponds to "B% section%", such as "I think he will come," It can be translated by applying a pattern. In addition, "% clause%" of the symbol of B% clause% in the above-mentioned parallel translation pattern means that the application of the parallel translation pattern is limited when the portion corresponding to this free part is a clause, and by this, the parallel translation pattern It is possible to more appropriately control the application range of.

According to another aspect of the present invention, in the pattern matching combined type translation processing device, the postword located at the end of the clause, phrase or word that can be applied to the free part of the predetermined pattern by the postword designating means. Is arbitrarily specified. For example, "* A surely B% section% think" → "I'm sure #B #
If you want to make the free part B of A. "optional, then if there is no part corresponding to the free part B, delete the accompanying case particle" to "and apply the bilingual pattern to the part corresponding to the free part B. If there is, it is possible to apply the bilingual pattern on condition that the case particle “to” is attached. That is, when the predetermined pattern is registered as “* A surely [B] think”, the sentence “I think” in which the free portion B is missing or a portion corresponding to the free portion B exists. It is possible to match sentences such as "I think it will rain" with the above-mentioned predetermined pattern.
Since the part "nuisance" corresponding to does not meet the condition that the designated postword is "to", it is not matched with the above-mentioned predetermined pattern, and the misapplication of the bilingual pattern can be suppressed. In addition,
The symbol “[,]” of [B and] in the parallel translation pattern represents an existing arbitrary free part with a postfix designation.

In the pattern matching combined type translation processing device according to the fourth aspect, the preposition designating means arbitrarily designates the type of the preposition and the preposition phrase or the conjunction located at the head of the free part of the translation pattern. . This allows
Use the pre-specified exact translation even for prepositional translations, etc. that are originally determined by the words to which the free part relates and that cannot be determined only from the clauses, phrases, or words corresponding to the free part. So, for example, the parallel translation pattern "Please pass A to [B]" → "Please hand #
If you register "A #B [to].", The part corresponding to the free part #B of the translation pattern will be designated as a translated word such as a noun corresponding to the free part B of the given pattern. An accurate translated sentence is generated simply by applying the translated word with the addition of "." Note that the symbol "[,]" in #B [to] of the above bilingual pattern indicates the preposition and preposition phrase at the beginning or the kind of conjunction. Represents the translation of a preposition and a prepositional phrase or conjunction of a free part.

In the pattern matching combined type translation processing device according to the fifth aspect, the meaning category of the central word is set for the clause, phrase or word that can be applied to the free part of the predetermined pattern by the meaning category limiting means. The semantic category is collated with respect to the object limited and inputted by the semantic category collating means. As a result, it is possible to prevent erroneous application of the bilingual translation pattern due to the difference in meaning, and to convert sentences having the same shape in the surface layer into different translated sentences depending on the difference in meaning. For example, by registering the parallel translation pattern "Please give A to [B] $ human $" → "Please hand #A #B [to].", The Japanese sentence "Please give the book to the boy" With regard to, since the part corresponding to the free part B, "boy (ni)", is a noun included in the specified semantic category "human", the bilingual pattern is applied, and "the book "Please pass" is not a noun that is not included in the specified meaning category "human", which is the part corresponding to the free part B, "today (ni)",
Although this Japanese sentence corresponds to the bilingual pattern in the surface layer, it is not applied to the translation process due to the difference in meaning. The symbol "$ Human $" of $ human $ in the bilingual pattern [B] means that the application of the bilingual pattern is limited when the meaning category of the central word of the portion corresponding to the free portion is human. Thus, the range of application of the bilingual pattern can be controlled more appropriately.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of a pattern matching combined type translation processing apparatus of the present invention will be described below with reference to FIG. 1 when the target language is English.

The pattern matching combined type translation processing apparatus of this embodiment corresponds to the input section 1 for inputting Japanese sentence data, the fixed section corresponding to the fixed word portion of the Japanese sentence pattern, and any other character string. A bilingual pattern database 2 for storing a bilingual pattern composed of a fixed pattern composed of a free part and a translation pattern of a target language sentence corresponding thereto, and an input sentence sent from the input part 1 and a bilingual pattern database 2 stored therein. A pattern translation control unit 3 as a pattern translation unit for controlling the pattern translation process from the pattern matching with the existing fixed pattern to the setting of the translation process result into the translated pattern, and the entire input sentence or the input sent from the pattern translation control unit 3. The analysis processing unit 4 that analyzes the syntactic structure of the sentence portion and the analysis result obtained by the analysis processing unit 4 are used to construct the target language. A conversion generating unit 5 for generating a target language statements into the structure, the output unit 6 for outputting the output text generated by the pattern translation controller 3 or the conversion generation processing unit 5
And a translation processing dictionary unit 7 for accumulating necessary data in the analysis processing unit 4 and the conversion generation processing unit 5.

Further, as shown in FIG. 2, the pattern translation control unit 3 has a portion corresponding to the free portion of the fixed pattern in the input sentence so that one bilingual translation pattern can correspond to a plurality of input sentences. The pattern matching means 8 that enables matching regardless of whether or not, and the target applicable to the free part in the fixed pattern or the translated pattern of the bilingual pattern data is expanded to not only phrase but also clause,
A syntax category limiting unit 14 that can limit to a specific syntax category such as a phrase or a noun, a syntax category determining unit 9 that can determine to which syntax category the input object belongs, and a fixed pattern A postposition word designating means 10 for arbitrarily specifying the type of postposition word located at the end of a clause, phrase or word that can be applied to the free part, and a preposition located at the head of the free part of the translation pattern. And a preposition designating means 11 for arbitrarily designating a preposition phrase or a kind of conjunction, and a clause applicable to a free part of a fixed pattern,
It is provided with a meaning category limiting means 12 for limiting the meaning category of a central word with respect to a phrase or a word, and a meaning category matching means 13 for matching the meaning category of an input object.

Next, the operation of this embodiment will be described with reference to the flowchart of FIG.

In the input unit 1, one sentence is extracted from the Japanese sentence data, and the extracted sentence is sent to the pattern translation control unit 3 (step S1). It is judged whether or not there is the corresponding bilingual pattern data in the bilingual pattern database 2 (step S2). If there is no corresponding bilingual pattern data, the translation processing is performed without using the pattern matching. Information flag is set (step S3). If it is determined that there is the corresponding parallel translation pattern data, the pattern translation control unit 3 retrieves one corresponding parallel translation pattern data from the parallel translation pattern database 2 as shown in FIG.
As shown in FIGS. 5B and 5C, the fixed pattern and the translated pattern are separately set in the fixed portion and the free portion and set in the buffer. At this time, if the free part of the fixed pattern exists, it is discriminated whether it is optional or required, and if there is a postfix designation, the designated postword, if there is a semantic category designation such as "human" or "abstract concept", If there is a preposition designation in the meaning category and in the free part of the translation pattern, the designated preposition is also written in the buffer in a predetermined format (step S4). Then, the input Japanese data and the fixed pattern are collated, and it is determined whether or not the input Japanese sentence data and the fixed pattern match (step S5). If the Japanese sentence data and the fixed pattern do not match, the process returns to step S2, and matching with another corresponding parallel translation pattern data is continued. Also, if there is a match, in the state in which the portions corresponding to each free part are extracted from the input sentence and set, respectively, the information that “the pattern translation process is performed” is received in response to the input sentence matching the fixed pattern. The flag is set (step S6). It is judged whether or not there is a free part of the corresponding translation pattern in which the translation result is not stored among the free parts of the fixed pattern (step S
7) If there is an unprocessed free part, the character string extracted from the input sentence and set in the free part of the fixed pattern in which the translation result is not stored in the free part of the corresponding translation pattern is sent to the analysis processing part. It is sent (step S8).

Then, the analysis processing unit 4 performs analysis processing on the sent character string. For example, when the character string “his son or daughter” is sent, morphological analysis and syntactic analysis are performed, and finally, FIG.
A dependency structure representing a dependency structure as shown in (3) is formed (step S9). If the information flag indicating “to perform pattern translation processing” is set, the analysis result is once returned to the pattern translation control unit 3. Here, it is determined whether or not there is a syntax category specification or a semantic category specification in the free part (step S10). As a result of the analysis process, it is determined whether or not the specified condition of the specified syntax category or the semantic category is satisfied (step S11). If the condition is not satisfied, the process returns to step S2, and another corresponding parallel translation pattern is obtained. Continue to collate with the data. When the condition is satisfied, the analysis result is sent to the conversion generation processing unit 5. At this time, as shown in FIG. 8D, the highest bunsetsu in the dependency structure of the analysis result is set as the central word bunsetsu, and it is judged whether or not the specified conditions of the syntax category and the semantic category are satisfied. Done by clause. That is, if the syntactic category is specified as "comprehension",
If the central phrase is a "boiler phrase", it is judged that the specified condition is satisfied. If the central phrase is "a phrase", it is judged that the specified condition is satisfied. If the word is designated as a "word", it is determined that the designated condition is satisfied unless a lower clause is present with respect to the central word clause in the formed dependency structure. Further, with respect to the designation of the semantic category, it is determined whether or not the independent word in the central word bunsetsu that can be a bunsetsu alone satisfies the condition based on the semantic feature possessed as dictionary data.
For example, the third element (FIG. 8) of the fixed pattern of FIG.
No. in (c). In the case of the free part of 3 heading “B”), the condition that the semantic category is “human” is specified, but as shown in FIG. 8D, the independent word in the central phrase of this free part Is a "daughter", and if the dictionary data of the noun "daughter" has a semantic feature of "human", the input sentence part "his son or daughter" is a semantic category in this case. It is judged that the specified condition of is satisfied.

Then, the conversion generation processing unit 5 performs conversion generation processing on the sent character string (step S12). It is judged whether or not it is the normal translation processing (step S13), and if the information flag indicating "it is the normal translation processing" is set, the conversion generation result is sent to the output unit 6 and sent by the output unit 6. The translated result is output (step S14). If it is not the normal translation process, the conversion generation result is sent to the pattern translation control unit 3, and the pattern translation control unit 3 changes the result as shown in FIG.
The conversion generation result is stored in the free part of the translation pattern corresponding to the free part, and the process returns to step S7 to perform processing on the remaining free parts of the fixed pattern. In addition, step S7 described above.
If there is no unprocessed free part, the translated words of the fixed parts and free parts stored in the translation pattern are arranged in the specified order (step S15), and the translated text is sent to the output part 6 as the translation result. And the translation result is output.

Next, referring to the flow charts of FIGS. 4a and 4b and FIGS. 6 and 7 showing an embodiment in which this processing is applied to an actual example sentence, the Japanese sentence data is collated with a fixed pattern. The method will be described.

In FIGS. 6 and 7, the shaded portion indicates the portion being collated, the position of the black arrow in the Japanese sentence data indicates the next collation start position, and the position of the middle arrow indicates the virtual start of the previous free part. The position is shown. Fig. 6 shows an example of a bilingual pattern that begins with a fixed pattern. The bilingual pattern is "Make sure to give A to [B] $ human $" → "Please besu
"re to hand #A #B [to]." and the input sentence to be matched is "Make sure to give this book to his son or daughter." On the other hand, FIG. 7 shows an example of the bilingual pattern in which the fixed pattern starts from the free part.
[] Give me $ human $ "→" Please hand #A #B [t
o]. "and the input sentence to be matched is" please give this book to his son or daughter. "

The initial letter of the input Japanese sentence is set as the collation start position (step S21), and it is determined whether or not the fixed pattern starts from the fixed portion (step S22). When the fixed pattern starts from the fixed part, it is determined whether or not the length of the input Japanese sentence is equal to or longer than the fixed portion (step S23), and the length of the input Japanese sentence is equal to or longer than the fixed portion. If it is determined, it is determined whether or not the fixed part character string and the character string at the beginning of the input Japanese sentence of the number-of-characters sentence match (step S24). If it is determined that they match, the collation start position is moved to the "next collation start position" in FIGS. 6 and 7, which is the character after the portion that matches the character string of the standard part (step S25). For example, in the case of FIG.
Since the first element of the fixed pattern, "always", and the initial character string of the input sentence match, the matching process is continued. Then, it is determined whether or not the matching start position is at the end of the sentence of the input Japanese sentence (step S26). If it is determined that the matching start position is at the end of the sentence of the input Japanese sentence, all the fixed parts of the fixed pattern and It is determined whether or not the matching process for the free part is completed, or only one existing arbitrary free part without postfix designation remains (step S).
27). If it is determined to be in any state, it is determined that the input Japanese sentence and the fixed pattern match (step S28), and the matching process for the fixed pattern ends. For example, in the case of FIG. 6 (h) and FIG. 7 (g), since the collation start position is at the end of the Japanese sentence and the collation processing for all the fixed parts and free parts of the fixed pattern has ended, In the example as well, it is determined that the fixed pattern matches the input Japanese sentence.

If it is determined in step S26 that the collation start position is not at the end of the sentence, it is determined whether or not the collation processing has been completed for all the fixed parts and free parts of the fixed pattern (step S29). If it is determined that the matching process is completed, it is determined that the input Japanese sentence and the fixed pattern do not match, and the matching process for the fixed pattern is ended (step S30). If it is determined in step S23 that the input sentence length is not greater than or equal to the fixed phrase length, or if it is determined in step S24 that the fixed part does not match the character string at the beginning of the sentence, the fixed pattern is determined in step S27. Similarly, if it is determined that the collation processing for all the fixed parts and free parts of the above has not been completed, or if only one free arbitrary part that does not have a postfix designation remains Assuming that the Japanese sentence and the fixed pattern do not match, the matching process for the fixed pattern ends.

When it is determined in the above step S29 that the collation processing is not completed, it is determined whether or not the collation processing is completed for all the fixed parts and free parts of the fixed pattern in the next free part ( Step S3
1) If it is determined that the collation processing is completed, it is determined whether or not there is a postword for the free part (step S32). When it is determined that the postfix word is designated, it is determined whether the designated postword and the character string at the end of the input Japanese sentence of the number of characters of the designated postword match (step S33).
If they match, the process proceeds to step S28, and it is determined that the input Japanese sentence and the fixed pattern match, and the matching process for the fixed pattern ends. If they do not match, the process proceeds to step S30, and it is determined that the input Japanese and the standard pattern do not match, and the matching process for the standard pattern ends.

If it is determined in the above step S31 that the collation processing is not completed, it is determined whether or not the next free part is optional (step S34). If the next free part is not optional, the verification is performed. The start position is moved backward by one character (step S35). For example, in the case of the example of FIG. 6B, since the second element of the fixed pattern is an indispensable free part, the character string next to “must” of the fixed part which is the first element of the fixed pattern has at least one character. The above must apply to the free part of the second element. Therefore, by setting the collation start position of the fixed part of the third fixed pattern pattern element that is shifted backward by one character, the free part corresponding to the essential existence is secured. This is the same as in the case where the beginning of the input sentence corresponds to the free portion that is essential to exist, as shown in FIG. Then, it is determined whether or not the length of the input Japanese sentence after the collation start position is equal to or longer than the length of the next fixed part (step S36), and the length of the input Japanese sentence after the collation start position is of the next fixed part. If it is determined that the length is equal to or longer than the length, it is determined whether or not the fixed part character string and the character string for the input Japanese of the number of characters after the collation start position match (step S).
37). If they do not match, the process returns to step S35.
When it is determined that they coincide with each other, it is determined whether or not there is a postword designation in the free part immediately before the fixed part (step S3).
8). For example, as shown in FIG. 6D, the portion corresponding to the free portion of the second element of the fixed pattern is determined. This also applies to FIGS. 7B and 7C. Also,
In the case of FIG. 6 (f) or FIG. 7 (e), similarly, the part corresponding to the free part is determined, but since the postword is specified for the free part, is the input sentence applicable to the pattern? The determination as to whether or not to make the determination is made in the steps described below. When there is a postfix word designation, it is determined whether or not the designated postword and the character string at the end of the free portion for the number of characters of the designated postword match (step S3).
9). At this time, for example, as shown in FIG. 7 (e), assuming that the fixed part of the fourth element of the fixed pattern is collated with "pass me" after the 12th character of the input sentence example, the freedom immediately before this Since the postfix "ni" is specified for the free part of the third element (heading "B") of the fixed pattern which is the part, matching has not been completed for the input corresponding to the free part. The end of "his son or daughter" after the 5th character and the specified postfix "ni" are collated, and as shown in FIG. Continued. This also applies to FIGS. 6F and 6G.

In the input Japanese sentence, the character string of the part corresponding to the free part immediately before the fixed part is set in the buffer (step S40), and the operation proceeds to step S25 to repeat the operations thereafter. For example, in the case of the example of FIG. 7, the buffer state at the time when the free part of the third element of the fixed pattern is determined in FIG. 7 (f) and the corresponding part is set in the buffer is conceptually shown in FIG. 8 (c). Shown.

Further, NO in the above steps S36 and S39.
If it is determined that the input Japanese and the fixed pattern do not match, the collation process for the fixed pattern ends.

As described above, according to the pattern matching combined type translation processing apparatus of the present embodiment, the degree of freedom in bilingual pattern creation is expanded, the application range of one bilingual translation pattern is expanded, and the misapplication of bilingual translation patterns is prevented. It is possible to suppress, reduce the labor of creating the parallel translation pattern, and improve the accuracy of the translation processing result of the sentence having a similar structure.

[0035]

According to the translation processing apparatus with combined pattern matching of the first aspect, the application range of one parallel translation pattern can be widened. As a result, it is possible to save labor in creating a parallel translation pattern and improve the accuracy of translation processing.

According to the pattern matching combined type translation processing device of the second aspect, it is possible to extend the degree of freedom of creating the parallel translation pattern and to suppress the misapplication thereof. As a result, it is possible to secure the degree of freedom in creating the bilingual translation pattern and suppress misapplication, and it is possible to reduce the number of cases in which bilateral translation patterns cannot be registered due to consideration of misapplication.

According to the pattern matching combined type translation processing device of the third aspect, the kind of the postfix word located at the end of the clause, phrase or word that can be applied to the free part of the fixed pattern is arbitrarily designated. It is possible to suppress erroneous application. As a result, it is possible to improve the accuracy of selecting a translated word for a postfix word and reduce the number of cases in which a bilingual translation pattern cannot be registered due to misapplication.

According to the pattern matching combined type translation processing device of the fourth aspect, the kind of the preposition and the preposition phrase or the conjunction located at the head of the free part of the translation pattern is arbitrarily designated. As a result, it is determined by what kind of words the dependent part of the free part is originally related to, and even for prepositional translations that cannot be determined only from the clauses / phrases / words corresponding to the free part, the correct prespecified translations are By using it, it is possible to make the translation word selection more accurate when using the bilingual translation pattern.

According to the pattern matching combined type translation processing device of the fifth aspect, with respect to a clause, a phrase or a word which can be applied to the free part of the fixed pattern, the semantic category of the central word is limited and inputted. The semantic category is matched with respect to the target object. As a result, it is possible to prevent erroneous application of the bilingual translation pattern due to the difference in meaning, and to convert sentences having the same shape in the surface layer into different translated sentences depending on the difference in meaning. As a result, it is possible to register a bilingual translation pattern that is applied only in a specific case, which could not be registered due to the possibility of incorrect application in the related art, and improve the effectiveness of the pattern matching combined translation process.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a pattern matching combined type translation processing apparatus according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of a pattern translation control unit of the embodiment of FIG.

FIG. 3 is a flowchart showing the operation of the embodiment of FIG.

4a is a part of a flowchart showing the operation of the matching process of the embodiment of FIG.

FIG. 4b is another part of the flowchart showing the operation of the matching process of the embodiment of FIG.

FIG. 5 is an explanatory diagram showing the concept of setting bilingual pattern data in the buffer in the embodiment of FIG.

FIG. 6 is an explanatory diagram of a matching method of a fixed pattern and an input sentence when the fixed pattern starts in a fixed portion in the embodiment of FIG.

FIG. 7 is an explanatory diagram of a matching method of a fixed pattern and an input sentence when the fixed pattern starts in a free portion in the embodiment of FIG.

8A and 8B are explanatory diagrams of a state of setting an input sentence matching a fixed pattern of the embodiment of FIG. 1 in a buffer, a relationship between a dependency structure of an analysis result and a central word phrase, and a method of outputting a translation result.

[Explanation of symbols]

 1 input part 2 bilingual pattern database 3 pattern translation control part 4 analysis processing part 5 conversion generation processing part 6 output part 7 translation processing dictionary part 8 pattern matching means 9 syntax category discrimination means 10 postposition word designation means 11 preposition designation means 12 Meaning category limiting means 13 Meaning category matching means 14 Syntax category limiting means

Claims (5)

[Claims] 1. A translation processing device for converting a Japanese input sentence text into a translation output sentence of a target language, wherein a fixed part corresponding to a fixed word part and an arbitrary character string other than the fixed part are included in the input sentence text. When there is a predetermined pattern composed of the corresponding free part, it has a pattern translation means for translating a translation pattern of the target language sentence corresponding to the predetermined pattern, and the pattern translation means has a plurality of parallel translation patterns. A pattern matching combined type having a pattern matching means that enables matching when the fixed part matches even if the part corresponding to the free part of the predetermined pattern does not exist in the input sentence so as to be compatible with the input sentence. Translation processor. 2. The pattern translation means extends the target of the free part of the predetermined pattern or translated pattern to not only a phrase but also a clause, and further, the target is a specific syntactic category such as a clause, phrase or noun. 2. The pattern matching combined type translation processing device according to claim 1, further comprising: a limiting unit that can limit the input target and a determining unit that can determine which syntax category the input target belongs to. 3. The pattern translation means further comprises designation means for arbitrarily designating a type of a postpositional word located at the end of a clause, phrase or word to be a target of a free part of the predetermined pattern. 3. A pattern matching combined type translation processing device according to item 1 or 2. 4. The pattern translation means further comprises designation means for arbitrarily designating a preposition and a preposition phrase or a conjunction which are located at the head of the free part of the translation pattern. A translation processing device according to item 1, which is used in combination with pattern matching. 5. The limiting means for limiting the meaning category of a word, which is the center of a clause, a phrase, or a word, which is a target of the free portion of the predetermined pattern, by the pattern translation means, and the limiting means for the input object. 5. The pattern matching combined type translation processing device according to claim 1, further comprising a matching unit that matches the semantic category.