JPH05108705A - Machine translation device - Google Patents

Abstract

(57) [Summary] [Purpose] When translating from a source language into a partner language, it is more accurate by using a knowledge base that collects the relationships between words to distinguish between words based on polysemy. Moreover, it aims at performing translation processing efficiently. When the source language character string input by the input means 1 is translated into a partner language and it is necessary to perform transcribing based on the polysemy of words, the polysemy resolution unit 6 makes the knowledge base 5 Refer to the data that describes the relations between words that appear in ordinary sentences to resolve polysemy and perform translation.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a machine translation device for translating a source language character string into a partner language character string.

[0002]

2. Description of the Related Art In recent years, there has been an increasing demand for natural language processing technology as represented by machine translation.

When considering translation between two languages, resolution of polysemy becomes a big problem. For example, when thinking about translating between Japanese and English, when trying to translate "solve a problem" and "unravel a package" into English, in the above example, the word "solve" is so
lve, and untie is suitable in the example below.
In order to perform such translation based on the polysemy of "solve," the object of the verb (in this case, "solve") (in this case, "problem" and "luggage") is semantically classified. Every
When a certain semantic category is entered, a certain translated word is applied, and when another semantic category is entered, another translated word is applied.

[0004]

In the method of eliminating polysemy by means of semantic classification and selecting an appropriate word as described above, it is necessary to classify meanings for a very large number of words, and inconsistencies. It may be difficult to create a criterion for the semantic classification so that the above does not occur. Moreover, such semantic classification does not eliminate all ambiguities.

[0005]

[Means for Solving the Problems] In order to solve the above problems, consider using a knowledge base that collects co-occurrence relationships between words. In other words, information about co-occurrence of words (for example, “solve a problem”, “unravel a package”, etc.) is extracted from the sentence that is actually used, and the corresponding partner Expression in language (in this example, "s
"olve a problem", "untie a"
A large amount of data including "package" and the like) is prepared. When a problem of polysemy occurs, appropriate translation can be performed by referring to the stored data. Then, with respect to the polysemy of “solve”, it can be seen that if the sentence “solve a problem”, solve is appropriate as a translation.

[0006]

When a translation between two languages is performed, a knowledge base that collects the relations between words is prepared and used for differentiating according to the polysemy of words, so that efficient translation processing can be performed. become able to.

[0007]

DESCRIPTION OF THE PREFERRED EMBODIMENTS A description will be given below with reference to the drawings.

FIG. 1 is a functional block diagram of a machine translation device according to an embodiment of the present invention. In the figure, reference numeral 1 is an input means for inputting a character string in the source language. Reference numeral 2 denotes an input character string storage unit that stores a character string in the source language input from the input means 1.
Reference numeral 3 is a translation dictionary unit having information about each language and bilingual translation information, and the like, and 4 is a dictionary search unit that searches a translation dictionary for a word to be analyzed. Reference numeral 5 is a knowledge base unit that collects relations between words in the source language and their translations, and 6 is a ambiguity resolution unit that uses the knowledge base 5 to perform translation according to the polysemy of the words. Reference numeral 7 is an analysis result output unit that outputs an analysis result, and 8 is a translation processing control unit that controls translation processing.

FIG. 2 is a flow chart showing the flow of processing in one embodiment of the present invention. A Japanese-English translation will be taken up as an embodiment of the present invention, and the operation of the present apparatus will be described with reference to this flowchart.

First, in step 1, the translation processing control unit 8 searches the translation dictionary unit 3 for the input Japanese character string by the dictionary search unit 4 using a known search method to obtain information about words. And proceed with the translation process.

Next, in step 2, in the process of translation processing,
Determine whether there is a need for translation due to the polysemy of a word. If there is a place to be translated, the process of step 3 is performed, and if not, the process proceeds to step 7 as it is.

In step 3, the disambiguation unit 6 is used for translating the word in question in the knowledge data described in the knowledge base unit 5 of FIG. Determine if there is something you can do. If there is no applicable data here, the process directly proceeds to step 5.

If there is applicable knowledge, ambiguousness resolution processing is performed in step 4. Next, in step 6, it is judged whether or not the processing for disambiguation has been completed. If completed, the processing proceeds to step 7, and if not completed, the processing in step 3 is repeated.

In step 5, the translation processing control unit 8 determines one translated word by using information such as the frequency of use.

Further, in step 6, the remaining processing is performed, and in the next step 7, the translation result is output to the translation result output unit 7.

The actual processing will be described in more detail with reference to specific examples. Consider the following example sentence. "He solved the problem yesterday." First, in step 1, the dictionary search unit 4 searches the translation dictionary unit 3 word by word to proceed with the translation process.
In this example, as shown in FIG. 3, it is necessary to separately translate the "solved" portion due to polysemy (step 2).

Next, in step 3, the data regarding "solve" are searched in the knowledge base section of FIG. Here, since knowledge data about "solving" is found, step 4
Then, the ambiguity resolution unit 6 uses the data to perform the ambiguity resolution processing of this portion. In this case, there are two data for “solve” and “solve luggage” for “solve”, and the input sentence corresponds to “solve the problem”, so “solve”
It is determined that solve is appropriate as a parallel translation of. Here, if there is no appropriate knowledge data, one of them is selected according to the information such as the frequency of use as described above (step 5). Next, in step 6, it is determined whether or not there is another portion that needs to be disambiguated. In this example, since there is nothing else, the process proceeds to the next step 7, and the remaining processing is performed.

In the next step, the final result of the translation is output to the translation result output unit 7. In this example, the result shown in FIG. 5 is output.

Although the operation of disambiguation has been described by taking a Japanese-English translation as an example, it can be similarly performed in the case of English-Japanese. As in the example of take in FIG. 3, the object is “act.
Although the translation varies depending on "ion" and "an examination", the translated word can be selected using the knowledge data even in such a case.

[0020]

As described above, in translating between two languages, a knowledge base that collects relations between words is prepared and used for transcribing according to the polysemy of words.
It becomes possible to perform efficient translation processing.

[Brief description of drawings]

FIG. 1 is a functional block diagram of a machine translation device according to an embodiment of the present invention.

FIG. 2 is a flowchart showing an operation of a translation process in the embodiment of the present invention.

FIG. 3 is a diagram showing an example in which polysemy occurs in one embodiment of the present invention.

FIG. 4 is a diagram showing an example of knowledge base data according to an embodiment of the present invention.

FIG. 5 is a diagram showing a result of disambiguation according to an embodiment of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Input means 2 Input character string storage unit 3 Translation dictionary unit 4 Dictionary search unit 5 Knowledge base unit 6 Ambiguity resolution unit 7 Analysis result output unit 8 Translation processing control unit

Claims (1)

[Claims] 1. An input means for inputting a source language character string, an input character string storage section for storing the input character string, a morpheme information of a source language word and a partner language, a bilingual information of a source language and a partner language, and the like. A translation dictionary unit that holds the translation dictionary, a dictionary search unit that searches the translation dictionary, a co-occurrence relationship between words in the source language, and a knowledge base unit that collects corresponding expressions in the other language, and the knowledge base. Section, the ambiguity resolution unit that resolves the ambiguity that occurs when translating an input character string into the partner language, the translation result output unit that outputs the translation processing result, and the translation process between the source language and the partner language. A machine translation device comprising a translation process execution control unit for controlling.