JP3244286B2 - Translation processing device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a translation processing device.

[0002]

2. Description of the Related Art In recent years, research and development of a machine translation system for translating a sentence from a Japanese language into English or from English into Japanese by a computer has become active. Such a machine translation system generally includes three processes of analysis, conversion, and generation. First, a syntactic relationship and a semantic relationship are analyzed by an analysis process, and as a result, a concept (semantic) structure is obtained. The processing converts the concept structure into the concept structure of the target language. Further, in the generation processing, a translated sentence is generated based on the conceptual structure.

A conventional apparatus of this kind is capable of analyzing, converting, and generating the above-described data in units of one sentence cut out from an input sentence, using only information on the sentence without considering information included in sentences before and after the sentence. Since the processing was performed, there was a problem that it was difficult to automatically select an optimal translated word.

[0004] For example, when the following English sentence is given, these accounts are very misleading. . . "Accounts" in the text is "calculation" as a general translation
Although there may be translations such as "description", "article", and "reason", it is very difficult to determine which of these candidates is appropriate from this sentence alone.

[0005]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned circumstances, and determines the translation of a noun at the time of conversion from a source language to a target language. It is intended to provide a translation processing device capable of realizing efficient translation processing by utilizing a conversion result of a compound word including.

[0006] The translation processing apparatus of the present invention comprises: a word dividing means for dividing a character set inputted by the input means into words;
Detecting means for detecting a compound word as a noun in the words divided by the word dividing means, a target language translation corresponding to the compound word detected by the detecting means, and the compound word; A conversion means for obtaining a translation for the element being read , and when the conversion means obtains a plurality of translations corresponding to each element constituting the compound word,
A selection means for selecting the one completely included in the translation of the compound word as the optimal translation word, and a storage means for storing the optimal translation word selected by the selection means, When each of the above elements appears alone, there is provided a setting means for setting a translated word to a translated word corresponding to the element stored in the storage means.

[0007]

According to the translation processing apparatus of the present invention, the character set input by the input means is divided into word units by the word division means. Next, a compound word is detected as a noun in the words divided by the word dividing means by the detecting means. Further, the conversion means obtains a translation of the target language corresponding to the compound word detected by the detection means and a translation of each element constituting the compound word. At that time, when a plurality of translations corresponding to each of the above elements is obtained, the selecting means selects the plurality of translations.
The best translation is the one that is completely included in the translation of the above compound word.
It is selected as a word and stored in the storage means. Thereafter, when each of the elements appears alone after the compound word, the translation unit sets the translation to the translation corresponding to the element stored in the storage unit.

[0008]

FIG. 1 is a system configuration diagram of a translation processing apparatus according to an embodiment of the present invention.

In FIG. 1, an original text input unit 11 is realized by input means such as a keyboard. The one-sentence cutout unit 12 cuts out the input sentence one by one by detecting punctuation marks and the like of the input sentence, and the translation main processing unit 1.
3. This translation main processing unit 13
Is composed of an original sentence analysis unit 131, a conversion processing unit 132, a translated sentence generation unit 133, and an optimum translated word selection process 134, and includes a data processor including a CPU and the like including the one sentence cutout unit 12. The translated sentence output unit 14 outputs the translated sentence generated by the translation main processing unit 13 to the outside, and includes a sentence memory including a RAM and a CRT for displaying the contents of the memory.

As shown in FIG. 4, the original sentence analysis dictionary 15 is composed of a morphological analysis dictionary for managing headings, parts of speech, and utilization, and a syntax analysis dictionary for managing rules such as sentence patterns. As shown in FIG. 5, the dictionary 16 manages headings, parts of speech, and translated words in correspondence. The translation generation dictionary 17 manages rules such as ending and utilization, and each of the dictionaries is a dictionary memory for executing the translation processing of the translation main processing unit 13, and is composed of a RAM, a ROM, or the like. Is done.

Here, regarding the translation main processing unit 13,
The configuration will be described in detail.

In particular, the original sentence analysis unit 131 has three functions including a morphological analysis function, a syntax analysis function, and a semantic analysis function, and each processing will be described below.

1. Morphological analysis function: It works as a word division unit that divides a character set input by the input unit into words. That is, processing is performed with reference to the information managed by the morphological analysis dictionary of the original sentence analysis dictionary 15. Further, it operates as a detecting means for detecting a compound word of a noun from the divided words by the part of speech setting processing for each word. In addition to this, analysis processing such as detection of parts of speech and inflected forms is executed.

2. Syntax analysis function: Based on the information obtained by the morphological analysis function, a structural rule such as a dependency relationship is analyzed by referring to the information managed by the syntax analysis dictionary of the original text analysis dictionary 15. .

3. Semantic analysis function: Analyzes the (language-specific) semantic structure of the input sentence based on the result of the syntax analysis function.

On the other hand, the conversion processing unit 132 converts a target language corresponding to the compound word detected by the morphological analysis function into a target language,
And serves as a conversion means for obtaining a translation for an element constituting the compound word, that is, processing is performed with reference to the information managed by the translation conversion dictionary 16. Further, when each of the above-mentioned elements appears alone after the above-mentioned compound word, it also operates as a setting means for setting the translated word to the translated word corresponding to the element stored in the translation conversion dictionary 16. .

The translation generation unit 133 also converts the conversion processing unit 132 based on the analysis result of the original text analysis unit 131.
An appropriate translated sentence is generated by converting the ending, utilization, etc. of the translated word obtained in step (1). That is, the processing is executed by referring to the information managed by the translation generation dictionary 17.

The optimal translated word selection processing unit 134 selects the optimal translated word from the translated words corresponding to each component of the compound word based on the translated word of the compound word obtained by the morphological analysis function of the original sentence analyzing unit 131. It operates as selection means.

To summarize the above structure, a character set input by a keyboard or the like is divided into words by the morphological analysis function of the original sentence analysis unit 131. Detect words. Further, the conversion processing unit 132
Then, a translation of the target language corresponding to the compound detected by the morphological analysis function and a translation of each element constituting the compound are obtained. At this time, when a plurality of translations corresponding to the above elements are obtained, the optimal translation selection process 134
Then, based on the translation of the compound word, an optimal translation is selected from among the translations corresponding to each element, and stored in the translation conversion dictionary 16 or the like. Thereafter, when each of the above elements appears alone after the above compound word, the translation processing unit 132 sets the translation to the translation corresponding to the element stored in the translation conversion dictionary 16 or the like.

The operation of the apparatus of the present invention shown in FIG. 1 when the input English text shown in FIG. 2 is input from the original text input unit 11 will be described below.

First, the text file is cut out one sentence at a time by the one sentence cutout unit 12 and sent to the translation main processing unit. For example, the sentence "He has been described
insome news accounts. . . 21 is cut out, the translation main processing unit 13 first compares the information with the above information managed in the morphological analysis dictionary of the original sentence analysis dictionary 15 incorporated in the system, and Recognition (part of speech and inflected forms) is performed.

As shown in FIG. 3, the morphological analysis dictionary stores, in addition to headings (original forms), information on parts of speech and utilization in a table format. In the case of dictionary lookup, the information is registered in the dictionary. A word that matches in a longer unit is prioritized as a dictionary lookup result.

When the sentence 21 is input, a dictionary lookup result of each word is obtained as shown in FIG. 4 and stored in an array. At this time, "news accou
When the word "nt" is registered as a heading,
The part "news accounts" is analyzed as one word.

That is, as shown in the figure, the heading "he" corresponds to the part of speech "pronoun" and the inflected "original form", and the heading "ha"
"ve" corresponds to the part of speech "auxiliary verb" and the conjugation "third person". Similarly, "be", "describe", "in", "som
For "e" and "news account", the correspondence with the part of speech and utilization as shown in FIG.

Each word (content word such as noun, verb, etc.) recognized by the morphological analysis is associated with a translation conversion dictionary (dictionary for selecting a translation word) 16 as shown in FIG. In the case where a plurality of translated words are considered in the translation conversion dictionary, a plurality of translated word candidates are arranged.

In the translation processing apparatus of the present invention, when the above-described morphological analysis is performed, for a compound word composed of a plurality of words, the components of the compound word are also exemplified as follows. Then, a dictionary lookup is performed to determine the correspondence between the translated words of each component, and the optimal translated word selection process is performed. That is, for example, the "news account"
Is translated into a dictionary as a compound word, and the components “news” and “account” are also translated into a dictionary to obtain information on translated words. That is, as shown in FIG. 6, a Japanese translation of "news article" 62 is obtained by dictionary lookup of "news account" 61, and a dictionary lookup of "news" 63 and "account" 64 constituting the compound word. For each of the Japanese translations, candidates for "news", "news", "news", and "account" for "calculations", "descriptions", "articles", and "reasons" are looked up in a dictionary.

Next, FIG. 7 is a flowchart for narrowing down the translated words in the translation processing apparatus according to one embodiment of the present invention.

In the figure, the heading of the compound word is Xe, and the translated word is Xj. When the head of the component of the compound word is Xe [m] (m> 0), the translated word is Xj [m, n] (n> 1). Further, Xe [1]
When the translation corresponding to has multiple possibilities, the translation is Xj
[1,1], Xj [1,2], Xj [1,3]. . . Becomes

At this time, Xj [m, n] (m> 0, n>
0) is included (even partially) in Xj, the heading Xe
Even when [m] is used alone, it can be said that Xj [m, n] is appropriate for the translated word in this text.

That is, in the flow shown in the figure, at step S1, it is determined whether or not Xj [m, n] (m> 0, n> 0) is included (even partially) in Xj. If it is determined that it is included, information that prioritizes Xj [m, n] as a translation of Xe [m] is stored in step S2.

Next, the processing for narrowing down the translated word will be described in further detail, taking as an example the processing for the above-mentioned compound word “news account”.

The translated word “news article” of the compound word “news account” includes “news” among the translated word candidates (“news” and “news”) 5 of “news”, and in the text of FIG. As the translation of "news", "news" is judged to be appropriate. Similarly, the compound word "news accoun
The translated word “news article” for “t” includes “article” among the candidate translations for “account” (“calculation statement”, “description”, “article”, “reason”). It is determined that “article” is appropriate as a translation of “account”. (In this example, the case where the translation of the element is completely included in the translation of the compound word is shown, but partial matching may be used if they do not completely match.) That is, this determination result (in the text of FIG. 2, “ The word "news" should be translated as "news", and the word "article" could be translated into "account".
Is stored in the translation conversion dictionary 16 as learning data. As a method of storing this learning data,
The following two examples are possible.

(1) A frame in which the priority order can be described is provided in the translation conversion dictionary 16, and the priority of the corresponding translated word is increased and stored.

(2) In addition to the translation conversion dictionary 16, a translation learning dictionary is provided, and the translated word information of the corresponding word is described therein.

After performing the above-described optimal translation selection processing, syntax and semantic analysis are performed to perform conversion processing. When a translation word is obtained by the conversion process, the translation word selected in this process is prioritized for "news" and "account" that have been optimized thereafter, and for other words that have not been optimized yet, General translations are given. Based on this conversion result, a translated sentence is generated from the conceptual structure, and is displayed on the CRT of the translated sentence output unit 14.

The translation process proceeds while performing the above operations in units of one sentence. Here, when the sentence 22 “these account are very mislcanding” is cut out from the input text of FIG. 2 and passed to the translation main process, a morphological analysis result as shown in FIG. 8 is obtained. Here, for the translation of “account”, the translation conversion dictionary has been updated by the optimal translation selection process applied when processing sentence 21, and by referring to the stored learning data, the “article” Can translate.

[0037]

According to the present invention, by utilizing information on compound words appearing in an input text, words can be appropriately translated according to a text (scene), thereby improving translation accuracy. be able to.

[Brief description of the drawings]

FIG. 1 is a functional configuration diagram of a translation processing apparatus according to an embodiment of the present invention.

FIG. 2 is a diagram showing an example of an input English sentence text of a translation processing apparatus according to an embodiment of the present invention;

FIG. 3 is a diagram showing a morphological analysis dictionary of the translation processing apparatus according to one embodiment of the present invention;

FIG. 4 is a diagram illustrating an example of a morphological analysis dictionary lookup result of the translation processing apparatus according to an embodiment of the present invention;

FIG. 5 is a diagram showing English and Japanese translation conversion dictionaries of the translation processing apparatus according to one embodiment of the present invention;

FIG. 6 is a diagram showing a correspondence between a compound word and a translation of a component in the translation processing apparatus according to the embodiment of the present invention;

FIG. 7 is a flowchart of narrowing down a translation word in the translation processing apparatus according to one embodiment of the present invention;

FIG. 8 is a diagram showing another example of a morphological analysis dictionary lookup result of the translation processing apparatus according to the embodiment of the present invention;

[Explanation of symbols]

 11 Original sentence input unit 12 1 sentence extraction unit 13 Translation main processing unit 14 Translation output unit 15 Original sentence analysis dictionary 16 Translation conversion dictionary 17 Translation sentence generation dictionary 131 Original sentence analysis unit 132 Conversion processing unit 133 Translation sentence generation unit 134 Optimal translation word selection processing

Claims (2)

(57) [Claims] 1. A word dividing means for dividing a character set inputted by an input means into words, a detecting means for detecting a compound word by a noun among words divided by the word dividing means, A conversion unit for obtaining a translation of the target language corresponding to the compound word detected by the detection unit, and a translation for an element forming the compound word; and each element forming the compound word by the conversion unit. When a plurality of translations corresponding to are obtained, the translation is completely included in the translation of the compound word among the plurality of translations.
Comprising a selection means for selecting what is a best translation, and storage means for storing an optimal translation elected by該選detecting means, after said compound word, when each element is encountered alone, that A translation processing apparatus, comprising: setting means for setting a translated word to a translated word corresponding to the element stored in the storage means. 2. The compound word is constituted by the conversion means.
When multiple translations corresponding to each element are obtained,
Of the translations, those completely included in the translations of the above compound words
If there is, select it as the best translation and it is completely included
If there is no match, a partial match with the translation of the above compound word
It is equipped with a selection means for selecting the best translation.
2. The translation processing device according to claim 1, wherein: