JPH0636168B2 - Machine translation processor - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Field of Application] The present invention makes it possible to search for terms in two or more fields with arbitrary priority without sacrificing access efficiency in a machine translation processing apparatus. The present invention relates to a machine translation processing device.

[Conventional technology and its problems]

 FIG. 3 shows an explanatory view of the prior art.

When performing automatic translation using an electronic computer, various methods have been considered for selecting which translated word when one original word has a plurality of translated words having different meanings. As one method, in general, the meaning of a term is often specified for each field related to the original sentence, and it is possible to prepare a plurality of dictionaries for each field, for example, as shown in FIG. Has been done.

For example, the word "reactor" is often used to mean "reactor" in the field of nuclear power and "reactor" in the field of chemistry. in this case,
Appropriate translated words for each field are registered in advance in the nuclear field dictionary 3A, the chemical field dictionary 3B, and the more general standard dictionary 3C. Then, the translation processing unit 1 notifies the dictionary search unit 2 of the dictionary types to be searched in accordance with the priority order, and specifies the priority order of the dictionary to be searched in translation. Specify as a headword. If “nuclear power” is designated as the first priority and then the field of “chemistry” is designated, the dictionary search unit 2 searches the nuclear field dictionary 3A for each headword, and translates it. If it does not exist, the chemistry field dictionary 3B is searched, and if it does not exist, the standard dictionary 3C is searched.

According to this method, it is possible to use terms in various fields on the same system and change the priority order among the fields. However, since the terms of each field are stored in separate dictionaries and these are logically linked and accessed as necessary, for example, when the number of dictionary files is n, the time required for dictionary access Since the standard dictionary is usually searched in many cases, the number is approximately n times as large as when there is only one dictionary file.

In order to solve this problem, for example, if one priority dictionary is determined and an integrated dictionary is created according to the priority, the dictionary cannot be used for other priorities and the system There is a problem that the flexibility is lost and the maintenance and management of the dictionary file becomes complicated.

[Means for solving problems]

The present invention solves the above problems and provides a machine translation processing device that enables access to terms in various fields without sacrificing the efficiency of dictionary access. Therefore, the machine translation processing device of the present invention is provided with a dictionary that stores an original word and its translated word in association with each other, and in the machine translation processing device that searches the dictionary and translates the input original sentence, A translation table that stores each source word as a headword in association with a translation word in a field in which each source word is used and a field code that identifies the field is configured to be registered in association with the information of the headword, and The translation table having the same headword is configured to be collected and stored in one physical record that can be read by one input / output operation, and represents a plurality of fields to be the reference for translation selection in translation. A field table that stores field codes together with their respective priorities, and a dictionary reading unit that reads out the translation table related to the headwords with the source word to be translated as a headword from the dictionary. A translation word selection unit that sequentially searches the translation table read by the dictionary reading unit based on the field code according to the priority order stored in the field table, and selects the translation word when the translation word corresponding to the field code exists It is characterized by having and.

〔Example〕

FIG. 1 shows the configuration of an embodiment of the present invention, and FIG. 2 shows the process explanatory diagram of the translated word selection unit shown in FIG.

In the figure, 1 is a translation processing unit that translates an input original sentence by parsing and semantic analysis, 2 is a dictionary search unit that accesses a dictionary according to an instruction from the translation processing unit 1, and 3 is a standard term and a special for each field. An integrated dictionary in which terms and words are integrated into one dictionary, 4 and 5 are input devices, 6 is an input unit for inputting field priority information and original text to be translated, 7 is a field table according to a specified field priority A field table creating unit, 8 is a field table that stores information about specified field priorities, 9 is a dictionary reading unit that reads a translation table from the integrated dictionary 3 according to a headword specified by the translation processing unit 1, and 10 is a reading unit. Translation table buffer that stores the issued translation table, 11
Is a translation word selection unit that selects a translation word based on the translation word table and the field table and notifies the translation processing unit 1; 12 is an output unit that outputs the translation result; 13 is an output device such as a line printer, a display, or an external storage device; Reference numeral 20 represents headword information, and 21 represents a translation word table.

In the case of the present invention, a field code for identifying the field is assigned to each field in advance. Then, in the dictionary, for example, as shown as an integrated dictionary 3 in FIG. 1, a translation table 21 in which field codes and translations in the field indicated by the field code are paired and stored corresponding to each headword information 20. Is set.
The type and range of field may be arbitrarily determined according to the scale of the device.

For example, in the dictionary for Japanese to English translation, when the entry word is "frequency", the translation table 21 includes
As shown in Fig. 1, the translation of "frequency" corresponds to the field code of "communication", the translation of "frequency" corresponds to the field code of "chemistry", and the translation of "frequency" as standard. be registered. The order in which the field codes appear in the translation table 21 does not necessarily have to be special. By setting the terms in various fields in each translation table 21 in this way, the integrated dictionary 3 in the present invention does not have a different file structure for each field, but is a single integrated file. . It should be noted that it does not matter at all that the index word information 20 and the pointer information to the translated word table 21 corresponding thereto are provided as separate files as index files.

In this embodiment, the translation word table 21 having the same headword in the integrated dictionary 3 is collected and stored in one physical record. That is, in the case of the present embodiment, each translated word table 21 is set so as not to span a plurality of physical records. Therefore, at least one whole translation table 21 is read into the translation table buffer 10 by one input / output request.

In translating, the field table creation unit 7 creates the field table 8 in accordance with the field priority information input from the input device 5. The field priority information may be sequentially input in a menu method for inquiries, or may be input together with the input original text. It may also be registered in the system in advance. Field table 8
In the field, a specified number of field codes are stored in a specified order.

When the input original sentence is sent to the translation processing unit 1 via the input unit 6, the translation processing unit 1 performs syntax analysis and the like, and activates the dictionary search unit 2 with the extracted word as a headword. Note that the main processing in the translation processing unit 1 may be the same as that of the conventional technique, so a detailed internal description is omitted.

In response to the activation from the translation processing unit 1, the dictionary searching unit 2 uses the dictionary reading unit 9 based on the headword specified.
The integrated dictionary 3 is accessed and the corresponding translated word table 21 is read into the translated word table buffer 10. The translated word selection unit 11 refers to the translated word table 21 read out to the translated word table buffer 10 and executes the processing shown in FIG. 2 to select an appropriate translated word.

That is, the translated word selection unit 11 first initializes the index value N to the field table 8 to "1" by the processing 30 shown in FIG. Then, by the process 31, the translation word table 21 is searched to check whether or not the Nth field code in the field table 8 is in the translation word table 21.

If it is determined by the determination in the process 32 that the field code exists, the process proceeds to the process 33, the corresponding translated word is obtained from the translated word table 21, and the translation processing unit 1 is notified of it. If the first field code does not exist, in process 34, “1” is added to the index value N, and in process 35,
The index value N is the size M of the field table 8, that is, the field table 8
It is determined whether it is larger or smaller than the number of field codes stored in. If the index value N is not larger than the size M, the process returns to step 31 and similarly the translation table 21 is searched for the next field code.

In this way, the processes 32 to 35 are repeated, and when the field code stored in the field table 8 is not found in the translation table 21, the process 36 selects a general standard translation registered regardless of the specialized field. Then, it reports it to the translation processing unit 1.

The translation processing unit 1 obtains translated words by sequentially designating headwords and calling the dictionary search unit 2 as described above, and assembles these translated words to generate a translated sentence. The generated translated sentence is output to the output device 13 via the output unit 12.

The present invention may be configured as a processor dedicated to automatic translation, or may be realized on a so-called general-purpose computer. The field code can be defined using English letters, for example, "EL" in the field of electrical engineering and "PH" in the field of physics. And, for example, the translated word "frequency" is "frequency" in the field of electrical engineering,
Even in the field of optical engineering, when it is common like "frequency", by storing the translated words for each field regardless of the common translated words, it is possible to construct a dictionary that is resistant to revisions and changes. On the other hand, the field code is set to a numerical value such as "64" in the field of electrical engineering and "16" in the field of optical engineering so that the duplicated bits do not become "1" when converted to binary numbers. Oki, "64" and "16" for translated words common to electrical engineering and optical engineering
It is also possible to set a composite field code of "80", which is a combination of and, in the translated word table 21 to save the translated word area.

〔The invention's effect〕

As described above, according to the present invention, terms in multiple fields can be referred to at the same time when translation is executed, and priorities of fields can be freely changed when translation is executed. Therefore, it is possible to give flexibility to the target field of translation and to guarantee the same level of access time as in the case of one dictionary file containing only one field. Further, since one dictionary is integrated, there is also an effect that it is easy to create and maintain the dictionary.

Further, since the translation table having the same headword is collected in one physical record and stored in the dictionary regardless of the translation field, the translation table can be referenced very quickly even when there are many fields. There is also an effect that can be done.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention, FIG. 2 is an explanatory view of the processing of a translation word selection unit shown in FIG. 1, and FIG. In the figure, 1 is a translation processing unit, 2 is a dictionary search unit, 3 is an integrated dictionary, 8 is a field table, 9 is a dictionary reading unit, 11 is a translation word selection unit, and 21 is a translation word table.

Claims (1)

[Claims] 1. A machine translation processing apparatus comprising a dictionary for storing an original word and its translated word in association with each other and searching the dictionary to translate an input original sentence, wherein each of the original words to be translated is used as a headword in the dictionary. A translation word table that stores a translation word in a field in which each of the original words is used and a field code for identifying the field in association with each other is configured to be registered in association with the information of the above entry word, and has the same entry word. The translation table is configured to be collected and stored in one physical record that can be read by one input / output operation, and when translating, field codes that represent a plurality of fields to be used as a reference for selecting a translation word are stored. A field table to be stored together with the priority order, a dictionary reading unit for reading out the translation table related to the headword from the dictionary with the original word to be translated as a headword, and stored in the field table. A translation word selection unit that sequentially searches the translation table read by the dictionary reading unit based on the field code according to the priority order and selects the translation word when the translation word corresponding to the field code exists. Machine translation processor.