JPH04296969A - Mechanical translation device - Google Patents

Abstract

PURPOSE:To attain the effective partial translation of an original sentence when the analysis of the original sentence fails by having such a constitution of a symbol inserting means where the symbols are put into the divided original sentences if an input original sentence could not be totally analyzed. CONSTITUTION:An input original sentence is sent to 8 translation module 14 under the control of a main CPU 10. The module 14 translates the original sentence With use of a dictionary, the grammatical. rules, the tree structure conversion rules, etc., stored in a dictionary/grammar memory 15. Then the module 14 prints out the translated sentence through a printer 16. A means is added to divide the input original sentence into plural parts together with a symbol inserting means which puts the punctuation symbols into the divided areas of the original sentence, and a translation means which translates those divided parts of the original sentence based on the added punctuation symbols. Thus the original sentence is divided into parts and translated when the analysis of the original sentence fails. For this purpose, the punctuation symbols are put into the original sentence.

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.

0002

2. Description of the Related Art Conventional machine translation devices have conventionally used a backtracking method (also referred to as backtracking) parsing using a context-free grammar.

[0003] The above-mentioned backtrack method means that when there are multiple possibilities for syntactic analysis, those multiple possibilities are tried sequentially, and if the analysis reaches a dead end at a certain point, another possibility exists from that point on. This is a method of backtracking to a certain point and trying another candidate.

[0004] A machine translation device using the above syntax analysis is
The input sentence (original text) written in the source language entered from the keyboard is input into the translation module under the control of the main CPU (central processing unit), and the text is translated into the input source language. is configured to be translated into a target language by a translation module using a dictionary, grammar rules, and tree structure conversion rules stored in a memory.

[0005] The specific structure of the conventional machine translation device described above is disclosed in Japanese Patent Application Laid-Open No. 63-8864 and Japanese Patent Application Laid-Open No. 1-11387.
It is stated in No. 0.

[0006]

[Problems to be Solved by the Invention] In the conventional machine translation device described above, when analysis of the entire original text fails, the original text is divided into small parts and translated, such as translating the original text for each predetermined phrase. However, it is difficult to distinguish between phrases in the translated text that correspond to phrases in the original text, and the correspondence between the original text and the translated text is difficult to understand, which hinders post-editing and pre-editing.

[0007] Furthermore, since it is difficult to determine which part of the original text is failing in analysis, it is difficult to know which part should be improved when improving the performance of a machine translation device, and it is difficult to improve performance. The problem is that it is inefficient.

[0008] In view of the above-mentioned problems of conventional machine translation devices, the present invention provides a machine translation device that is capable of efficiently translating parts of an original text when analysis of the original text fails and the original text is translated part by part. .

[0009]

[Means for Solving the Problems] The present invention includes a dividing means for dividing an input original text into a plurality of parts, a symbol inserting means for inserting a symbol indicating a division break into a division part of the divided original text, The apparatus includes a translation means for translating each of the plurality of divided parts based on the symbol indicating the division break applied to the original text, and the symbol insertion means inserts the symbol when the input original text cannot be analyzed as a whole sentence. This is achieved by a machine translation device that is configured to insert into the segmented original text.

0010

[Operation] The dividing means divides the input original text into multiple parts, and the symbol insertion means inserts a symbol indicating the division break at the dividing point of the divided original text when the input original text cannot be analyzed as a whole sentence. Then, the translation means translates each of the plurality of divided parts based on the symbol indicating the division break applied to the original text.

[0011]

Embodiments Hereinafter, embodiments of the machine translation apparatus of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a flowchart showing the operation of the machine translation apparatus of the present invention.

FIG. 2 is a block diagram showing the configuration of an embodiment of the machine translation apparatus of the present invention.

The machine translation device shown in FIG. 2 includes a main CPU (central processing unit) 10, a main memory 11, a CRT (cathode ray tube) 12, a keyboard 13, a translation module 14, a dictionary for translation, grammar rules, and trees. It is comprised of a dictionary/grammar rule memory 15 that stores structure conversion rules, etc., and a printer 16 that prints the translation results on a sheet such as paper.

The translation module 14 is configured to, when an input sentence (original sentence) written in the source language is input, translate the original sentence and output an output sentence (translated sentence) written in the target language. ing.

The input original text is controlled by the main CPU 10 and sent to the translation module 14. The translation module 14 uses the dictionary, grammar rules, tree structure conversion rules, etc. stored in the dictionary/grammar rule memory 15 to translate the input original text by the method described later, and sends the obtained translated text to the printer 16. Print and output.

FIG. 3 shows a specific configuration of the translation module 14 of this embodiment.

Next, the operation of the translation module 14 will be explained using FIGS. 2 and 3.

First, the dictionary lookup morphological analysis unit 18 reads the original sentence and divides it into each morpheme string (word string) using the dictionary in the dictionary/grammar rule memory 15, and extracts grammatical information such as part of speech and translation for each word. as well as tense/person/
Analyze numbers, etc.

The syntactic analysis unit 19 determines a structural analysis tree indicating the dependency relationship between each word based on the dictionary and grammar rules in the dictionary/grammar rule memory 15.

The conversion unit 20 has a dictionary/grammar rule memory 15.
The structure of the parse tree for the input original text is converted into the structure of the parse tree for the translated text using the tree structure conversion rules. The translated sentence generation unit 21 attaches appropriate particles and auxiliary verbs as a translated sentence in the target language, outputs the obtained translated sentence from the translation module 14, and outputs it from the printer 16.

Next, ``processing for inserting division symbols into the original text'' will be described, in which when the original text fails to be analyzed, the original text is divided into small parts, translated, and symbols indicating the breaks in the original text are inserted.

In this embodiment, it is assumed that only the following grammar rules are included.

(a) Sentence → Declarative sentence + End of sentence (b) Declarative sentence → Noun phrase + Verb phrase (c) Noun phrase → Noun (d) Verb phrase → Verb + Noun phrase (e) Verb phrase → Verb (f) Sentence → Declarative sentence (for division) (g)
Sentence → Conjunction (for division) Grammar rule (a) above indicates that a "sentence" consists of a "declarative sentence" and "sentence final", and grammar rule (b) indicates that a "declarative sentence" consists of a "noun phrase" and a "verb". ``phrase'', and grammar rule (c) is ``phrase''.
Indicates that a ``noun phrase'' consists of ``nouns,'' and uses the grammatical rule (d
) indicates that a ``verb phrase'' consists of a ``verb'' and a ``noun phrase.''

[0025] Also, the above grammar rule (e) is a "verb phrase"
The grammar rule (f) is a rule that is only used during split translation (when the split flag is on) and indicates that the "sentence" consists of a "declarative sentence".
g) is a rule that is used only during split translation (when the split flag is on) and indicates that a "sentence" is composed of "conjunctions". The division flag is turned on when analysis of the original text fails, and is cleared when processing starts.

Next, application of the above rules by the machine translation device of FIG. 1 will be explained using example sentences.

First, a case where "Mark ate apples." is input as a first example sentence to the machine translation apparatus shown in FIG. 1 will be described.

Perform the dictionary lookup morphological analysis process shown in FIG. 1 (step S1), clear the buffer at the original text break position (step S2), and reset the division flag (step S3).
By doing so, the part-of-speech string shown in Table 1 is obtained.

[0029]

[Table 1]

[0030] Then, the above grammar rules (c), (c)
, (d), (b), and (a) in this order to perform syntactic analysis (step S4), and check whether or not the parsing was successful as a "sentence" in step S4 (step S5). If the syntax analysis is not successful in step S5, the division flag is turned on (step S6), and the process returns to step S4 to analyze the syntax again.

On the other hand, if the syntax analysis is successful in step S5, the parsed syntax is converted (step S7) and then generated (step S8).

Next, it is checked whether the division flag is on or not (step S9). If the division flag remains off, the process proceeds to step S10 of FIG. ” is displayed on the CRT.

Next, as a second example sentence, “Mark sa
id John ate apples and Ma
ryate oranges. ” is input.

As in the case of the first example sentence described above, the part-of-speech sequence shown in Table 2 is obtained by the dictionary lookup morphological analysis process (step S1) shown in FIG.

[0035]

[Table 2]

[0036] This original sentence should originally be translated as "Mark said that John ate the apple and Mary ate the orange." However, the built-in grammar is, for example, (h). Verb phrase → verb + declarative sentence (i) Declarative sentence → declarative sentence + conjunction + declarative sentence Because the analysis of the entire sentence fails due to the lack of grammatical rules, it is necessary to divide the original sentence into small parts. Translate from.

That is, step S in the flowchart of FIG.
In step 6, the division flag indicating that parsing of the entire sentence has failed and that segmental translation is to be performed is set on, so that the grammar rules for segmentation can also be used.

That is, steps S4 to S1 in FIG.
By loop 3, “Mark said” (grammar rules (c), (e), (b), (f) are applied in this order), “John ate apples” (grammar rules (c), (c), (d) , (b) , (f
) in this order), “and” (grammatical rules (g
), “Mary ate oranges.
”(Grammar rules (c), (c), (d), (b)
, (a) in this order) are analyzed. Now,
The above analyzed word positions "1", "4", "5", ("
9" is omitted because it is the end of the sentence) is set in the original text break position buffer (step S11).

Next, the translation is set in the translation buffer (step S12), and it is checked whether or not there are any unprocessed words (step S13). When there are no unprocessed words in step S13, the table is As shown in 3, a delimiter "‖" is inserted immediately after the word position of the original text set in the original text delimiter position buffer.

[0040]

[Table 3]

[0041] Through the above processing, once the division symbol "‖" is inserted into the original text, it is possible to translate the entire sentence without failure in analysis.

In that case, it is necessary to incorporate the following grammar in order to process the division symbol "‖".

(j) Sentence → subtree + sentence end (k) subtree → declarative sentence (l) subtree → conjunction (m) subtree → subtree + ‖+ subtree grammar (
Analyze and translate the input original text by applying the grammar rules a) to (g) and (j) to (m). FIG. 4 shows an example of the analytic tree obtained.

[0044]

[Effects of the Invention] According to the machine translation device of the present invention, there is a dividing means for dividing an input original text into a plurality of parts, and a symbol inserting means for inserting a symbol indicating a dividing line into a dividing part of the divided original text. and a translation means for translating each of the plurality of divided parts based on the symbol indicating the division break applied to the original text, and the symbol insertion means is used when the input original text cannot be analyzed as a whole sentence. It is configured to insert symbols into the divided original text, so when the original text fails to be analyzed and the original text is translated part by part, the parts of the original text that were successfully parsed and the parts that failed are inserted into the original text. This can be easily determined based on the correspondence between the delimiter symbol and the delimiter symbol in the translated text, and as a result, translation can be performed efficiently.

[Brief explanation of the drawing]

FIG. 1 is a flowchart showing the operation of a machine translation device of the present invention.

FIG. 2 is a diagram showing the configuration of an embodiment of the machine translation device of the present invention, showing the operation of FIG. 1;

FIG. 3 is a diagram showing the configuration of the translation module in FIG. 2;

FIG. 4 is a diagram illustrating an example of the structure of an analysis tree produced by the machine translation device of FIG. 2;

[Explanation of symbols]

10 Main CPU 11 Main memory 12 CRT 13 Keyboard 14 Translation module 15 Dictionary/grammar rule memory 16 Printer

Claims (1)

[Claims] 1. A dividing means for dividing an input original text into a plurality of parts; a symbol inserting means for inserting a symbol indicating a division break into a division part of the divided original text; and translation means for translating each of the plurality of divided parts based on a symbol indicating a delimiter, and the symbol insertion means replaces the symbol with the symbol when the input original text cannot be analyzed as a whole sentence. A machine translation device characterized in that it is configured to be inserted into a divided original text.