JPH1166069A - Machine translation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a machine translation device which generates a suitable translated sentence by canceling the polysemy of noun phrases included in a source sentence. SOLUTION: The source sentence is inputted to an input part 1, stored in a source sentence storage part 2 and inputted to a translation part 3. This translation part 3 generates the optimum tree structure of the source sentence through a grammar rule dictionary 4b by performing the morpheme analysis of the source sentence while using a word dictionary 4a at a translation dictionary part 4. The translation part 3 has a noun phrase analytic processing part and concerns the main noun of the noun phrase in the optimum tree structure and when it is noun from a verb derivatived, it is invertigated whether the corresponding verb requested term structure has successivety to the noun phrase or not. Further, using a term structure correspondence dictionary 4c, the semantie role of a phrase modifying the main noun is determined so as to perform the semantic analysis of the noun phrase. Based on this analysis, translation is executed, the result is stored in a translated sentence storage part 5 and the source sentence and the translated sentence are displayed on a display part 6. These source sentence and translated sentence are edited by editing control 7.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a machine translation apparatus, and more particularly, to a machine translation apparatus that generates an appropriate translated sentence by eliminating polysemy of a noun phrase included in an original sentence to be translated.

[0002]

2. Description of the Related Art In a conventional machine translation apparatus, various ambiguities are eliminated according to the stage of translation processing, such as determination of part of speech or determination of a syntax structure. For example, in the Japanese noun phrase expression “A, B, C”, polysemy elimination is performed using semantics and examples to determine the semantic relationship between A, B, and C (“natural language processing”). ”, Makoto Nagao, 199
6). Japanese Unexamined Patent Publication (Kokai) No. 63-213066 discloses techniques for eliminating polysemy in noun phrases and analyzing semantics.
The one disclosed in JP-A-4-343172 is known. The technique disclosed in Japanese Patent Publication No. 63-213066 relates to a method for resolving a polysemy relationship between two nouns constituting a compound word, and examines the lexical classification of nouns, semantic attributes, and inheritance relationships of semantic attributes. Is used to determine the relationship between nouns.

The technique disclosed in Japanese Patent Application Laid-Open No. 4-343172 relates to a technique for generating a natural translation by clarifying the dependencies within a complicated noun phrase and determining the operability of the main noun of the noun phrase. If this noun is derived from a verb, the dependency in the noun phrase is clarified based on the case frame of the verb, and the natural noun is obtained by translating the main noun into a verb. is there. In this case, the verb case frame contains both mandatory and optional case information,
A semantic marker of a noun that can be an individual case element is described, and is used to determine which case role other than the main noun of the noun phrase plays.

[0004]

Problems to be Solved by the Invention
According to the technology of JP-A-66, only the semantic analysis of a compound noun composed of two nouns can be performed, and information is used as to whether or not the main noun in the compound noun is a verb-derived noun. It does not do. Even if “noun word classification” corresponds to the above information, this is not clearly described in the official gazette. In addition, Japanese Unexamined Patent Publication
According to the technology disclosed in Japanese Patent No. 343172, a noun derived from a verb does not always inherit the case information of the original verb, and when the case information of the verb is inherited, the noun is always referred to as a prepositional phrase. It plays a case role in the form, and in that case, the correspondence between the case role and the prepositional phrase is determined depending on the lexical characteristics of the verb, so it is not necessary to determine the case role using the semantic marker.

The present invention has been made in view of the above situation, and focuses on the main noun of the noun phrase included in the original sentence. If the main noun is a verb-derived noun, the corresponding verb requests. Investigate the inheritance of the noun structure to the noun phrase, the number information of the main noun, the number information of the phrase modifying the main noun, and the structural and semantic information, and resolve the polysemy of the noun phrase, and a more appropriate translation To provide a machine translation device capable of generating a translation.

[0006]

According to a first aspect of the present invention, there is provided an input unit for inputting an original sentence to be translated, an original sentence storage unit for storing the original sentence input from the input unit, A translation dictionary unit containing knowledge information used for translation processing, a translation storage unit storing translations obtained as a result of translation processing of the original text, a display unit displaying the original text and the translation text, In a machine translation device having an editing control unit that performs editing processing of the displayed original sentence or the translated sentence,
The translation unit has a noun phrase analysis processing unit that performs a semantic analysis of the noun phrase, thereby making it possible to obtain a translated sentence with excellent understandability.

According to a second aspect of the present invention, in the first aspect of the present invention, when the main noun of the noun phrase is a verb-derived noun, the noun-phrase analysis processing unit requests a verb corresponding to the verb-derived noun. The semantic analysis of the noun phrase is performed by examining the inheritance of the noun phrase to the noun phrase, thereby making it possible to eliminate the polysemy of the noun phrase.

According to a third aspect of the present invention, in the second aspect of the present invention, the noun phrase analysis processing section includes the number information of the noun phrases, the number information of the words modifying the main noun, and structural information.
It is characterized by performing semantic analysis using semantic information, thereby enabling to determine the semantic role of a phrase that modifies the main noun of a noun phrase.

[0009]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a schematic diagram for explaining an embodiment of a machine translation apparatus according to the present invention.
Is an input unit for inputting an original sentence to be translated, 2 is an original sentence storage unit for storing the input original sentence, 3 is a translation unit for executing translation processing of the original sentence, 4 is a word dictionary 4a, a grammar rule dictionary 4b,
A translation dictionary unit having a term structure correspondence dictionary 4c, a translated sentence storage unit 5 for storing translated translations, a display unit 6 for displaying original sentences and translated sentences, and a central processing function 7, which is displayed on the display unit 6. The editing control unit performs editing processing of the original or translated sentence.

In FIG. 1, when an original sentence to be translated is inputted to an input unit 1, the original sentence is stored in an original sentence storage unit 2 and is translated by a word dictionary 4a of a translation dictionary unit 4, a grammar rule dictionary 4b, and a term. Translation processing is performed using the structure correspondence dictionary 4c. The translated sentence is stored in the translated sentence storage unit 5. The original sentence and the translated sentence are displayed on the display unit 6. The editing control unit 7 centrally processes the above-mentioned units, and performs an editing process of the original sentence and the translated sentence displayed on the display unit 6.

Next, FIG. 2 is a flowchart of the translation process performed by the translation unit 3, and the manner in which the translation process is performed will be described with reference to the flowchart. When the original sentence is input (Step 21), a morpheme is extracted from the original sentence using the word dictionary 4a of the translation dictionary unit 4, and an analysis is performed on how the extracted morphemes are combined to form a word. Is performed (Step 22). The result of this morphological analysis is passed to the next syntax analysis processing (Step 23). In this parsing process (Step 23), a tree structure is generated using parsing rules in the grammar rule dictionary 4b based on the part of speech information obtained by the morphological analysis. Next, using the predicate of the original sentence as a key, a tree structure is selected from all the syntactic structures that the predicate can take (Step 24). After this, a noun phrase semantic analysis process is executed (Step 25). When the semantic analysis of the noun phrase is completed, this result and the syntax structure obtained as a result of the syntax analysis are converted and generated into the target language (Step 2).
6). Here, translation is performed in the target language using the conversion / generation rules in the grammar rule dictionary 4b, and the result is output (Step 27).

FIG. 3 is a flowchart of the semantic analysis processing of the noun phrase. The manner in which the semantic analysis processing is performed according to this flowchart will be described. First, a noun phrase included in the original sentence is read (Step 31). Next, the word dictionary 4a is searched to determine whether the main noun which is the head (main word) of the read noun phrase has a verb derivative noun (Step 32).

FIG. 4 shows an example of the word dictionary 4a, in which nouns in the dictionary are given features indicating whether they are verb-derived nouns. If the main noun is a verb derivative noun, the process proceeds to the next process (Step 33), and if not, the process proceeds to a process for determining whether or not there is another noun phrase (Step 38). If the main noun of the read noun phrase is a verb derivative noun, the inheritance to the noun phrase of the term structure required by the corresponding verb is checked by referring to the inheritance determination table (Step 3).
3).

FIG. 5 shows an example of the inheritability judgment table, which is an "Argument Structur".
e ”According to the concept of J. Grimshaw (1990), verb derivative nouns are changed to“ result nominal, ”“ simple event nominal, ”“ comp
lex event nominal ”.
The above “result nominal,” “simple event nomina
l, ”and“ complex event nominal ”are used as they are. For example, the verb derivative noun“ assignment ”
As shown in the following example, "result nominal"
(1) They study the assignment. (“result nomin
al ”) (2) The assignment of difficult problems always
causes problems. (Use as "complex event nominal") The problem here is the polysemy of the noun.

The term structure required by a verb can be inherited only in the case of "complex event nominal".
Therefore, if the noun has no ambiguity by referring to the dictionary description information of the noun in the word dictionary 4a, it is compared with the inheritance determination table (FIG. 5) to determine whether the noun has the inheritance of the term structure of the verb. Is determined. If it is determined that there is no inheritance, the process proceeds to Step 38. If it is determined that there is inheritance and there is no ambiguity, the semantic role determination processing (Step 3)
Proceed to 7).

On the other hand, if the verb-derived noun is a polynomial noun and cannot be determined with respect to inheritance even if it is compared with the inheritance determination table (FIG. 5) (for example, the above-mentioned “assign
ment "), that is, if there is a possibility of inheritance, the process proceeds to the next polysemy elimination process (Step 35).

FIG. 6 shows an example of the number information / modifier collation table. In the polysemy elimination processing (Step 35), the polysemy elimination of the verb derived noun is referred to by referring to the number information / modifier collation table. Perform By this processing, the subject noun inherits the noun structure of the verb “complex event no
If it is determined to be “minal” (Step 36), the process proceeds to the next semantic role determination process (Step 37).

FIG. 7 shows an example of a term structure correspondence dictionary. In the semantic role determination processing (Step 37), the meaning role of a phrase modifying a main noun is determined by referring to the term structure correspondence dictionary. To determine. At this point, the semantic analysis processing for the read noun phrase ends. Next, it is checked whether another noun phrase exists in the original sentence (Step 3).
8). If there is another noun phrase, the process returns to Step 31; otherwise, the process ends.

Here, "The constant assignment of u
An example of translating the English sentence “nsolvable problems led to disaster.” into Japanese will be described as an example of the semantic analysis processing of the noun phrase in the embodiment of the present invention. ), The structure tree (FIG. 8) is selected as the optimal solution, and the processing shifts to the semantic analysis processing of the noun phrase (FIG. 3).

When the structure tree is searched for a structure corresponding to the noun phrase (N) from the beginning of the sentence, first, "The constant assignm
ent of unsolvable problems ”(Step 3
1). Therefore, the main noun of this noun phrase “assignmen
It is checked whether or not "t" is a verb-derived noun. When the word dictionary 4a is searched and checked, it is found that the assignment in FIG.
al, ”and“ complex event nominal ”. Therefore, it is not possible to make a clear judgment on the inheritance even if it is compared with the inheritance judgment table (FIG. 5). (Step 3
3).

Next, by referring to the numerical information / modifier collation table (FIG. 6), a process for eliminating polysemy is performed. This requires that the word “assignment” is singular and that the article that precedes it is a definite article,
tant ”modifiers (aspectual modifie
r), and with the postfix modifier of the prepositional phrase,
“t” is determined to be “complex event nominal” (Steps 35 and 36).

Then, the semantic role of the phrase modifying the main noun is determined by referring to the term structure correspondence dictionary (FIG. 7) (Step 37). As a result, "of unsolvabl"
e problems ”is given a semantic role of“ object case ”. At this point, the semantic analysis processing for the read noun phrase ends. Next, the structure tree is searched to determine whether another noun phrase exists (Step 38). Then "di
The noun phrase "saster" is found, and this noun phrase is read. However, when the word dictionary 4a is searched, it is found that this noun is not a verb derivative noun, and the subsequent processing is not performed. It is checked whether another noun phrase exists, but there is no such phrase. At this point, the noun phrase semantic analysis process ends.

When the semantic analysis of the noun phrase is completed,
By executing the generation processing, the noun phrase “the cons
"Tant assignment of unsolvable problems" translates to "giving unsolvable problems constantly." this is,
"No unsolvable problem"
This is because “ms” is given the semantic role of “target case” (of the verb “assign”).

[0024]

According to the first aspect of the present invention, an input unit for inputting an original sentence to be translated, an original sentence storage unit for storing the original sentence input from the input unit, and a translation process of the original sentence A translation dictionary unit containing knowledge information to be used for translation, a translation storage unit that stores translations obtained as a result of the translation processing of the original text, a display unit that displays the original text and the translation text, and a display unit that displays the translation text. In the machine translation device having an editing control unit that performs editing processing of the original sentence or the translated sentence, the translation unit includes a noun phrase analysis processing unit that performs semantic analysis of a noun phrase, so that a translated sentence with excellent understandability can be obtained. Obtainable.

According to the second aspect of the present invention, in addition to the effect of the first aspect, the noun phrase analysis processing unit, when the main noun of the noun phrase is a verb-derived noun, Since the semantic analysis of the noun phrase is performed by checking the inheritance of the noun phrase required by the verb corresponding to the noun phrase to the noun phrase, the polysemy of the noun phrase can be eliminated.

According to the third aspect of the present invention, in addition to the effect of the second aspect of the present invention, the noun phrase analysis processing section includes the number information of the noun phrases and the number information of the phrases modifying the main noun. And semantic analysis using structural and semantic information,
The semantic role of the phrase that modifies the main noun of the noun phrase can be determined.

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram of a machine translation apparatus to which the present invention is applied.

FIG. 2 is a flowchart of a translation process according to the present invention.

FIG. 3 is a flowchart of a semantic analysis of a noun phrase according to the present invention.

FIG. 4 is a diagram illustrating an example of a word dictionary used by the machine translation device to which the present invention is applied.

FIG. 5 is a diagram showing an example of a succession determination table used in a machine translation apparatus to which the present invention is applied.

FIG. 6 is a diagram showing an example of a numerical information / modifier collation table used in a machine translation apparatus to which the present invention is applied.

FIG. 7 is a diagram showing an example of a term structure correspondence dictionary used in a machine translation apparatus to which the present invention is applied.

FIG. 8 is a diagram showing an example of a structure tree selected as an optimal solution used in a machine translation apparatus to which the present invention is applied.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... input part, 2 ... original sentence storage part, 3 ... translation part, 4 ... translation dictionary part, 4a ... word dictionary, 4b ... grammar rule dictionary, 4c ... item structure correspondence dictionary, 5 ... translation sentence storage part, 6 ... display part , 7 ... Edit control unit.

Claims (3)

[Claims] An input unit for inputting an original sentence to be translated, an original sentence storage unit for storing the original sentence input from the input unit, and a translation containing knowledge information used for translation processing of the original sentence A dictionary unit, a translation storage unit that stores a translation obtained as a result of the translation process of the original text, a display unit that displays the original text and the translation text, and an editing process of the original text or the translation text displayed on the display unit Wherein the translation unit has a noun phrase analysis processing unit that performs a semantic analysis of the noun phrase. 2. The noun phrase analysis processing unit, when the main noun of the noun phrase is a verb derivative noun, examines the inheritance of the term configuration required by the verb corresponding to the verb derivative noun to the noun phrase. 2. The machine translation apparatus according to claim 1, wherein a semantic analysis of the noun phrase is performed. 3. The method according to claim 1, wherein the noun phrase analysis processing unit performs a semantic analysis using the number information of the noun phrases, the number information of the phrases modifying the main noun, and the structural / semantic information. 3. The machine translation device according to claim 2, wherein