JPH01166258A - Mechanical translation device - Google Patents

Abstract

PURPOSE:To interpret with higher accuracy whether an original sentence is equal to a sentence or a title and to ensure correct mechanical translation by dividing the original sentence into morpheme strings and extracting corresponding translated words. CONSTITUTION:A read original sentence is divided into morpheme strings by a dictionary consulting morpheme analyzing par 11 via a dictionary of a memory 6. Grammar information and the translated word are obtained for each word and tense, person, numeral, etc., are analyzed. A syntax structure analyzing part 12 decides a syntax structure analyzing tree showing the active/passive relation between words according to grammar rules and by reference to the dictionary of the memory 6. A converting par 13 converts the structure of the syntax structure analyzing tree corresponding to the input original sentence into the structure of the syntax structure analyzing tree corresponding to the translated sentence based on the tree structure converting rule of the memory 6. Then a translated sentence producing part 14 adds proper postpositional words and auxiliary verbs for production of a translated sentence of a target language. The translated sentence is supplied to a main memory 2 and a CRT 3 respectively via a translation module 5.

Description

DETAILED DESCRIPTION OF THE INVENTION <Industrial Application Field> The present invention relates to a machine translation device that can determine whether an input word string is a sentence or a title and translate it.

<Prior Art> Conventionally, there are the following machine translation devices. This machine translation device inputs the source language input from the keyboard into the translation module under the control of the main CPtJ, and the translation module transfers the input source language to the dictionary, grammar rules and book structure stored in the memory. It uses conversion rules to translate into the target language. The above grammar rules apply to the input sentence (hereinafter,
It is used to determine the syntactic structure of the original text.

<Problems to be Solved by the Invention> However, when translating with the above-mentioned conventional machine translation device, the syntactic structure of the original text is determined according to the grammatical rules stored in memory, and the Since it is not possible to determine whether the original text is a sentence or a title (generally a noun phrase), even though the input original text is a title, the grammar rules for normal sentences stored in the memory are The problem is that when applied, the original text is often misinterpreted. Additionally, since the first letter of each word in the word string representing the title is usually capitalized, when looking up a dictionary, each word in the title is not matched with a registered word in the dictionary, and the above words are treated as proper nouns. There is a problem in that the word is determined to be an unknown word, making it difficult to identify the part of speech of the word, and erroneous analysis is likely to occur.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a translation device that can determine whether an original text is a sentence or a title, and can more accurately interpret the original text based on the result of the determination.

<Means for Solving the Problems> In order to achieve the above object, the present invention includes a morphological analysis unit that divides an input sentence into morphemes using a dictionary and obtains information such as parts of speech, and the morphological analysis unit described above. a syntactic analysis unit that analyzes the syntax of the morpheme sequence divided by using a dictionary and grammar rules, a conversion unit that converts the syntactic structure obtained by the syntactic analysis unit into a syntactic structure in the target language, and the conversion unit A machine translation device having a translated sentence generation section that generates a translated sentence according to the syntactic structure of the target language obtained by the section, a capital letter discrimination means for determining whether or not the beginning of an input morpheme is a capital letter; a memory means for storing a morpheme determined by the means to have a capital letter at the beginning; and a memory means for storing a morpheme whose first character is a capital letter; title identifying means for identifying that the morpheme stored in the means is a title;
When the title specifying means specifies that the morpheme stored in the memory means is a title, the morpheme stored in the memory means is provided with an upper case/lower case converting means for converting the first capital letter of the morpheme into a lower case letter, and the morpheme is changed to a title. It is characterized by being able to perform translations according to the user's needs.

<Operation> The input sentence is divided into morphemes by the morpheme analysis section, and the capitalization determining means determines whether the beginning of the morpheme is a capital letter. As a result, all morphemes determined to have a capital letter at the beginning are stored in the memory means. If the morpheme following the morpheme stored in the memory means has the final form rsT: and is not a period or a question mark, the title specifying means determines that the morpheme stored in the memory means is a title. be identified. Then, the first capital letter of the form BT: stored in the memory means is converted to a lower case letter by the case/lower case conversion means. Thereafter, information such as the part of speech of the form B element converted to lowercase letters is obtained using a dictionary.

Therefore, it is possible to determine whether or not the input sentence is a title, and if it is a title, to perform translation according to the title.

<Example> Hereinafter, the present invention will be explained in detail with reference to illustrated embodiments.

FIG. 1 is a block diagram of an embodiment of a machine translation apparatus according to the present invention. l is the main CPU (central processing unit),
2 is a main memory, 3 is a CRT (Cathode Ray Duplex), 4 is a keyboard, 5 is a translation module, and 6 is a memory that stores a dictionary 1 for translation, grammatical rules, tree structure conversion rules, etc.

When the source language is input, the translation module 5
It translates it and outputs the target language.

That is, the source language input from the keyboard 4 is sent to the translation module 5 under the control of the main CPUI. The translation module 5 uses the dictionary 9 grammar questions and tree structure conversion rules stored in the memory 6 to translate the input source language into the target language as will be detailed later. The results are temporarily stored in the main memory 2 and displayed on the CRT 3.

FIG. 2 shows a specific configuration of the translation module 5 of this embodiment.

The operation of the translation module will be explained below using FIGS. 1 and 2.

First, the read original text is divided into each morpheme string (word string) by the dictionary reference morphological analysis unit II using the dictionary in the memory 6, and grammatical information such as part of speech and translation for each word are obtained, and further, Tense, person, number, etc. are analyzed. In the syntactic analysis unit 12, in accordance with the dictionary in the memory 6 and the grammatical rules, the dependencies between each word are determined as shown in FIG. 5 by creating a structural analysis tree indicating the relationship as will be described later.

The conversion unit I3 uses the tree structure conversion rules in the memory 6 to convert the structure of the parse tree for the input original text into the structure of the parse tree for the translated text. Then, the translated sentence generation unit 14 adds particles and auxiliary verbs suitable for the translated sentence in the target language, and the obtained translated sentence is output from the translation module 5, stored in the main memory 2, and displayed on the CRT 3. be done.

Next, the dictionary lookup morphological analysis unit 1 of the translation module 5
In operation 1, title determination processing for determining whether an input sentence is a title will be described in detail. Here, it is assumed that the memory 6 stores only the grammar rules shown in Table 1.

Table 1 The grammar rules in Table 1 above represent the following:

That is, grammar rule (a) indicates that a noun phrase is made up of nouns and nouns. Grammar rule (b) indicates that a noun phrase is made up of nouns, nouns, and nouns. Grammar rule (C) indicates that verb phrases are made up of verbs. Grammar rule (d) indicates that a declarative sentence is made up of a noun phrase and a verb phrase. Grammar rule (e) indicates that a sentence is made up of a declarative sentence and a sentence ending. The grammar rule (``) is
Indicates that a sentence is made up of a noun phrase and the end of the sentence.

FIG. 3 is a flowchart of title discrimination processing according to this embodiment. The operation of the title discrimination process will be explained below with reference to FIG.

In step S1, a pointer is set to the first morpheme.

In step S2, it is determined whether the beginning of the morpheme (word) to which the pointer was set in step S1 is a capital letter. As a result, if the character is an uppercase character, the process proceeds to step S3; otherwise, the process proceeds to step SIO.

In step S3, words with initial capital letters are stored in the buffer.

In step S4, the pointer is set to the next p4 element.

In step S5, it is determined whether the morpheme to which the pointer is set is a morpheme representing the end.

If the morpheme is the final morpheme, the process advances to step S6, otherwise the process returns to step S2.

In step S6, it is determined whether the morpheme determined to be the terminal in step S5 is a period or a question mark. If the result is a period or question mark, proceed to step SIO,
Otherwise, the process proceeds to step S7.

In step S7, step S5. The terminal morpheme determined to be not a period or a question mark in step S6 is stored in the buffer.

In step S8, the morpheme stored in the buffer is identified as a title, and a title flag is set.

In step S9, the first uppercase letter of each word stored in the buffer is converted to lowercase letter.

In step SIO, a dictionary lookup is performed according to words whose first letters are lowercase letters to obtain information such as part of speech.

In step Stt, it is determined whether or not the process is finished. If the process is finished, the title specifying operation is finished, and if not, the process proceeds to step S12.

In step S12, the pointer is set to the next shape spasticity.

The title translation processing operation based on the above-mentioned title identification processing will be explained in more detail using a specific input sentence.

Here, as an input sentence, (a) r Surface Tempera
turcs Rise J(b) r S ur
race temperatures rise,
Assume that J is input.

When the input sentence (a) is input, since the beginning of each word is a capital letter, step S2. of the flowchart of FIG. S
3. S4. By S5 each word r S surface r T emperatures J and r Rise
J is stored in the buffer.

, then 'rRiseJ is followed by a space, so
Step S5. At S6, the word string already stored in the buffer is recognized as a title.

Then, in step S8, a title flag is set, and in step S9, the first capital letters S, T, and rL of each word stored in the buffer are converted to lowercase letters. Thereafter, in step S10, a dictionary lookup is performed using the word whose first letter is converted to lower case. As a result, the obtained part-of-speech information is stored in the dictionary lookup result buffer within the translation module 5, as shown in FIG. At this time, in order to facilitate the processing of the dictionary lookup morphological analysis unit 11, a sentence ending that does not exist in the input sentence (a) is added to the buffer.

Next, the syntax analysis section! 2 is the content of the dictionary lookup result buffer shown in Figure 4 and the grammar rules (a) and (a) shown in Table 1.
c), (d) and (e), the syntax is analyzed as shown in FIG. 5(a). Further, the horizontal text analysis unit 12 checks the title flag set in step S8 of FIG. Then, since the title flag is set, the input sentence (a) is determined to be a title, and the lateral sentence analysis unit 12 fails in the lateral sentence analysis using the verb phrase, as shown in Figure 5 (ra). let As a result, the parser 1
In 2, backtracking works, and the part of speech of rriseJ stored in the dictionary lookup result buffer shown in Figure 4 is changed to the second part of speech "noun", and the grammar rule is changed to (b) of the grammar rule shown in Table 1. and (r) by changing
Syntax analysis is performed as shown in FIG. 5(b). Then, since the structure of the resulting syntax is a noun phrase, the syntactic analysis is successful, and the source language parse tree shown in FIG. 5(b) is determined. Thereafter, the conversion unit 13. The process moves to the translated sentence generation section 14, the translation result of the input sentence (a) is outputted as shown in FIG. 6(a), and the title translation process ends.

On the other hand, when input sentence (b) is input, the word r S ur
Since the beginning of faceJ is a capital letter, step S2. of the flowchart of FIG. Word r S u by S3
rfccJ is stored in the buffer. next word rt
Since the beginning of empires J is a lowercase letter, the word r teIIlperature is entered in step S2.
Since sJ is not determined to be a title and step S8 is skipped, the title flag is not set. Then, the word r SurfaceJ stored in the buffer
At the same time, a dictionary lookup is performed in step SIO. Furthermore, step S2. A dictionary lookup for the next word rrIseJ is performed at SIO. As a result, the syntax analysis section 12 determines a syntax analysis tree as shown in FIG. 5(a). Conversion processing and translated sentence generation processing are then performed based on this parse tree. The translation result is shown in Figure 6(b).
The translation process ends with output as shown in .

In this way, by using the fact that a title is generally a noun phrase and that the first letter of each word in the word string representing the title is usually capitalized, we can determine whether the input morpheme starts with a capital letter or not.
By determining whether the final morpheme is a period or a question mark, it is determined whether the input original text is a title or a sentence. Therefore, based on the discrimination results, if it is a title, the first capital letter of each word is converted to lower case, and translation according to the title can be performed. without mistranslating or incorrectly parsing the words in the title.
Can always translate correctly.

<Effects of the invention> As is clear from the above, this invention has the form @search solution Il'
A machine translation device including an r section, a syntax analysis section, a conversion section, and a translated sentence generation section is provided with an upper case discrimination means, a memory means 1 title identification means, and an uppercase/lower case conversion means. The morpheme determined to be a title is stored in the memory means, and when the title identifying means identifies the morpheme stored in the memory means as a title, translation according to the title can be executed. Even if a sentence is input that contains a mixture of ``and'' normal sentences, the syntax can be parsed accurately, making highly accurate translation possible.

Furthermore, when translating a title, the case converting means converts the first capital letter of the morpheme stored in the memory means to a lower case letter, so that the word starting with a capital letter in the word string is treated as a registered word in the dictionary. Matsushi can serve,
Highly accurate translation can be performed without the words in the title being identified as incorrect parts of speech or being treated as proper nouns.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of an embodiment of the machine translation device of the present invention, Fig. 2 is a functional block diagram of the translation module in the above embodiment, Fig. 3 is a flowchart of title discrimination processing according to the above embodiment, and the lower 4 figures is a diagram showing the contents of the dictionary lookup buffer, and Figure 5 is a diagram showing an example of the determined parse tree.
FIG. 6 is a diagram showing an example of an input sentence to the machine translation device of FIG. 1 and an example of an output sentence (translated sentence) therefor. 1... Main Or'U, 2... Main memory, 3... CRT, 4... Keyboard,
5...Translation module, 6...Memory, ■...
Dictionary guide! @Elementary analysis unit, 12...Syntax analysis unit, 13...Conversion unit, 14.
...Translation generation unit. Figure 4 Figure 6 Figure 5 (a) Bu (b) Fu

Claims (1)

[Claims] (1) A morphological analysis unit that divides the input sentence into morphemes using a dictionary to obtain information such as parts of speech, and analyzes the syntax of the morpheme string divided by the morphological analysis unit using a dictionary and grammar rules. a syntactic analysis unit; a conversion unit that converts the syntactic structure obtained by the syntactic analysis unit into a syntactic structure in the target language; and a translation unit that generates a translated sentence according to the syntactic structure of the target language obtained by the conversion unit. A machine translation device having a sentence generation unit, comprising: an uppercase character determination means for determining whether or not the beginning of an input morpheme is an uppercase letter; and a memory means for storing a morpheme determined to have an uppercase character at the beginning by the uppercase character determination means. , title specifying means for specifying that the morpheme stored in the memory means is a title if the morpheme following the morpheme stored in the memory means is a terminal morpheme and is not a period or a question mark; , when the title specifying means specifies that the morpheme stored in the memory means is a title, case converting means converts the first uppercase letter of the morpheme stored in the memory means into a lowercase letter; A machine translation device characterized by being able to perform translations according to.