JP3985066B2 - Machine translation device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a machine translation apparatus that translates a source language into a target language.
[0002]
[Prior art]
In a machine translation device, it is important not only to be able to translate correctly, but also to be able to translate as desired by the user. In conventional machine translation technology, only part-of-speech terms and unique translation patterns can be registered in the dictionary. It is difficult to register wordings and expressions in parts of speech. A translation pattern may be able to replace a numerical value or a specific term in the original text as a variable, but basically the original text and the translated text have a one-to-one correspondence. In order to cover a wide range of expressions, a large number of patterns must be prepared, and it takes much time to create and search for a translated sentence pattern, which is not efficient.
For example, in Japanese Patent Laid-Open No. 5-290082, by providing a translation pattern that can be created using a variable and a nested structure, it is possible to provide a translation pattern that can be easily created by a user and that can be searched efficiently. A device that can be used is proposed.
[0003]
[Problems to be solved by the invention]
However, the above prior art describes that English and German can be translated into Japanese, but it is redundant to create a combination of source language pattern and target language pattern for each language to be translated. In addition, when the source language pattern and the target language pattern are not in one-to-one correspondence, there is a problem that it cannot be simply created.
[0004]
For example, to translate Japanese into English, we want to translate “will improve” to “will improve”, so the source language pattern is Japanese, the target language pattern is English, and “improve” “improve” ”And“ will be planned ”, and“ will improve ”is also translated as“ will improve ”, so“ will be ”and“ will ... ”are also created.
Thus, in order to translate English into Japanese, it is redundant to re-create the source language pattern in English and the target language pattern into Japanese, and “will…” "It will be ~" corresponds to both, so you must delete one.
[0005]
[Means for Solving the Problems]
The present invention employs the following configuration in order to solve the above-described problems.
<Configuration 1>
A machine translation apparatus including a bilingual dictionary in which a first language pattern and a second language pattern, which are composed of mutually different languages and have a bilingual relationship, are stored in correspondence with each other. A translation direction code for permitting bi-directional translation to the language pattern and the second language pattern or restricting translation in any one direction is attached, and the input source language sentence is the first and second Judgment is made between languages, and if a translation direction code permitted by the parallel translation is added to the language pattern constituting the source language sentence by referring to the bilingual dictionary, the corresponding language pattern of the other language is selected, and the parallel translation is selected. A machine translation device comprising translation means for deselecting a language pattern of another language corresponding to a translation direction code regulated by
[0006]
<Configuration 2>
In the translation dictionary, a variable that can be replaced with respect to the first language pattern composed of one word is attached, and a part of the first language pattern composed of a plurality of words is replaced with a variable. 2. The machine translation apparatus according to claim 1, further comprising conversion means for replacing the first language pattern constituting the source language sentence composed of any of the languages described above with a variable to convert it into a nested source language pattern combination sentence.
[0007]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail using specific examples.
"Concrete example"
<Constitution>
FIG. 1 is a block diagram showing a specific example of the machine translation apparatus of the present invention.
The apparatus shown in the figure includes a selection unit 10, an input unit 11, a pattern dictionary 12, a conversion unit 13, a development unit 14, and an output unit 15.
[0008]
The machine translation device is realized by a personal computer, for example. The selection means 10 is a means for selecting a source language and a target language to be translated, and has a function as a user interface that accepts display to the user and designation of a translation direction from the user. And hardware such as a processor and a memory for executing this. The input means 11 is means for inputting the original text, and is composed of an interface for taking in the original text from a recording medium such as a CD-ROM or a floppy disk, a network interface, or a keyboard for directly inputting the original text. .
The pattern dictionary 12 is a bilingual dictionary indicating a source language pattern, a target language pattern corresponding to the source language pattern, a translation direction of the source language pattern and the target language pattern, and variables, and is provided in a magnetic disk device or the like. ing.
[0009]
FIG. 2 is an explanatory diagram illustrating an example of the pattern dictionary 12.
As shown in the figure, the pattern dictionary 12 is a dictionary composed of variables, source language patterns, translation directions, and target language patterns. In this specific example, the source language is English and the target language is Japanese. At least one of each language pattern has a nested structure that can be included in another language pattern as a corresponding variable. For example, an English pattern (it) can be included in another English pattern (improve OBJ) as a variable (OBJ). The translation direction is the direction to convert the language pattern, => from left to right (English to Japanese: Note that this pattern is omitted in the example shown), <= from right to left (Japanese to English) , = Is bidirectional.
[0010]
Based on the translation direction selected by the selection unit 10, the conversion unit 13 refers to the source dictionary input from the input unit 11, refers to the pattern dictionary 12, and replaces the corresponding source language pattern with a variable. The input sentence including the variable is searched for whether it corresponds to the source language pattern in the pattern dictionary 12, and by repeating this process, the input sentence is converted into a nested source language pattern in which all of the input sentences are replaced with variables. It is. The expansion unit 14 is a unit that expands the sentence replaced by the variable by the conversion unit 13 using the target language pattern of the pattern dictionary 12 and obtains a translation result of the target language. The converting means 13 and the expanding means 14 constitute a translation means 16. The output unit 15 is a functional unit for outputting the translation result obtained by the expansion unit 14.
The conversion unit 13 and the expansion unit 14 are realized by software corresponding to the function of each unit and hardware such as a processor or a memory that executes the software. The output unit 15 includes a display for displaying the translation result, a printer for printing, an output interface for outputting the translation data, and the like.
[0011]
<Operation>
Hereinafter, the operation of this example will be described using an actual input sentence as an example.
FIG. 3 is a flowchart showing the translation processing of this example.
First, the language to be translated is selected by the selection means 10 (step S1). In this example, we choose to translate English into Japanese.
[0012]
Next, the original text is input to the input means 11 (step S2). Here, “I will improve it.” Is input as an example of English text. The conversion means 13 refers to the pattern dictionary 12 to search for an English pattern corresponding to a part or the whole of the original sentence (step S3), and converts the original sentence into an English pattern so that no variable remains (step S4). Here, in the pattern dictionary 12, only the pattern whose translation direction is from English to Japanese => or = is searched and converted. In the pattern dictionary 12 shown in FIG. 2, (1) (2) (4) (5) (6) corresponds. Then, the conversion means 13 converts the original text into an English pattern so that no variable remains.
“I will improve it.”
→ (5) SUBJ will improve (6) OBJ.
→ (5) SUBJ will (4) VERB.
→ (5) SUBJ (2) VERB.
→ (1) SENTENCE
still,
“I will improve it.”
→ (5) SUBJ will improve (6) SUBJ. For example, this combination is not adopted because a portion that cannot be converted into a variable such as improve SUBJ remains.
As a result, the conversion result is [[I (5)] [will [improve [it (6)] (4)] (2)]. (1)].
[0013]
Next, the converted English pattern is converted into a Japanese pattern and developed (step S5).
[[I (5)] [will [improve [it (6)] (4)] (2)]. (1)]
→ [[I (5)] [will [improve [it] (4)] (2)]. (1)]
→ [[I (5)] [will [Improve it] (2)]. (1)]
→ [[I (5)] [I plan to improve it]. (1)]
→ [[I] [I plan to improve it]. (1)]
→ [I plan to improve it. ]
The output means 15 outputs “I will improve it” as the translation result (step S6).
[0014]
Next, the case where the direction which wants to translate is reverse is demonstrated.
The selection means 10 selects to translate Japanese into English (step S1), and inputs "I plan to improve it" to the input means 11. The conversion means 13 refers to the pattern dictionary 12 to search for a Japanese pattern whose translation direction is from Japanese to English <= or =, and converts the original text into a Japanese pattern so that no variable remains ( Steps S3 and S4). In the illustrated pattern dictionary 12, (1), (2), (4), (5), and (6) correspond, so that the same conversion as described above is performed as follows.
“I plan to improve it.”
→ (5) SUBJ will improve (6) OBJ.
→ (5) SUBJ is scheduled for (4) VERB.
→ (5) SUBJ is (2) VERB.
→ (1) SENTENCE
still,
“I plan to improve it.”
→ (5) SUBJ will improve (6) SUBJ. In this case, for example, a portion that cannot be converted into a variable such as [Improve SUBJ] remains, so this combination is not adopted.
As a result, the conversion result is [[I (5)] will improve [[[It (6)] (4)] (2)]. (1)].
[0015]
The converted Japanese pattern is converted into an English pattern in the same manner as in the above example, and the pattern is developed (step S5).
[[I (5)] is going to improve [[[It (6)] (4)] (2)]. (1)]
→ [[I (5)] is going to improve [[it] (4)] (2)]. (1)]
→ [[I (5)] is [[improve it] (2)]. (1)]
→ [[I (5)] is [will improve it]. (1)]
→ [I will improve it] (1)]
→ [I will improve it.]
The output means 15 outputs “I will improve it.” As a translation result (step S6).
[0016]
In addition, when "I will improve it" as an input sentence, it becomes as follows.
The conversion means 13 refers to the pattern dictionary 12 to search for a Japanese pattern whose translation direction is from Japanese to English and is converted into a Japanese pattern so that no variable remains ( Steps S3 and S4). In the illustrated pattern dictionary 12, (1), (3), (4), (5), and (6) correspond, so the following conversion is performed.
“I will improve it.”
→ (5) SUBJ will improve (6) OBJ.
→ (5) SUBJ will be (4) VERB.
→ (5) SUBJ is (3) VERB.
→ (1) SENTENCE
As a result, the conversion result is [[I (5)] will improve [[[6]] (4)] (3)]. (1)].
[0017]
The converted Japanese pattern is converted into an English pattern in the same manner as in the above example, and the pattern is developed (step S5).
[[I (5)] will [[[It (6)] improve (4)] (3)]. (1)]
→ [[I (5)] will [Improve [[it] (4)] (3)]. (1)]
→ [[I (5)] would be [[improve it] (3)]. (1)]
→ [[I (5)] is [will improve it]. (1)]
→ [I will improve it] (1)]
→ [I will improve it.]
The output means 15 outputs “I will improve it.” As a translation result (step S6).
[0018]
<effect>
As described above, according to the specific example, since the direction to be translated is added to the pattern of the pattern dictionary 12, it is not necessary to create a pattern for each language to be translated, and the dictionary can be created even if it is not one-to-one correspondence. There is an effect that can be done. As a result, the capacity of the dictionary as a translation dictionary can be reduced.
[0019]
In the above specific example, the information on the translation direction is applied to the pattern dictionary having variables. However, the translation means includes a bilingual dictionary indicating the target language pattern corresponding to the source language pattern, and the translation means translates using the bilingual dictionary. Any machine translation apparatus that performs the same can be applied as in the specific example.
[Brief description of the drawings]
FIG. 1 is a configuration diagram showing a specific example of a machine translation apparatus of the present invention.
FIG. 2 is an explanatory diagram showing an example of a pattern dictionary of the machine translation apparatus of the present invention.
FIG. 3 is a flowchart showing the operation of the machine translation apparatus of the present invention.
[Explanation of symbols]
10 Selection means 12 Pattern dictionary (Bilingual dictionary)
13 Conversion means 14 Expansion means 16 Translation means

Claims (2)

A machine translation device comprising a bilingual dictionary in which a first language pattern and a second language pattern, which are composed of different languages and have a bilingual relationship, are stored correspondingly,
The bilingual dictionary is given a translation direction code that permits bi-directional bi-directional translation to the corresponding first language pattern and second language pattern or restricts bi-directional translation in either direction,
It is determined whether the input source language sentence is the first language or the second language, and the translation direction code permitted by the parallel translation is added to the language pattern constituting the source language sentence with reference to the parallel translation dictionary. A translation unit that selects a language pattern corresponding to another language when attached and includes a translation unit that deselects the language pattern corresponding to another language when the translation direction code that restricts parallel translation is attached;
A machine translation device characterized by that. In the translation dictionary, a variable that can be substituted for a first language pattern that consists of one word is attached, and a part of the first language pattern that consists of a plurality of words is substituted with the variable,
The translating means has conversion means for converting the source language sentence composed of the first language into a source language pattern combination sentence having a nested structure by replacing the first language pattern constituting the source language sentence with the variable. The machine translation apparatus according to claim 1, wherein the machine translation apparatus is a machine translation device.

Images