JPS60200368A - Mechanical translation system - Google Patents

Abstract

PURPOSE:To improve the probability of production of a correct sentence using a target language and to enable even a user having no knowledge of the target language to use a mechanical translation system, by providing a meaning mark processing part, a meaning relation mark processing part and a concept structure processing part. CONSTITUTION:When a sentence of a language to be translated is fed to a subject language input part 8, a word dividing part 9 decomposes the sentence into words while referring to a dictionary file 6. A meaning mark processing part 10 obtains the meaning marks corresponding to each word, and a meaning relation mark processing part 11 discriminates the relation among meaning marks and delivers the meaning relation marks. A concept structure processing part 12 produces a meaning network of the language to be translated from the output of the part 11 and performs the concept conversion into a meaning network of a target language to produce a sentence of the target language through a target language producing part 13. In case the translation fails in this process, the procedure proceeds to a concept structure designating input mode of a user. Thus the probability of afilure is reduced for translation and even a user having no knowledge of the target language can use easily this mechanical interpretation system.

Description

DETAILED DESCRIPTION OF THE INVENTION (1) Technical Field of the Invention The present invention relates to a machine translation method that analyzes an input sentence, develops it into a conceptual structure called a semantic network, and then translates it into another language.

(2) Prior art and problems Conventionally, machine translation devices using semantic networks (conceptual structures) analyze an input sentence (target language) to create a conceptual structure corresponding to the target language, and then create a conceptual structure corresponding to the target language. is converted into a conceptual structure corresponding to the target language (translated language), and sentences in the target language are generated based on this. In such conventional machine translation systems, all of the above-mentioned processes are not always successful in generating correct sentences in the target language, and the success rate is not necessarily high.

Therefore, it is usually necessary to inspect the generated target language and make corrections, but this has the disadvantage that this cannot be done smoothly if the user of the machine translation system is not familiar with the target language. Ta.

(3) Purpose of the Invention In view of the above-mentioned conventional drawbacks, the present invention has a high probability of successfully generating correct sentences in the target language, and even if the user lacks knowledge of the target language, as long as the user understands the target language. The purpose is to provide a machine translation device that is easy to use.

(4) Structure of the invention According to the present invention, the object is to decompose a sentence in a language to be translated into words, and to extract a meaning symbol corresponding to each word and each meaning symbol. In a translation system for translating a language to be translated into a target language based on a conceptual structure based on semantic relational symbols indicating relationships between the two, a means for accepting the semantic symbols and semantic relational symbols selected by the user is provided. This is achieved by a machine translation method characterized in that the conceptual structure of a sentence consisting of a language to be translated is determined based on the selected semantic symbols and semantic relationship symbols, and a sentence in a foreign language is generated from the conceptual structure. .

(5) Embodiment of the Invention FIG. 1 is a block diagram showing an example of a device for carrying out the present invention, in which 1 is a display device, 2 is a keyboard, 3 is a light pen, 4 is a control device, and 5 is a memory. , 6 is a dictionary file,
7 represents a printer.

2 is a block diagram illustrating the control device 4 in more detail, in which 8 is a target language input section, 9 is a word division processing section, 10 is a semantic symbol processing section, 11 is a semantic relation symbol processing section,
12 is a conceptual structure (semantic network) processing section, 13 is a target language generation section, and 14 is a selection input section.

When a sentence in the language to be translated is input to the target language input section 8 of the apparatus shown in FIGS. 1 and 2, the word division processing section 9 decomposes the sentence into words while referring to the dictionary file 6. The semantic symbol processing section 10 determines the meaning symbol corresponding to the word, and the semantic relation symbol processing section 11 determines the relation between the semantic symbols and outputs the semantic relation symbol indicating the relation. The conceptual structure processing unit 12 generates a semantic network of the language to be translated from the semantic symbols and semantic relationship symbols, conceptually converts the semantic network into a semantic network of the target language, and the target language generating unit 13 generates a meaning network of the target language. Generate sentences in the target language based on the network.

If the translation fails in the above process, the semantic symbol processing unit 10 and the semantic relation symbol processing unit 11 shift to a conceptual structure designation input mode by the user, and display the screen shown in FIG. 3 on the display device 1.

In the figure, numerals 15 to 18 represent display sections, where the display section 15 shows the original text, the display section 16 shows the semantic symbols, the display section 17 shows the semantic relationship symbols, and the display section 18 shows the semantic network. This shows a state in which the progress is displayed.

As shown in the figure, in this example, the original text is r The
first impression of sitti
A case is shown in which ngis rather hard J is input. Then, the control device 4 finds the meaning symbols corresponding to the words in the original text and displays them in the display section 16, and the meaning-related symbols (including additional symbols) in the display section 17.

When the user selects a meaning symbol from the display section 9 with the light pen 3, the selection input section 14 selects a meaning symbol from the display section 1.
8 (in the figure, HARD is selected).

Then, the display direction of the arc is specified and a rightward arc is displayed. (This corresponds to the line segment with the arrow in the display section 18 in Fig. 2.) Then, use the light pen 3 to specify one of the semantic symbols displayed in the display section 17 for this arc. , gives a semantic relation symbol to the arc. (<OBJ><ST> is given in the figure)
Furthermore, a semantic symbol that becomes a node connected to the arc (<OBJ>) is selected.

By performing these operations one after another, a semantic network diagram as shown in FIG. 4 is finally formed in the display section 18 of FIG. 3.

Then, when the semantic network is completed, keyboard 2
In response to a user's command from , the concept S construction processing unit 12 uses the semantic network as input and converts it into the conceptual structure of the object word, and also searches the dictionary file 6.
A matching object word or phrase is selected, a sentence using the object word is generated, and the sentence is output to the display device 1 or printer. In the unlikely event that the translation fails, a message to that effect is output, and the user displays the semantic network related to the original text on the display device 1 again, corrects it using the same procedure as when creating it, and then translates it again. All you need to do is give instructions.

Adding some additional information to the conceptual structure (semantic network), semantic symbols, and meaning-related symbols explained above, a conceptual structure expresses the meaning of a sentence in the form of a semantic network, and the nodes that make up the network (
A semantic symbol) represents a concept, and an arc (semantic relationship symbol) represents a relationship between concepts. The relationships between concepts represented by rightward arcs are 〈CAUSE), (SEQUENC
Relationships that connect sentences such as E) and (A G E
There are deep case relationships such as NT), (OBJECT>, and (GOAL).There are also special arcs that have no partner node, and this represents additional semantic information for the node.

In addition, the meaning symbols corresponding to each word in the original text are selected from the dictionary file, but some words do not have meaning symbols.

As already explained, “The firat im-p
rsssion of sitting is rat
The semantic symbol generated from the sentence herhard J is [
5ITJ, rFIR8T IMPRESSIONJ, “
HARD”.

Semantic relationship symbols serve as arcs of a semantic network, and some arcs are marked with a square symbol. Some examples of semantic relationship symbols are shown in Table 1, and some examples of additional symbols are shown in Table 2.

Table 1 Table 2 (6) Effects of the Invention As explained in detail above, the value of the machine translation method of the present invention is that the method generates sentences in the target language by allowing the user to directly specify the conceptual structure of the target language. Therefore, the probability of translation failure is small, and it can be used even by users without knowledge of the target language, so it is highly effective.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an example of a device for carrying out the present invention, FIG. 2 is a block diagram showing details of the control device shown in FIG. 1, FIG. 4, and FIG. 3 is a diagram showing an example of a display screen. FIG. 3 is a diagram showing a semantic network. DESCRIPTION OF SYMBOLS 1...Display device, 2...Keyboard, 3...Light Ben, 4...Control device, 5...Memory, 6...
Dictionary file, 7... Printer, 8... Target language input section, 9... Word division processing section, 1o... Semantic symbol processing section, 11... Semantic relation symbol processing section, 12... - Conceptual structure processing unit, 13...Target language generation unit, 14...Selection input unit, 15-18...Example of display screen. Representative Patent Attorney Hiroshi Matsuoka! Figure 2 Figure 3 @ 1, Risa 4 times

Claims (1)

[Claims] A sentence in the language to be translated is broken down into words, a conceptual structure is determined from the semantic symbols corresponding to each word and the semantic relationship symbols indicating the relationship between each semantic symbol, and the language to be translated is determined based on the conceptual structure. In a translation system that translates into the target language,
A means is provided to accept the meaning symbols and meaning-related symbols selected by the user, a conceptual structure of a sentence in the language to be translated is determined based on the selected meaning symbols and meaning-related symbols, and a sentence in another language is created from the conceptual structure. A machine translation method that is characterized by generating .