JPH0312344B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to machine translation of natural languages, particularly machine translation from English to Japanese.

[Prior Art] In recent years, an English-Japanese machine translation method that uses a dictionary and grammar rules registered in a storage device to analyze and convert an English input sentence to generate a Japanese translated sentence has been attracting attention.

Generally, most practical English-Japanese machine translation methods use the transfer method. The transfer method first analyzes the input sentence according to the analysis dictionary registered in the storage device and the rules of the analysis method.
Obtain an intermediate representation of the input sentence, then convert this intermediate representation into an intermediate representation of the output sentence (translated sentence) according to the conversion dictionary and conversion rules registered in the storage device, and finally This method uses this intermediate representation to generate a target translated sentence according to a generative dictionary and generative grammar rules.

[Problems to be solved by the invention] In the transfer method of the conventional English-Japanese machine translation system, an input English sentence is parsed to a certain extent to obtain an intermediate representation representing the syntax, which is then converted to produce a Japanese sentence. There are two methods: one that deeply parses the input English sentence, performs further semantic analysis, obtains an intermediate representation that represents the meaning, converts it, and outputs a Japanese sentence. According to the former method, the structure of the English-Japanese machine translation device is simpler and smaller, and the translation speed is faster, but it has the disadvantage that the quality of the translated text is poor. On the other hand, according to the latter method, the quality of the translated text improves, but the structure of the English-Japanese machine translation device becomes complicated and large, and the amount of language data to be registered in the storage device becomes enormous.
Translation speed is slowed down, further maintenance of translation equipment,
The drawback is that it is not easy to manage. In addition, because the analysis is done deeply into the intermediate expressions of the English sentence that express the meaning, it is difficult to reflect the subtle nuances and idiomatic expressions of the surface English sentence into the Japanese sentence.

Therefore, an object of the present invention is to exploit the advantages of an English-Japanese machine translation method in which a parse tree obtained by parsing an input English sentence is used as an English sentence intermediate representation and is transferred to a corresponding Japanese intermediate representation. It can be realized with a relatively simple system configuration and processing procedure using a relatively small amount of linguistic data, and in particular, it is based on the characteristics that almost no processing related to writing style is required.
In order to significantly improve the poor quality of translated sentences, which is a drawback of this method, a processing step is inserted to transform the English sentence parse tree into another English sentence parse tree as necessary before the English-Japanese conversion. First, we transform unique English expressions, including idiomatic phrases, into English expressions that are easy to convert directly into Japanese, improve the accuracy of subsequent English-Japanese conversion, and improve the English-Japanese transfer part. The English-Japanese translation system is designed to be easy to maintain and manage, speed up the English-Japanese transfer process, and output high-quality translations at high speed as a whole. The purpose is to provide a machine translation method.

[Means for Solving the Problems] In the English-Japanese transfer method of the present invention, first, an English input sentence is parsed according to the parsing dictionary and parsing rules registered in the storage device, and as an intermediate representation,
Obtain a parse tree representing the syntax of the input English sentence. Next, starting from the top node of this English sentence parse tree and working your way down to the lower nodes, a group of transformation rules grouped by type of node in the parse tree are applied to each node along the structure of this parse tree. By doing this, the structure of an English sentence parsing tree is transformed into another English sentence parsing tree, and at the same time, at each node, by applying a group of conversion procedure rules grouped according to the type of each node, Determine the procedure for the English-Japanese conversion process for a one-level range between that node and the group of child nodes of that node, that is, the conversion process from the English sentence parse tree to the corresponding Japanese tree structure, and store these in the storage device. I will make it memorized. The above transformation rules are rules for transforming idiomatic expressions and special constructions of English sentences into basic sentence patterns. This transformation process is performed to facilitate the later English-Japanese conversion process. Further, the actual conversion process is not performed at the time when the conversion process procedure is determined. When the transformation process from upper to lower direction and the determination and storage of the conversion procedure reach the lowest node of the English parse tree, the actual conversion process is performed according to the already memorized English-Japanese conversion processing procedure. We will carry out the conversion process. This conversion processing procedure is executed by a group of specialized structural conversion execution devices according to the type of each node and the status of its child nodes.
The Japanese intermediate representation corresponding to the input English sentence parse tree, that is, the Japanese tree structure, is created one after another from the lower node to the upper node, and finally the English-Japanese conversion is performed for the top node of the input English sentence parse tree. Perform processing to obtain the desired Japanese intermediate representation.

Next, from this intermediate representation of the Japanese sentence, a Japanese sentence generating device outputs a surface Japanese sentence using a Japanese sentence generating dictionary.

[Example] An example of the present invention will be described below with reference to the drawings. FIG. 1 shows an overview of the process of automatically translating an English sentence 100 into a Japanese sentence 120 using the English-Japanese machine translation method of the present invention.

When an English sentence is input through the input device 101, words, idioms, and technical terms are recognized using the English sentence analysis dictionary 106, that is, morphological processing is performed. Next, the input English sentence is analyzed by the English sentence analysis device 102 according to the analysis rules of the English sentence analysis rule dictionary 107, and an intermediate representation of the English sentence, that is, an English sentence parse tree is output, and the English sentence intermediate representation storage device 112 is stored in Next, the English-Japanese intermediate expression transformation conversion device 103 converts this English sentence parse tree into the English sentence intermediate expression transformation rule dictionary 108, the English-Japanese intermediate expression conversion procedure rule dictionary 109, and the English-Japanese conversion dictionary 1.
10, the intermediate representation of the Japanese sentence is transferred and stored in the Japanese sentence intermediate representation storage device 113. This intermediate representation is also a tree structure expressing the syntax of a Japanese sentence. Subsequently, a Japanese sentence is generated from this intermediate representation of the Japanese sentence by the Japanese sentence generating device 104 using the Japanese sentence generating dictionary 111. Finally, the output device 105 outputs the target Japanese sentence 120.

In the following, the English-Japanese machine translation system according to the present invention will be explained in more detail using an example, focusing on the processing by the English-Japanese intermediate representation transformation conversion device, which is the central part of the present invention, that is, the English-Japanese transfer processing. .

(1) As an example, we will explain the case of translating the simple English sentence shown in sentence (1) above into a Japanese sentence. first,
When the input English sentence shown in sentence (1) is analyzed by the English sentence analysis device 102 shown in FIG. 1, the English sentence parsing tree shown in FIG. 2 is obtained, and the English sentence intermediate representation storage device 1 shown in FIG.
12. In Figure 2, "DECL"
is a sentence, "NP" is a noun phrase, "VP" is a verb phrase,
"DET" is a determiner, "N" is a noun, "VERB" is a verb, "ADJ" is an adjective, and the lowest "THE", "PROGRAM", "PROVIDES",
"THESE" and "FUNCTIONS" represent the English words themselves. Further, the number of each node is added to identify each node. Next, the English sentence intermediate representation storage device 11 shown in FIG.
The English sentence parsing tree in Figure 2 stored in 2 is the first
This becomes the input to the English-Japanese intermediate representation transformation conversion device 103 shown in the figure.

FIG. 3-1 shows an outline of the English-Japanese intermediate representation transformation conversion device. The English sentence parsing tree input by the English sentence parsing tree input unit T1 is stored in the transformation conversion tree structure storage unit T2.
The contents of T2 at this time are as shown in Fig. 3-2. This T2 is a first-in, last-out stack, and stores subtree structures that have not yet been transformed. Next step T
Reference numeral 3 denotes a tree structure selection section which takes out the data in the storage section T2 in order from the most recently entered data. In this example, the tree structure shown in FIG. 2 is used.
The next step T4 is a transformation rule selection section, and T4 is a transformation rule selection section.
Only one applicable rule is selected from the group of transformation rules corresponding to the type of vertex node of the tree structure selected in step 3. The transformation rule group corresponding to each node used in this step is DECL of T51.
VP transformation rule storage unit of T52, NP transformation rule storage unit of T53, and T54 transformation rule storage unit.
It is stored in advance in the PP transformation rule storage unit, etc. In the case of the example in Figure 2, the vertex node is "DECL", so the transformations that can be applied to the tree structure in Figure 2 are selected from among the DECL transformation rules stored in the DECL transformation rule storage section of T51. Search for rules. However, in the case of this example, there is no transformation rule to be applied, and the tree structure in FIG. It will be done.

Here, a specific example of transformation processing for the "DECL" structure will be explained.

It is assumed that the record is torated on
(2) For example, in the case of the input English sentence shown in sentence (2) above, the English sentence parse tree shown in Figure 11a will be obtained, and from the transformation rule group for the "DECL" structure, The rule shown in the following equation (3) will be applied.

<(NP(PRON IT))(VP(BE-V IS)
(VERB*1) (NP(COMP THAT)*2))>
→＜(NP(COMP THAT) *2)(VP(BE−V
IS) (VERB*1))> (3) In the transformation rule of formula (3), the left side of the → mark describes the condition for whether or not to apply this rule, and this condition is is satisfied if the pattern matches for each child structure of . On the other hand, the right side of the → mark describes the subtree structure sequence that will be newly generated when the subtree structure below the "DECL" node in Figure 11a satisfies the matching condition of the pattern on the left side of the → mark. ing. For the sake of explanation, the content of equation (3) can be expressed graphically as shown in FIG. 11b. Next, the modification processing here will be explained based on FIG. 11b.
First, the subtree structure 2005 in FIG.
The pattern matches the subtree structure 2001 in Figure 1a. Next, the structure 2006 of FIG.
Structure 200 is compared with structure 2002 in figure a, but structure 200
The symbols *1 and *2 in 6 are variables, and when matching a pattern, they match any subtree structure existing at that location, and at the same time, the matched structure is assigned to these variables. Ru. Therefore, structure 20
The pattern for 06 also matches the structure 2002 in Figure 11a, the structure 2003 in Figure 11a is substituted for *1, and the structure 2002 for *2.
4, the structure shown in FIG. 11c is substituted. Here, all the subtree structures on the left side of Figure 11b have been processed, and at the same time, all child nodes of the "DECL" node in Figure 11a have also been processed, so the conditions of this transformation rule are satisfied. That means that. Therefore, structures below the "DECL" structure are generated according to the right side of the → mark in FIG. 11b. First, according to structure 2007, the 11d
A diagram structure is generated. However, here the structure that has already been assigned to *2 is used. Next, according to the structure 2008 of FIG. 11b, the structure 2008 of FIG.
A diagram structure is generated. In this case as well, the structure assigned to *1 is used. Finally, combine these structures as child nodes of the "DECL" node to obtain the result in Figure 11f. This transformation is intended to translate the English IT-THAT syntax into proper Japanese.

Next, we will show rules for transforming the English THERE-BE construct and examples of their application.

There are three types of commands. (4) The English sentence parse tree for the input English sentence of sentence (4) is as shown in Figure 12a, and the applied transformation rule is the following equation (5).

<(AVP(ADV THERE))(VP(BE-V
BE) (NP*1))＞→(NP*1)(VP(VERB
EXIST)> (5) Figure 12b shows the transformation results. In this variant, "THERE" and "ARE" are removed and the verb "EXIST" is used instead as shown at 2101 in Figure 12b.
has been introduced.

As a final example of transformation processing for the "DECL" structure, an example for transformation of the English TOO-TO syntax will be shown next.

The file is too large for you to use. (6) The parse tree for the input English sentence (6) is shown in Figure 13a, and the applied transformation rules are as follows.

<(NP*1)(VP(VERB*2)(AJP(ADV)
TOO) (ADJ*3) (PP (PREP FOR) (NP*
4) (INFCL(INFTO TO) *5)))>→<
(SUBCL(CONJ BECAUSE)(NP*1)(VP
(VERB*2) (AJP (ADV VERY) (ADJ*
3))))(NP*4)(VP(AUXV CAN)(ADV
NOT) *5)> (7) Figure 13b shows the transformation results. Transformation rules (7)
*3, *4, and *5 described therein are the same variables as *1 and *2 in the previous explanation. In this version, the TOO-TO syntax is transformed into a syntax using "BECAUSE".

By the way, the data sent to the transformation result storage section T7 in FIG. 3-1 has the tree structure shown in FIG. 2, and is stored there. The next step T8 is a conversion procedure rule selection section, which performs a one-level English-Japanese conversion between the apex node of the tree structure stored in the transformation result storage section T7 and its child nodes. The procedure rules are retrieved from a conversion procedure rule storage section prepared in advance for each type of node. These storage units include the DECL conversion procedure rule storage unit of T91, the VP conversion procedure rule storage unit of T92, the NP conversion procedure rule storage unit of T93, and the PP conversion procedure rule storage unit of T94.
For example, for AJP (adjective phrase) in T95.
In this example, the apex node of the tree structure in Figure 2 is "DECL", so the DECL conversion procedure rule group stored in the DECL conversion procedure rule storage section of T91 is searched, and the following application is applied. Get possible rules.

<(DECL*1)>→<(P-DECL*1)> (8) The left side (DECL*1) of → represents the matching pattern with the input tree structure, and its contents are as follows.
It is a tree structure with a vertex node called "DECL", and the structure of its child nodes is arbitrary, meaning that the child nodes are set to a variable called *1. On the other hand, the right side of → describes a procedure to perform a transformation called "P-DECL" using data in which all the child nodes set to *1 have been transformed. In the next step T10, the conversion procedure rule application section applies the conversion procedure rules to obtain the conversion procedure data shown in FIG. 3-3.

The conversion procedure data shown in FIG. 3-3 has a list structure including three data.
A Japanese intermediate expression corresponding to the "DECL" level in the English sentence parsing tree is created using the results of each conversion execution of the data 342.

The conversion procedure data shown in FIG. 3-3 has a list structure including three data.
It represents the name of the procedure for creating a Japanese intermediate expression corresponding to the "DECL" level in the English sentence parsing tree using the results of the conversion execution of the data 342. In other words, here the procedure name is "P-
DECL”. Generally, the format of conversion procedure data is such that the first data represents a conversion procedure name, and the second and subsequent data are data that will be used in the conversion procedure represented by this conversion procedure after the conversion is executed. .

This data is stored in the conversion procedure storage section of T11. The storage section of T11 is configured as a first-in, last-out stack, and the contents at this point are as shown in Figures 3-4. In Figure 3-4, the number added to the leftmost edge represents the node number of the node to which this conversion procedure data corresponds in the original English sentence parse tree; This is the node number of the corresponding node on the Japanese intermediate representation created by execution.

The next step T12 is a conversion execution determination section,
The contents of the conversion procedure storage unit of T11 and the conversion execution result storage unit of T16 are searched, and if the conversion procedure data stored most recently in T11 is executable, that data is retrieved and the data is retrieved from the next T13. The data is sent to the conversion procedure judgment unit. In the case of this example, the conversion execution determination unit of T12 first examines the procedure data most recently stored in T11, that is, the data shown in FIG. 3-3. Then, the conversion execution data for node numbers 2 and 3 is searched in the conversion execution result storage section of T16, but since nothing is stored in T16 yet, the process goes to the child node tree structure selection section of step 17. Go ahead and select the tree structure in which the apex nodes are 1 | DECL child nodes, 2 | NP and 3 | VP in Figure 2 from the procedure data in Figures 3-4, and
2 is stored in the tree structure storage unit for transformation conversion. The contents of the storage section T2 at this point are as shown in FIG. 3-5.

Next, the tree structure selection unit T3 first selects the third-
A tree structure with 2|NP in Figure 6 as the vertex node is extracted. In the next step T4, the transformation rule selection section, since the apex node of the target tree structure is "NP", the NP transformation rule group stored in the NP transformation rule storage section of T53 is searched, and the applicable transformation Select only one rule. However, in this example, there is no transformation rule to be applied, and the target tree structure passes through the transformation execution unit T6 and is stored as it is in the transformation result storage unit T7.
The contents of item 7 are as shown in Figure 3-6.

Here, a specific example of transformation processing for the "NP" structure will be explained.

It ensures proper execution of the
functions. (9) For example, in the case of the input English sentence above, Section 17a
The English sentence parse tree shown in the figure is obtained, and now consider the case where the "NP" structure 2601 is transformed. "NP"
The following rules are applied to the structure 2601 from among the transformation rules for the "NP" structure.

<(ADJ*1)(N EXECUTION)(PP
(PREP OF) *2) >>→ (COMP THAT)
(NP*2) (BE-V BE) (ADV*1) (VERB
EXECUTED) > (10) For the format and application of this transformation rule,
This is the same as described in the description of the transformation of the "DECL" structure, but in this case, for the child structure sequence of the subtree structure 2601 in FIG. 17a, that is, structures 2602, 2603, and , pattern matching processing is performed. As a result of the application of the rules,
These structures 2602, 2603, 2604 are
Structures 2606, 2607, 260 in Figure 17b
8, 2609, and 2610. When these are combined as a child structure of the "NP" node, the first
The “NP” structure shown in Figure 7c is obtained. When the ``NP'' structure that has undergone this deformation is restored to its original structure, it becomes as shown in Figure 17d. This variant is called “proper
A process to translate "execution of the functions" into a more preferable Japanese expression, such as "the correct execution of the function," rather than the literal translation of "the correct execution of the function." It is.

Next, an example of the "NP" structural modification for the idiomatic expression "a high level of ~" is shown.

The facility implements a high level of
data security. (11) The parse tree of this input English sentence is as shown in FIG. 18a. Let us now consider the case where the "NP" structure 2701 is transformed.

<(DET A) (ADJ HIGH) (N LEVEL)
(PP (PREP OF) *1)＞→＜(ADJ A-HIGH
-LEVEL-OF) *1> (12) According to the above transformation rule, "NP" in Figure 18b
Structure 2702 is transformed into “a high
"level of~" is one adjective "a-high-level-"
It is summarized as "of". This transformation results in
The subsequent English-Japanese conversion process will be considerably simplified.

As a final example of the modification of the "NP" structure, we will show an example of processing expressions using present participles.

You use the file containing the record.
(13) The parse tree of the input English sentence above becomes as shown in Figure 19a, and when the "NP" structure 2801 is transformed according to the transformation rule (14) below, it becomes the "NP" structure 2802 in Figure 19b. get good results.

<*1 (PRPRTCL (VERB CONTAINING)
*2)>→<*1(RELCL(PRON THAT)
(VERB CONTAIN) (NP*2))> (14) In this transformation, the present participle expression is replaced with a relative pronoun expression that has the same meaning.

By the way, the next step T8 in Figure 3-1
is a conversion procedure rule selection unit, and since the apex node of the tree structure data of the transformation result storage unit T7 is “NP”, the group of conversion procedure rules for NP stored in the conversion procedure rule storage unit for NP of T93 Search for
Get only one applicable rule. Among the conversion procedure rules for NP, as shown in equation (15) below,
NP containing a prepositional phrase (PP) as in Figure 3-7a
There are rules that can be applied to structures and rules that can be applied to NP structures that include relative pronoun phrases (RELCL) as shown in Figure 3-7b, as shown in equation (16).
In this example, such a rule is not applicable, and the rule in equation (17) below is selected.

<(NP(NP*1)(PP*2))>→<(P-NP
-PP*1*2)> (15) <(NP(NP*1)(RELCL*2))>→<(P-
NP-RELCL*1*2)> (16) <(NP*1)>→<(P-SNP*1)> (17) In the conversion procedure rule application part of the next step T10,
This rule is applied to obtain the procedural data shown in Figures 3-8. This data is stored in the conversion procedure storage section T11, and the contents of this storage section T11 are the third
- It will look like Figure 9. The conversion execution determination unit T12
Among the contents of the storage unit T11, the most recently stored procedure data, that is, the data shown in FIGS. 3-8, is checked to see if conversion is executable. code,
The conversion execution data of the tree structure with 4DET and 5N as vertex nodes is searched in the conversion execution storage section of T16, but since nothing is stored in T16, this procedure data is not executed and the process is moved to the next step T17.
In the child node tree structure selection part, these nodes,
A tree structure having 4DET and 5N as vertex nodes is selected and stored in the deformation transformation tree structure storage unit of T2. The contents of T2 at this point are as shown in Figure 3-10.

Next, the tree structure selection unit T3 selects and retrieves the most recently stored tree structure shown in FIG. The transformation rule selection part of T4 tries to select a transformation rule for the vertex node of this tree structure, but since the transformation rule group for DET is not prepared,
This tree structure passes through the transformation execution unit at T6 as it is, and is sent to the transformation result storage unit at T70. next T8
In the conversion procedure rule selection section, the previous "DECL",
As in the case of "NP", search the DET conversion procedure rule group to obtain the following rule (18), apply it in the conversion procedure rule application section T10, and send the result to the conversion procedure storage section T11. to be memorized.

<(DET*1)>→<(P-DET*1)> (18) The contents of T11 at this point are as shown in Figure 3-12.

Next, the most recently stored procedure data from the conversion procedure storage section of T11, that is, (P-DET
THE) is extracted and the conversion execution determination unit examines the feasibility of conversion. In the case of this procedure data, the conversion procedure name, the data used by P-DET, is a conversion of the English word "THE", so
It is determined that this is executable, and the next conversion procedure determination unit at T13 determines that it should be processed by the DET structure conversion execution unit at T149, and the English word "THE" is sent there. In T149, the determiner English-to-Japanese conversion dictionary storage unit in T155 is searched, the Japanese translation word ``Sono'' for the English word ``THE'' is retrieved, and the subtree structure of the Japanese intermediate expression shown in Figure 4a is created. . This result is stored in the conversion execution result storage section of T16, and T16 becomes as shown in FIG. 3-13. And, as the node number of "JDET", the procedure data in 3-12 (P-DET
The number "4" added to THE) is used.

Here, the conversion execution determination unit T12 performs the 3rd-14th
Conversion procedure storage unit T11 in the state shown in the figure
The most recent procedure data and the contents of the conversion execution result storage unit T16 as shown in Fig. 3-13 are searched, and a tree whose apex node is 4DET used by the procedure name P-SNP of this procedure data is created. Confirm that the structure conversion result has already been obtained at T16, replace the 4DET structure part of this procedure data with the conversion result, and change the contents of the conversion procedure storage unit as shown in Figure 3-15. do. Furthermore, it is checked whether the conversion result for the tree structure with 5N as the apex node of this procedure data is in the conversion execution result storage unit of T16, but the converted tree structure with 4JDET as the apex node has already been retrieved. Since nothing is stored in , in the end, step P-
SNP becomes infeasible. In this case, originally
By performing the conversion of (P-DET THE), P-
Since the feasibility of SNP has been investigated, there is no new tree structure group for child nodes to be transformed, and the process proceeds to the next tree structure selection section T3. At this point, the contents of the deformation transformation tree structure storage section T2 are stored in the third-
The tree structure is as shown in FIG. 16, and T3 selects the tree structure whose apex node is the most recently stored 5N, and sends it to the modification rule selection section of T4. Since a transformation rule group for "N" is not prepared, this tree structure data passes through the transformation execution unit at T6 and is stored as is in the transformation result storage unit at T7. In the next conversion procedure rule selection section T8, T
The following rule (19) is selected from the N conversion procedure rule storage section of 97.

<(N*1>→<(P-N*1)>) (19) The conversion procedure rule application section T10 applies it and stores it in the conversion procedure storage section T11 as shown in Figure 3-17. be remembered.

Next, the most recently stored procedure data from the conversion procedure storage unit of T11, that is, (P-N
PROGRAM) and convert execution judgment part T.
12, the feasibility of the conversion is checked. In the case of this procedure data, the data used by the conversion procedure name P-N is a conversion of the English word "PROGRAM", so this is determined to be executable, and the next conversion procedure judgment unit of T13 Therefore, N of T145
The structure conversion execution unit determines what should be processed and sends the English word ``PROGRAM'' to it. In T145, the noun English-Japanese conversion dictionary storage unit in T151 is searched, and the Japanese translation for the English word "PROGRAM" is one of the meaning markers of the noun classified in a specific field. '' is extracted, and the subtree structure of the Japanese intermediate expression shown in Figure 4b is created. This result is stored in the conversion execution result storage unit of T16, and T16 becomes as shown in FIG. 3-18.

Here, the conversion execution determination unit T12 selects the 3rd-15th
The most recent procedure data of the conversion procedure storage unit T11, which is in a state similar to that shown in the figure, and the 3-1
8. Conversion execution result storage unit T1 as shown in Figure 8.
6 and search for the procedure name P- of this procedure data.
Confirm that the conversion result of the tree structure with 5N as the vertex node used by SNP has already been obtained in T16, and replace the 5N structure part of this procedure data with the conversion result. The contents are as shown in Figure 3-19. At this point, the conversion data 4JDET and 5JN used by the procedure name P-SNP of this procedure data are complete, so this procedure data is executed. That is, since the procedure name of this procedure data is P-SNP, it is sent by the conversion procedure determining section T13 to the SNP structure conversion execution section T142, where the conversion is executed. Here, the structures shown in Figures 4a and 4b are used to perform semantic processing to determine the translated word and word order. Since it is simple and consists of a 5N structure and there is no ambiguity regarding word order, a Japanese intermediate expression as shown in Figure 5 is created. This result is stored in the conversion execution result storage section T16, and T16 becomes as shown in FIG. 3-20. In addition, the conversion procedure memory T11
is as shown in Figure 3-21.

Next, the conversion execution determination unit T12 uses the most recent procedure data of the conversion procedure storage unit T11 shown in FIG. 3-21 and the conversion execution result storage unit T1 of FIG. 3-20.
6 and search for the procedure name P- of this procedure data.
Confirm that the conversion result of the tree structure with 2NP as the vertex node used by DECL has already been obtained in T16, replace the 2NP structure part of this procedure data with the conversion result, and then T
11 as shown in Figure 3-22. Furthermore, it is checked whether the transformation result for the tree structure with 3VP of this procedure data as the vertex node exists in the transformation execution result storage unit of T16, but since nothing is already stored in T16, the result is that the procedure P-
DECL makes it impossible to perform conversions. In this case, since there is no tree structure group for child nodes to be newly transformed, the process next proceeds to tree structure selection section T3. At this point, the contents of the transformation tree structure storage unit T2 are as shown in Figure 3-23, and the tree structure whose apex node is the most recently stored 3VP is selected by T3. Then, it is sent to the transformation rule selection section of T4. Here, a transformation rule applicable to this tree structure is searched from the VP transformation rule storage T52 corresponding to the type of vertex node of this tree structure, that is, VP. However, in this example, there are no transformation rules to be applied to this tree structure, so the third-
The tree structure shown in FIG. 23 passes through the transformation execution section at the next step T6, and is sent as is to the transformation result storage section at the next step T7.

Here, a specific example of transformation processing for the "VP" structure will be explained. For example, in the case of the next input English sentence (20), the English sentence parsing tree becomes as shown in No. 14a, and now consider the case where the "VP" structure 2301 is transformed.

The program makes efficient use of the
file. (20) <(VERB MAKE)(NP(ADJ*1)(N
USE) (PP (PREP OF*2))>→<(ADV*1)
(VERB USE) (NP*2)> (21) The above rule (21) is applied to the “VP” structure 2301 from the group of transformation rules for the general “VP” structure. The form and application of this rule is the same as described in the description of the transformation of the "DECL" structure, but in this case the example structure of the child of structure 2301 in Figure 14a,
That is, pattern matching processing is performed on the structures 2302 and 2303. The result of applying this rule is the "VP" structure 2304 in Figure 14b. This variant is similar to the English word "make use of
As a result, as shown in Figure 14b, the noun "USE" becomes the verb "USE" and the adjective "EFFICIENT" is processed.
is replaced with the adverb "EFFICIENTLY".
This allows the expression "make use of ~" to be translated into a more appropriate expression in Japanese.

Next, “VP” for the idiomatic expression “be to～”
An example of structural deformation is shown.

The file is to be created. (22) <(BE−V BE)(INFCL(INFTO TO)*
1)>→<(AUXV WILL)*1)> (23) The parse tree of the input English sentence 22 is as shown in FIG. 15a, and now consider the case where the "VP" structure 2401 is transformed. According to the transformation rule 23, the "VP" structure 2402 in FIG. 15b is transformed, and "be to
~' are grouped together as one auxiliary verb 'will'. This transformation simplifies subsequent English-Japanese conversions.

As a final example of a variation of the “VP” structure, IT−
An example of a variation of the TO syntax is shown below.

You found it deficult to modify the
record. (24) <(VERB FIND)(NP(PRON IT)(AJP
(ADJ*1) (INFCL(INFTO TO)*2)))＞→
<(VERB FIND)(NP(COMP THAT)
(INFCL(INFTO TO)*2)(VERB IS)
(ADJ*1))> (25) The parse tree of the input English sentence (24) becomes as shown in Figure 16a, and the “VP” structure 2501 is transformed using the transformation rule (25).
When deformed according to , we obtain a result like the "VP" structure 2502 in FIG. 16b. This transformation also greatly facilitates the subsequent English-Japanese conversion process.

As a specific example of transformation processing, so far,
Although we have explained the "DECL" structure, "NP" structure, and "VP" structure, in addition,
A specific example of transformation processing for the "PP" (prepositional phrase) structure will be described.

(26) For example, in the case of the input English sentence (26) shown above, the English sentence parse tree is shown in Figure 20a, and now,
Consider the case where the “PP” structure 2901 is transformed.

＜(PREP IN)(N GENERAL)＞→＜
[ADV GENERALLY]> (27) The rule (27) shown above is applied to the “PP” structure 2901 from the group of transformation rules for the general “PP” structure. The form and application of this rule are the same as described in the description of the transformation of the "DECL" structure, but in this case the structure sequence of the children of structure 2901 in Figure 20a, i.e. structure 2902 , 2093 is subjected to pattern matching processing. However, on the right side of the → mark in rule (27) [ADV
GENERALLY] means that if the conditions are met for this "PP" structure, instead of replacing only the child structure of the "PP" structure, the statement "PP"
It instructs to replace the entire section including the section with the "ADV" structure as shown in Figure 20b. The result of applying this rule is the "ADV" structure 2094 in Figure 20c. This variant is called “in
The idiomatic expression "generally" is replaced with the simpler adverb "generally," making future English-Japanese translations easier.

Next, an example of a modification of the "PP" structure for the idiomatic expression "to some extent" is shown.

The file is used to some extent. (28) The parse tree of this input English sentence (28) is as shown in FIG. 21a. Let us now consider the case where the "PP" structure 3001 is transformed.

<(PREP TO)(ADJ SOME)(N
EXTENT）＞→＜[ADV TO−SOME−
EXTENT〕> (29) According to this transformation rule (29), it is transformed like the "ADV" structure 3002 in Figure 21b, and the "TO
``SOME EXTENT'' is one adverb equivalent phrase ``TO
−SOME−EXTENT”. This transformation simplifies subsequent English-Japanese conversions.

As a final example of a variant of the “PP” structure, “on
An example of transforming the idiomatic expression of "purpose" is shown below.

You erase the file on purpose. （30） ＜(PREP ON)(N PURPOSE)＞→＜
[ADV INTENTIONALLY]> (31) The parse tree of the above input English sentence (30) becomes as shown in Fig. 22a, and when the “PP” structure 3101 is transformed according to the above transformation rule (31), it becomes as shown in Fig. 22b. We get a result like “ADV” structure 3102. This transformation also facilitates subsequent English-Japanese conversion.

By the way, the data sent to the transformation result storage section T7 in FIG. 3-1 has the tree structure shown in FIG. 3-24, and is stored in T7. In the next step, the conversion procedure rule selection unit T8 obtains the following applicable transformation rules from the VP conversion procedure rule storage unit T92 in accordance with the apex node "VP" of the tree structure stored in T7.

<VP*1)>→<(P-VP*1)> (32) In the next conversion procedure rule application part of T10, this rule is applied to the target tree structure, and the result is the conversion procedure of T11. It is stored in the procedure storage unit. T11 at this point is as shown in FIG. 3-25.

Next, the conversion execution determination unit T12 selects the storage unit T11.
It is checked whether or not it is possible to convert the most recently stored procedure data, that is, the procedure data in the upper row of FIG. 3-25. node
The conversion execution data of the tree structure with nodes 6VERB and 7NP as the apex nodes is searched in the conversion execution storage section of T6, but nothing is stored in T16, and the tree structure with nodes 6VERB and 7NP as the apex nodes is ,
Since it is not an English word itself, this procedural data is not executed, and in the next step T17, the child node tree structure selection section selects a tree structure with these nodes 6VERB and 7NP as vertex nodes, and transforms T2. After being stored in the conversion tree structure storage unit, the contents of T2 become as shown in FIG. 3-26.

Next, the tree structure selection unit T3 selects the most recently stored tree structure shown in FIG. 3-27 from among the stored contents of the transformation/transformation tree structure storage unit T2. T4
The transformation rule selection part tries to select transformation rules for the vertex nodes of this tree structure, but
Since the transformation rule group for VERB is not prepared,
This tree structure passes through the transformation execution section of T6 as it is and is sent to the transformation result storage section of T7. next T8
The conversion procedure rule selection unit searches the VERB conversion procedure rule group to obtain the following rule (33), which is applied in the conversion procedure rule application unit T10, and the result is stored in the conversion procedure storage unit T11. Remember.

<(VERB*1)>→<(P-VERB*1)
(33) The contents of T11 at this point are as shown in Figure 3-28.

Next, the most recently stored procedure data from the conversion procedure storage unit of T11, that is, (P-VERB
PROVIDES) and convert execution judgment unit T.
12, the feasibility of the conversion is checked. Procedure name P
- The data used by VERB is a converted version of the English word "PROVIDES", so it is judged to be executable and should be processed by the VERB structure conversion execution unit in T148 by the conversion procedure determination unit in T13. After determining that, I sent the English word "PROVIDES" there. In T148, T154
Search the verb English-to-Japanese conversion dictionary storage section to retrieve the dictionary description for the English word "PROVIDE" and create the tree structure shown in Figure 6.

Block 601 in FIG. 6 shows the contents of the English-Japanese conversion dictionary for the verb "PROVIDE." Here, in an English sentence with "PROVIDE" as the main verb, which case particles are selected when the subject, direct object, indirect object, and prepositional phrase are translated into Japanese sentences, and what Japanese nouns are selected. In addition to the order of the clauses, it also describes which translation to select for ``PROVIDE.'' Blocks 601 (a) to (d) describe four types of selections, and the appropriate one that meets the conditions is selected. The "SB" line selects a case particle based on the head noun of the noun phrase that makes up the English subject (the "H" line), as well as the semantic marker of that noun phrase (``LE'', ``UD'', etc.). It represents the selection of case particles based on the ``S'' (line of ``S''). The information in this "SB" line forms one noun clause in the Japanese sentence. "DO"
The line ``SB'' describes the noun phrase that constitutes the direct object in an English sentence,
As a result, one noun clause in the Japanese sentence is also formed. Also, when a specific preposition is written, as in the line "WITH" in block 601 (b), it is used together with the verb "PROVIDE" to express a prepositional phrase that expresses a specific meaning. It represents a description for converting a into a corresponding Japanese noun clause. In other words, for the noun phrase that follows this preposition, "SB",
As in the case of "DO", case particles that match the conditions are selected to form one Japanese noun clause. The line "PP" in part (c) of block 601 describes the English-Japanese conversion of a prepositional phrase that is often used with the verb "PROVIDE", and it is written as "FOR".
The description in the line indicates processing for a prepositional phrase using the preposition "FOR". The presence of a prepositional phrase in this case is different from the case of "WITH" in part (b),
Optional in English sentences. With this description, one Japanese noun clause is also formed in part (b), just as in the case of ``WITH''. In addition to this, there are also description formats similar to "SB" and "DO".
There is a description of the line "IO". This is a description of the English-Japanese conversion of indirect objects in English sentences,
One Japanese noun clause is formed using the same method as "SB" and "DO". The last line of each description part (a) to (d) of block 601, the line "P", is a description for forming a predicate part in a Japanese sentence,
Translation for "PROVIDE" and incidental information for generating the auxiliary verb part of this Japanese verb in later Japanese generation. For example, in part (a), "lower one step", "continuation", "transitive verb", etc. are described. has been done.

By the way, the tree structure shown in FIG. 6 is stored in the conversion execution storage section of T16, and T16 becomes as shown in FIG. 3-29.

Here, the conversion execution determination unit T12 selects the 3-30th
Conversion procedure storage unit T11 in the state shown in the figure
The tree structure with 6VERB as the apex node of the most recent procedure data is replaced with the content of the conversion execution result storage unit T16 that has already been obtained, that is, the conversion execution result of the tree structure with 6VERB as the apex node, and the conversion procedure is executed. Regarding the tree structure whose apex is the 7NP node in the procedure data currently being targeted, the contents of the storage unit T11 are as shown in FIG. Since no results have been obtained, it is impossible to execute this procedure data, and the contents of the tree structure storage unit T2 for transformation are changed from the state shown in Figure 3-32 to the tree structure selection unit T3 again. Then, the tree structure for deformation conversion is extracted. The subsequent processing procedure is almost the same as the deformation conversion process for the tree structure with the 2NP node as the top node shown in Figure 3-6. First, the tree structure with the 7NP node as the top node is transformed. , when the conversion procedure rules are applied and stored in the conversion procedure storage section of T11, the contents of T11 become as shown in FIG. 3-33. Among the contents of T11, for the procedure data whose most recently stored procedure name is P-SNP,
The transformation execution results of the tree structure with 8ADJ as the apex node and the tree structure with 9N as the apex node have not yet been obtained in the transformation execution result storage unit T16, and the tree structure of these child nodes is the transformation of T2. It is sent to the conversion tree structure storage unit. At this point, the contents of T2 are as shown in Figure 3-34.

The specific processing procedure has already been described, so it will be omitted, but first, among the tree structures shown in Fig. 3-34, the results of the transformation processing of the tree structure with 8ADJ as the top node are shown in Fig. 7a. This result is replaced with a tree structure with 8ADJ as the top node among the contents shown in Figure 3-33, and the contents of the conversion procedure storage section of T11 are changed to the tree structure shown in Figure 3-3.
It will look like Figure 5.

Next, similarly, for the tree structure with 9N as the top node, after the deformation conversion process, the result is as shown in FIG. 7b. In Figure 7b, "FA"
is a type of noun semantic marker in a specific field, and represents a certain function or ability.
Here, among the procedure data shown in Fig. 3-35,
The procedure data with procedure name P-SNP is now executable.
The execution result is T1 as shown in Figure 3-36.
6 is stored in the conversion execution result storage unit.

At this point, the conversion procedure storage unit T11 stores the third-
As shown in Figure 37, the transformation execution result of a tree structure with 7NP as the top node is stored in the transformation execution result storage unit T16 as shown in Figure 3-36. , replace the procedure data of the target procedure name P-VP with a tree structure with 7NP as the top node. Here, the procedure name P-
The VP procedure data becomes executable, and the execution results are stored in the conversion execution result storage section T16 as shown in FIG. In addition, the conversion procedure storage unit T1
The contents of 1 are as shown in Figure 3-38.

Here, the tree structure whose apex node is 3VP in the procedure data stored in the conversion procedure storage unit T11 is replaced with the conversion execution result already stored in the conversion execution result storage unit T16, and eventually,
The contents of the conversion procedure storage section T11 are as shown in FIG. 3-39.

At this point, the conversion execution determination unit T12 determines that
Procedure name P stored in the conversion procedure storage unit T11
- It is determined that the procedure data of DECL is executable, and the transformed tree structure with 2JNP as the vertex node and the transformed tree structure with 3JVP as the vertex node are created by the transformation procedure determination unit T13. DECL
It is sent to the structure conversion execution unit T141. At T141, semantic processing is performed using the sent tree structure data shown in FIGS. 5 and 8 to create an appropriate Japanese intermediate representation for the input tree structure shown in FIG. 3-2.

That is, in FIGS. 5 and 8, the subject of this input English sentence is block 501 in FIG.
The meaning marker is "LE", the direct object is block 802 in FIG. 8, and the meaning marker is "FA" in the verb conversion dictionary description 801 in FIG. Only part (a) is selected from. As a result, the subject "that program"
The case particle taken by ``ga'' is taken, and the case particle taken by the direct object ``these functions'' is ``wo'', and the translation of this verb ``PROVIDE'' is determined to be ``preparation.'' Furthermore, the order of these words in Japanese follows the order of descriptions such as "SB" and "DO" in block 801(a) of FIG.

If the subject and object are polysemous,
Appropriate translated words can be selected through semantic processing here. For example, when targeting the field of information processing, the following input English sentence (34)
In this case, FIGS. 9a and 9b are obtained as structures corresponding to FIGS. 5 and 8.

(34) In block 901 of FIG. 9a,
Two translations of ``THEOPERATOR'' are shown, each separated by a semantic marker. In other words, the translation of "HM" (semantic marker for human being) is "operator" and "LE"
(a semantic marker representing a symbol, etc.) is translated as ``operator''. In this case, according to the condition described in block 902 of Figure 9b, only the meaning marker "HM" is allowed as the subject of this English sentence, so as a result, the translation of "THEOPERATOR" is "Operator" will be selected.

Also, at the stage of processing the structures shown in FIGS. 5 and 8, if the selection option described in block 801
Regarding (a) to (d), if block 501 and block 802 do not satisfy the conditions of any of the options, the structures of FIGS. 5 and 8 are as follows:
The English parse tree shown in Figure 2 is rejected as meaningless in the target field.

Next, a Japanese sentence intermediate representation is formed based on the results obtained from the above semantic processing, and the structure shown in FIG. 10 is obtained. In Figure 10, "JDECL" is a Japanese sentence, "JNC" is a Japanese noun clause, "JPC" is a Japanese pragmatic part, "JNP" is a Japanese noun phrase,
"JPART" is a Japanese particle, "JVERB" is a Japanese verb, and "JVAUX" is a clause identifier representing the Japanese auxiliary verb part.

The results obtained here are stored in the conversion execution result storage unit at T16. At this point, nothing is stored in the conversion procedure storage section T11. Moreover, nothing is stored in the transformation/transformation tree structure storage unit of T2. Therefore, the conversion process for the English sentence parse tree shown in FIG. 3-2 input from the English sentence parse tree input section of T1 is completed, and the tree structure of FIG. The result is taken out from the result storage section and outputted by the Japanese sentence intermediate expression output section T18.

At this stage, the processing in the English-Japanese intermediate representation transformation conversion device 103 of FIG. 1 is completed, and the Japanese intermediate representation shown in FIG. 10 is stored in the Japanese sentence intermediate representation storage device 113 of FIG.

Note that details of the processing procedure of the conversion execution determination unit T12 in FIG. 3-1 are shown in FIG. 3-40. Next, the Japanese sentence generation device 104 shown in FIG. 1 generates a Japanese sentence by inputting the Japanese sentence intermediate expression shown in FIG. First, change "prepared" to "prepared" according to the description in the section "JVAUX" in Figure 10. Next, by extracting and arranging the terminal Japanese words in the structure of this Japanese sentence intermediate expression in the same order as they are, and adding a period mark "." at the end, a Japanese sentence as shown below is obtained.

The program has these functions.

Finally, this result is outputted to a printing device or a display device by the output device 105 of FIG. 1, thereby completing the entire process in the English-Japanese machine translation method according to the present invention.

Although preferred embodiments of the present invention have been described above, it goes without saying that various modifications can be made in carrying out the present invention.

Although the above embodiments can be implemented using LISP programs, the present invention can also be implemented using programs written in other languages or by hardware devices.

[Effects of the Invention] Hereinafter, the present invention will be compared with a conventional method, and advantages specific to the method of the present invention will be explained.

It has the ability to efficiently transfer practical-scale idiomatic expressions used in general English sentences into appropriate Japanese expressions.

None of the conventional English-Japanese machine translation methods has the practical functionality to translate a practical scale of idiomatic expressions into appropriate Japanese expressions.
For example, there is a system in which an idiomatic expression dictionary is searched for each English word during syntactic analysis of an English sentence, and in the case of an idiomatic expression, exceptional processing is performed. In this type of method, idiomatic expressions are exceptionally processed during analysis, so if you try to process almost all idiomatic expressions commonly used in English sentences, the processing itself will become complicated. , maintenance and management become difficult, and the efficiency of the analysis itself becomes worse. Therefore, in reality, we are now in a situation where only a limited number of people can process things in idiomatic expressions.

On the other hand, there are conventional systems that process idiomatic expressions after parsing an English sentence to obtain an English sentence parse tree. is the center,
There is almost no processing function for English expressions related to Japanese writing style and word order.

In the method according to the present invention, appropriate processing is performed for English expressions related to Japanese writing style and word order by transforming the English parse tree, and Japanese verbs, nouns, and It is designed to appropriately process English expressions related to the selection of translation words, such as particles, and can handle a variety of idiomatic expressions on a practical scale.

The processing device for English idiomatic expressions consists of two processes: transformation processing that transforms the English parse tree into English basic sentence patterns, and conversion processing that transforms it into the corresponding Japanese intermediate expression.
It is organized and divided into two parts, making maintenance and management very easy.

When trying to process a variety of English idiomatic expressions on a practical scale using conventional transfer methods, it is difficult to process expressions that are related to the Japanese writing style and word order, as well as the translation of verbs, nouns, particles, etc. This is done by repeatedly applying a set of transfer rules up to a point in time, but the variety of types of conversion makes the conversion system complex, making it difficult to maintain and manage the types of transfer rules on a practical scale. become.

In the method according to the present invention, the English parse tree is first transformed into a simpler English basic sentence pattern, and then the English This is dealt with by converting it into days. This transformation process is
Transformation processing within the same English parse tree, i.e.
Since it is a process of transforming idiomatic English sentence patterns into basic sentence patterns expressing the same meaning, it is very easy to understand and easy to maintain and manage. In addition, after transformation, English-Japanese conversion is performed, focusing on translating Japanese verbs, nouns, particles, etc., but the input is an English parse tree that has been transformed into a simple basic sentence pattern, and the processing is simple. As a result, maintenance and management become easier.

Transfer processing of English idiomatic expressions can be performed at high speed.

In conventional general tree structure transformation/conversion methods, a method is adopted in which a single transformation/conversion rule group is prepared as a so-called production rule and is applied to an input tree structure. According to this method, the applicability of the input tree structure is checked sequentially from the first rule in the group of transformation/conversion rules by pattern matching, and if applicable,
Apply the transformation/conversion rules to transform/transform the tree structure.
Convert. Next, the applicability of the tree structure obtained by this transformation is examined again from the next definition in the transformation/conversion definition group. and,
Once an applicable rule is found, it is applied to transform and transform the input tree structure to create another tree structure.
When this process of applying rules is completed up to the last rule in the group of transformation/conversion rules, the process returns to the beginning of the group of transformation/conversion rules and continues applying rules in the same way. Then, when no rule in the group of transformation/conversion rules is applied to a certain input tree structure, the transformation/conversion process for the entire input tree structure is terminated. The process of checking the applicability of rules and transforming/transforming them using tree-structured pattern matching inherently requires a lot of processing time. Since the rules for all node types in the tree structure are stored together in the transformation rule group, when examining the applicability of one rule, even if the rule is Even if it is a rule, all nodes in the input tree structure must be examined, resulting in extra processing. Further, since the application of rules to the input tree structure is repeated until the rules of the transformation/transformation rule group can no longer be applied, pattern matching is often performed a large number of times.

In the method according to the present invention, transformation/transformation rules are grouped and stored for each type of target node in the tree structure, thereby reducing the number of unnecessary pattern matching operations. In addition, when applying each rule group, the applicability is checked starting from the first rule, but only the first rule that becomes applicable is applied once, and then the transformation/conversion rule group is applied for that node. end the application.

Furthermore, in the method according to the present invention, in order to minimize the number of searches for substructures of the tree structure in the transfer of the input English parse tree, that is, in the transformation/conversion process, the English parse tree is first transformed from the top node to the lower nodes. Moving in the direction of the nodes, at the same time, at each node, following this transformation process, the procedure for the English-Japanese conversion process is determined and stored. After the transformation process has finished up to the lowest node of the English parse tree, the Japanese intermediate expression is constructed from the lower-level substructure to the upper-level structure according to the already memorized steps of the English-Japanese conversion process. I will do it. In this way, the search for the partial structure of the input English parse tree only needs to be performed once, which is the minimum necessary tree structure search process. In general, when searching a tree structure, starting from the highest node and proceeding toward lower nodes is much simpler than starting from the lowest node and proceeding toward higher nodes. Also, when transforming from a higher level to a lower level, the transformation process will be performed again later on the subtree structure that has already been transformed at the higher level, so the transformation rules can be written It has the advantage of being flexible.

For the above reasons, the English-Japanese transfer method according to the present invention can perform English-Japanese transfer processing at high speed.

[Brief explanation of the drawing]

Figure 1 is a diagram showing an overview of one embodiment of the English-Japanese translation system of the present invention, and Figure 2 is an English sentence analysis obtained by parsing example English sentences used to explain the embodiment of the present invention. Figure 3-1 is a detailed diagram of the English-Japanese intermediate representation transformation conversion device in the system shown in Figure 1, and Figures 3-2 to 3-39 are the English sentence parsing tree shown in Figure 2. is a diagram for explaining the English-Japanese intermediate representation in Figure 3-1 and the process of processing carried out in the transformation conversion device, and Figure 3-40 is a diagram for explaining the processing procedure of the conversion execution judgment unit T12 in Figure 3-1. Figures 4a and 4b showing details, and Figures 5 to 8 are diagrams showing the process of converting the English sentence parse tree in Figure 2 to a Japanese intermediate representation, Figures 9a and 9b. Figure 10 is a diagram explaining an example of selecting translated words using semantic markers, Figure 10 is a diagram showing an intermediate representation of a Japanese sentence corresponding to the English parse tree in Figure 2, and Figures 11a to 11f are diagrams of the DECL structure. Figures 12a and 12b are diagrams for explaining one example of modification, and Figures 13a and 13b are diagrams for explaining a second example of modification of the DECL structure.
Figures 14a and 4b are diagrams for explaining a third example of deformation of the DECL structure, and Figures 15a and 15 are diagrams for explaining one example of deformation of the VP structure.
Figure b is a diagram for explaining a second example of modification of the VP structure, Figures 16a and 16b are diagrams for explaining a third example of modification of the VP structure, and Figures 17a to 17a are diagrams for explaining a third example of modification of the VP structure.
Figure 17d is a diagram for explaining one example of deformation of the NP structure, Figures 18a and 18b are diagrams for explaining a second example of deformation of the NP structure, and Figures 19a and 19b are diagrams for explaining a second example of deformation of the NP structure. Diagrams for explaining the third example of deformation of the NP structure, Figures 20a to 20c are PP
In order to explain an example of a modification of the structure, Fig. 21a
Figures 21b and 21b are diagrams for explaining a second example of modification of the PP structure, and Figures 22a and 22b are diagrams for explaining the second example of modification of the PP structure.
FIG. 7 is a diagram for explaining a third example of modification of the PP structure.

Claims (1)

[Claims] 1. An English sentence analysis device that parses an input English sentence and outputs a parse tree of English sentence intermediate expressions, and converts idiomatic expressions and special constructions existing in the English sentence intermediate expressions into basic sentence patterns. A transformation rule dictionary that holds a group of transformation rules for transforming the above parse tree into a group of nodes for each type of node, and a group of English-Japanese conversion rules for converting the parse tree into a Japanese intermediate representation. The analytic tree is transformed by using an English-Japanese conversion rule dictionary that is maintained separately for each type of node, and by applying the above transformation rules at each node along the structure of the analytic tree from the top to the bottom of the analytic tree. and a means for forming a Japanese sentence intermediate expression by applying the above-mentioned English-Japanese conversion rules along the structure of the parse tree from lower to upper transformed branches of the parse tree. An English-Japanese machine translation device comprising: a device for generating a Japanese sentence; and a Japanese sentence generating device for generating a Japanese sentence from the Japanese sentence intermediate representation. 2. The English-Japanese intermediate expression transformation conversion device has a means for simultaneously determining the conversion procedure to Japanese sentences at each node when transforming the above-mentioned parse tree from the upper to the lower direction, and the The English-Japanese machine translation device according to claim 1, wherein when the determination of the transformation and conversion procedure is completed, the Japanese sentence intermediate expression forming means executes the conversion operation from the lower to the upper part of the parse tree according to the conversion procedure. .