JPS63255773A - Mechanical translation processing system - Google Patents

Abstract

PURPOSE:To increase the processing speed by extracting the same sentence elements from results of morpheme analysis and generating data of relations between the same elements to avoid repetition of the same processing. CONSTITUTION:A morpheme analyzing means 1 refers to a word dictionary to divide an input sentence into morphemes and analyzes connection relations between morphemes. A same element extracting means 3 extracts the same sentence elements from results of morpheme analysis in the morpheme analyzing means 1 to generate data 4 of relations between the same elements. In accordance with results of morpheme analysis, a syntactic analysis means 2 determines how sentence elements are connected, which syntactic functions the sentence elements fulfill, and how sentence elements constitute the sentence. In this case, data 4 of relations between the same elements generated by the same element extracting means 3 is used and preceding analysis results of the same elements are used.

Description

[Detailed Description of the Invention] [Summary] In machine translation, it is possible to perform morphological analysis of multiple sentences (
The system is structured so that the same sentence elements are extracted and the subsequent processing for them is standardized, thereby streamlining the processing and improving the processing speed.

FIELD OF INDUSTRIAL APPLICATION] The present invention relates to machine translation, and more particularly to a control method for extracting the same part of an original text and standardizing the processing of that part in a machine translation system using a computer.

[Prior Art] Conventional machine translation systems usually translate one sentence at a time.

That is, when the processing (analysis, transformation, generation) of one sentence is completed, the processing of the next sentence is started from the beginning.

Machine translation is usually performed through processing steps as shown in FIG.

The input sentence is first subjected to morphological analysis. This divides the input sentence into morphemes (words) by referring to a word dictionary, and analyzes the conjunctive relationships between the morphemes.

Next, based on the results of morphological analysis, syntactic analysis is performed to determine how the elements of the sentence are connected to each other, what syntactic roles they play, and how the sentence is constructed.

Next, from the results of the syntactic analysis, a semantic analysis is performed to extract a semantic structure with conceptual symbols and auxiliary symbols as nodes.

Next, the semantic structure is converted into an output language, and syntax is generated using the output language.

Morphological generation is performed from the result of syntax generation, and an output sentence is generated and output.

Semantic analysis may not be treated as an independent processing step, but may be included in syntactic analysis to form syntactic/semantic analysis. There is also a system that generates a conceptual structure based on the results of syntactic/semantic analysis using an intermediate representation that does not depend on the input/output language, and converts this into an output language.

In conventional machine translation, translation processing is performed one sentence at a time, and even if a phrase that is the same as a phrase that appeared in the previous sentence that has already been translated appears as a result of morphological analysis, the syntactic analysis for this phrase is not performed. It was something that was done repeatedly.

[Problem to be solved by the invention] Some sentences have similar structures, such as expressions using couplets or bullet points.

Conventional machine translation systems do not perform processing that spans multiple sentences, so the same processing is repeated even for the same portion.

The present invention aims to solve the problem of repetition of the same processing in conventional machine translation systems and to provide a machine translation side that improves processing speed.

Means for Solving Problem E] FIG. 1 shows a schematic diagram of the principle of the machine translation control system of the present invention.

In the figure, reference numeral 1 denotes a morphological analysis means, which divides an input sentence into morphemes (words) with reference to a word dictionary, and analyzes the connection relationships between the morphemes.

Reference numeral 3 denotes identical element extraction means, which extracts identical sentence elements from the results of morphological analysis by the morphological analysis means 1, and creates relationship data 4 between identical elements.

2 is a syntactic analysis means, which determines how the elements of a sentence are connected to each other, what syntactic roles they play, and how the sentence is constructed, based on the results of morphological analysis. At that time, the same element relationship data 4 created by the same element extracting means 3 is used, and the previous analysis result of the same element is utilized.

"Operation" Reads the input sentence, performs morphological analysis on each sentence, and saves the results.

At that time, mark the words that mark the breaks in sentences, such as conjunctions and separated parts, and store the words from one break to the other as one unit, along with the number of the sentence in which it appears and its position in the sentence. do.

When processing has progressed for a certain number of sentences, the same element extracting means 3 checks whether the previously stored elements are the same.

If the same elements are found, the same element relationship data 4 is created. This is marked with, for example, "The same element appears after r" or "The same element appeared before."

Specifically, for the entire part for which morphological analysis has been completed,
Starting from the last element, check whether the same element has appeared before, and if so, mark 1 for it and mark 2 for the one that appears later. Later, if the same element appears earlier, mark l for it, and mark 1 for it and mark 1 for it (
Change the mark 1 (first marked 1) to mark 2. Mark 1 means "J the same element appears later, mark 2 means the same unit appeared before r".

After that, parse the syntax from the beginning. At this time, the processing results of the elements marked with mark 1 are saved, and the stored processing results are used for the units marked with mark 2.

Simple pointer operations can reduce the effort required to reuse processing results to a negligible level.

According to this processing method, the same sentence unit (phrase, clause)
The processing time required for translation can be reduced because the processing can be performed in one process.

In the above processing method, "morphological analysis" may be replaced with "syntactic analysis j. Furthermore, the above processing may be performed at both the morphological analysis and syntactic analysis stages.

[Example] The present invention will be described in more detail below with reference to Examples shown in FIGS. 2 to 8.

FIG. 2 is an example input sentence for explaining processing according to an embodiment of the present invention.

FIG. 3 shows the division result of the input example sentence of FIG. 2 by morphological analysis.

FIG. 4 shows the results of the morphological analysis shown in FIG. 3 organized into clusters of clauses (independent words and ten adjunctive words).

The divided units in FIG. 4 are processing units according to this embodiment.

FIG. 5 is a diagram showing the configuration of a control block that shows the relationship between the same clauses, and shows processing for the morphological analysis result of FIG. 4.

If you look for phrases that appear earlier in sentence (2) in Figure 4, you will find the phrase ``Sora wo (not fly)''.
Since it is used in statement (11), create a control block for this.

The meaning of the control block is as follows.

PTRI: Pointer to the control block corresponding to the previously used location, or 0 (not previously used, indicating this is the first).

PTR2: Pointer to the control block corresponding to the next used location, or 0 (indicating the next least used, ie, this is the last).

PTL? 3: Pointer indicating the group of interest.

In the control block corresponding to the sentence (1), ``I will not fly in the sky'', PTRI is 0, PTR2 is ・,
In the control block corresponding to the sentence (2), r does not fly, PTRI is -, and PTR2 is O.

Similarly, “
A control block is also created for "Doesn't fly".

FIG. 6 is a flowchart showing the process of extracting the same portion in this embodiment.

In this manner, a chain of control blocks as shown in FIG. 5 is completed.

Next, a parsing process is performed.

FIG. 7 is a diagram showing the analysis processing status of the same part in this embodiment. That is, after finishing the analysis of sentence (1), sentence (2)
This shows the situation at the time the analysis started.

In the analysis of sentence (1), the analysis result of "I won't fly in the sky" is a verb clause called FLY, as shown in the rectangular frame in the figure, and LOG (location) is SKY (sky).
This indicates NOT (negation).

When the analysis of sentence (1) is completed, the pointer PTR3 of the control block for "Do not fly" is updated to point to the analysis result of "Do not fly".

In the analysis of sentence (2), “fly not” in sentence (2)
Since a control block has been created for , the control block pointed to by the pointer PTRI is referenced to obtain the analysis result of statement (1), ``Do not fly.'' The obtained analysis result is used as the analysis result of sentence (2).

The pointer PTR3 of the control program for ``J does not fly'' in sentence (2) may be made to point to the analysis result of ``J does not fly,'' as in sentence (11), or may be left as is. In the latter case, if the same part (``not flying'') appears in a later sentence, the chain of control blocks is returned to the beginning and the analysis result of sentence (11) is used.

FIG. 8 is a flowchart showing the process of saving and using analysis results in this embodiment.

[Effects of the Invention] As explained above, according to the present invention, in a machine translation system using a computer, it is no longer necessary to analyze the same element repeatedly, and the processing is made more efficient and the processing speed is improved. The effect is great.

[Brief explanation of the drawing]

Figure 1 is a block diagram of the principle of the present invention. Figure 2 is a diagram showing input example sentences. Figure 3 is a diagram showing the results of morphological analysis. Figure 4 is a diagram showing the results of morphological analysis summarized into phrases. , No. 5 is a system according to an embodiment of the present invention? Figure 6 is a flowchart showing the same part extraction process; Figure 7 is a diagram showing the analysis processing status of the same part; Figure 8 is a flowchart showing the process of saving and using analysis results; The figure is a diagram showing the stages of machine translation processing. In the drawings, 1 is a morphological analysis means, 2 is a syntactic analysis means, 3 is an identical element extraction means, and 4 is relationship data between identical elements. Input text Principle block diagram of the invention Figure 1 fil Fish do not fly. (2) Even if a carp is said to become a dragon, it cannot fly. (3) However, flying fish can fly above the sea. (4) How does it fly over the sea? +11
Fish do not fly in the air. Figure 3 fil Fish do not fly in the sky. (Even though I say I will be 21 years old, I will not fly in the sky. Halfway through the loaf:

Claims (2)

[Claims] (1) Morphological analysis means (1) that divides an input sentence into morphemes with reference to a word dictionary and analyzes conjunctive relationships between morphemes;
In a machine translation system equipped with a syntactic analysis means (2) that analyzes syntactic relationships between sentence elements from the results of morphological analysis, from the results of analysis of multiple sentences by the morphological analysis means (1),
It is equipped with a same element extraction means (3) that extracts the same sentence elements and creates relationship data (4) between the same elements, and after performing morphological analysis on multiple sentences, standardizes the syntactic analysis processing for the same sentence elements. A machine translation processing method characterized in that the machine translation processing method is configured to perform the processing using the following methods. (2) The same element extraction means (3) described above is a parsing means (
The same sentence elements are extracted from the analysis results of multiple sentences using 2), and the feature is that after syntactic analysis of multiple sentences, subsequent processing for the same sentence elements is performed in common. A machine translation processing method according to claim 1.