JPH0256066A - Processing system for converting translated word to compound noun - Google Patents

Abstract

PURPOSE:To decrease the monotonousness and the redundancy of the Japanese language which is generated by translating into a compound word having a structure of a prefix + a word base or a compound word having a structure of a word base + a suffix. CONSTITUTION:In a sentence structure analyzing part step 2, a structure of an inputted English sentence is analyzed, and in an English structure identifying part step 3, in order to a clarify a connecting relation, etc., a tree structure in which a verb node is a master is made. This tree structure is converted to an intermediate structure by which an expression of the Japanese language can be shown easily in a converting part step 4. In this case, a noun phrase constituted of an adjective + a noun is transformed to a Japanese language noun phrase constituted of a prefix + a word base, and a noun phrase having a constitution of a noun (phrase) + a preposition + a noun (phrase) is transformed to a Japanese language noun phrase of a constitution of a word base + a suffix. Subsequently, in a sentence structure generating part step 5, the Japanese language is generated by a sentence clause unit, and thereafter, in a morpheme generating part step 6, an array of particles is determined and a conjugation form of a verb is regulated.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a machine translation system, and particularly to an English-Japanese machine translation system.

In conventional machine translation systems, it was not appropriate to translate a single noun phrase. To explain an English-Japanese machine translation system as an example, the noun phrase 1, which is composed of adjective tone + noun contained in an input English sentence. There wasn't. In this case, you may want to use the prefix ``小'' to obtain a concise translation of ``small animal.'' Also, a noun phrase composed of noun + preposition + noun included in the input English sentence, for example, feeling of 1nferior.
When translating ity, the method used was to arrange two nouns using particles, so it was translated as ``sense of inferiority,'' but the preferred translation is ``sense of inferiority,'' a compound word that includes a suffix. .

In this way, translated words that do not use prefixes or suffixes sometimes lack conciseness as Japanese.

Purpose The present invention was made in view of the above-mentioned circumstances.
In particular, the purpose of this invention is to provide a processing method that can translate Japanese compound nouns using prefixes or suffixes.

Regarding prefixes, when translating a noun phrase in English that has the structure of adjective tone + noun into Japanese, it is usually translated as an adjective or an adjective verb ten nouns.

For example, new product means "new product J
, nextgeneration is "generation of law", 1
"ast emperor" is translated as "the last emperor," but the modifier parts in the previous two examples, "new" and "next," are replaced by the prefix "new," which consists of a single kanji character, instead of these adjectives. J "Next" is often used to mean "new product", "next generation J", etc.

The same goes for suffixes: 1 noun (phrase) + prefix + noun (
When English noun phrases with the structure ``clause'' are translated into Japanese, they are often translated as ``nouns, case particles, class ten nouns''. for example,
A passage to India is translated as "a passage to India." However, in some cases, it would be more natural to translate the word into Japanese by omitting the case particle that connects the nouns (phrases) and translating it as a compound noun with the structure of Chinese ten suffixes. For example, a feeligof 1nferio
Rity is a more appropriate translation of ``feeling of inferiority'' than ``feeling of inferiority.'' It is an object of the present invention to provide a processing method that can convert such a compound tone into a translated word.

4-1 In order to achieve the above-mentioned object, the present invention describes the word of the translation target language and its part of speech, the translated word Japanese and its part of speech in the translated word compound nounization processing method in a machine translation device. Japanese nouns are converted from the source language to the target language using a dictionary, the results of the parsing function performed using the dictionary, and a connection table between prefixes and/or suffixes and word bases. A dictionary that is characterized by being transformed into phrases, and further describes the words of the language to be translated and their parts of speech, as well as the translated word Japanese and its parts of speech.

and the result of the parsing function performed using that dictionary.

By using the prefix table and base table to move the nodes of the tree structure that is the analysis result in the conversion part from the yK language to the target language, when a noun phrase with the composition adjective + noun appears, A dictionary that is characterized by transforming phrases into Japanese noun phrases with a prefix + Chinese word structure, and also describes the words of the target language and their parts of speech, as well as the translated Japanese words and their parts of speech. , the results of the parsing function performed using the dictionary, and the word base suffix connection table are used to move and delete nodes in the tree structure that is the analysis result in the conversion part from the source language to the target language. By doing this, when a noun phrase with the structure of noun (phrase) + prefix + noun (phrase) is cursed, the phrase is transformed into a Japanese noun phrase with the structure of Chinese + suffix. This is what I did.

The following is an explanation based on an embodiment of the present invention.6 Figure 1 is 1.
The outline of the processing of one embodiment of the machine translation device is shown using an English-Japanese machine translation system as an example. In the figure, 1 is a dictionary,
Reference numeral 2 denotes grammar rules, 11 to 15 various files based on dictionaries, and 21.22 files for analysis rules based on grammar rules. Dictionary 1 describes one word form, conjugation, one part of speech, translation, etc. In this dictionary, for prefixes,
It is written as ``prefix,'' and nouns that can take a prefix are written as ``one word base.'' These codes distinguish whether a certain English noun phrase has a translation that can be translated as a compound noun. . The information files 11 to 15 contain dictionary information suitable for each of the 5 steps described below, and the grammar rules 2 and analysis rules 21 and 22 are also provided with grammar rules corresponding to each of the 5 steps.

The input English text is first recognized as individual words by a morphological analysis unit (stepl), and special character strings are grouped into phrases.

The structure of the input English sentence is analyzed in the syntax analysis section (step 2). For example, adjective tone + noun is analyzed as a noun phrase. In the English structure identification section (step 3), a tree structure is created with the verb node as the parent in order to clarify the dependencies and relationships. This tree structure is converted by the conversion section (stop 4) into an intermediate structure that can easily express Japanese phrases. Syntax generator (g
After generating Japanese phrases in units of clauses in step 5), the morpheme generation unit (step 6) determines the arrangement of particles and arranges the conjugation of words.

As part of the above-mentioned conversion section (step 4), prefix selection processing is performed, an example of which is shown in FIG.

First, starting from the start node (step 1) of the tree structure, the nodes are sequentially traced to find one noun node (step 2).

If it is not a noun node, the process ends and the process proceeds to the next node.If it is a 0 noun node, it is checked by referring to dictionary information whether the translation of the noun is a base word (stop 3). If the noun node is not a word base, the process ends.If the noun node is a zero word base, it is checked whether the noun node has a child node (step 4). If there are no child nodes, the process ends. If it has a child node, check the part of speech of the last child node among the child nodes (step 5). For efficiency, remove child nodes other than the last child node from the target of prefixing. If the part of speech of the last child node is not similar to an adjective (preposition ), the process ends. In the case of an adjective class, it is further checked whether the node of interest has a child node (step 6). If it has, the process ends. If it does not have a prefix, it is checked whether the translation of that adjective class has a prefix (step 7). If not, the process ends. If so, refer to the prefix word base connection table and check whether the connection is possible (step 8).

If the connection is possible, take that prefix as the translation (ste
p9), repeat the above process until there are no more nodes.

Next, a specific example will be explained.

(1) Input English sentence 1: Th1s is the 5pe
specification of our now prod
rct.

This English sentence is 1-1 in 5top2 of Figure 2.
From the analysis of table (a), it can be determined that there is one noun node "this", but since the translated word is not the base (second
Table 1) Finish the process and search for the next noun node.

The next noun node in Table 2-1 is “5specification” (
In (C) of Table 1-1, it can be seen that the translated word "spec" is a noun based on the dictionary information (Table 2-2).

Next, check whether the last child node is an adjective class. In this case, the youngest node is
Since this is not an adjective class, the processing ends.

The next noun node is "product", which is.

The dictionary information (Table 2-3) shows that the translation of "product" is a noun that is the base of the word.

Table 2-2 Checks whether this node has child nodes.

Table 2-3 Furthermore, from the analysis of (c) in Table 1-1, this noun node has a child node, so if we refer to the dictionary contents of the youngest child node, 'new' (Table 2-4), we get: It can be seen that the translated word has the prefix ``new''.

Table 2-4 So, I looked at the prefix word base connection table (Table 3) and found
Check whether "Product" is connected to "New". Since it is possible to connect, "new" is taken as the translation.

This youngest node is moved to the parent node to create a compound word node ((e) in Table 1-2).

Since (b) in Table 1-3 further has an adverb node ((C) in Table 1-3) that modifies it, the process of taking the prefix as a translation word is not performed.The zero-order noun node "■" ( 1st-
The translated word (d) in Table 3 does not serve as the base word, so the processing ends. The translation is ``fairly new product'' and ``fairly new product.''
It is not.

Then, when the processing is completed, the translated sentence is ``This is the specification of our new product.''

(2) Input English sentence 2: I bout a pret
ty new product.

Noun node “product” ((a) in Table 1-3)
There is.

According to dictionary information (Table 2-1), “Product”
The translation of ``product'' is the base noun. This noun node has child nodes, but the youngest child node "new" (the first -
Table 3 As a result of this, ``I bought a fairly new product.''
The following translation is obtained.

Figure 3 shows an example of compound nounization processing performed using open radicals and suffixes, and is performed as part of step 4 in Figure 1 above, so the explanation for Figure 1 is Don't repeat.

As part of the above-mentioned conversion unit (FIG. 1-12), first, starting from the start node (stopl) of the tree structure, the nodes are sequentially traced to search for a noun node (5 step 2). If not (
(such as in the case of imperative statements), the processing ends. If so, check the dictionary information to see if the noun has a suffix as a translated word (step 3).

If there is no suffix, processing ends. If there is a suffix, the suffix is stored in the storage unit (step 4). Next, check whether the noun node has a child node (st
ep5), if there is no child node, the process ends. If there are child nodes, search for preposition nodes among the child nodes. If there is no prepositional node, the process ends (step
6). If so, check whether the prepositional node has any further child nodes (step 7). If there are no s child nodes, the process ends. If there is a child node, the dictionary is checked to see if the word at that node has a word that can be used as a translation word (stsρ8). If there is no translation word that can serve as the base, the process ends.

If there is a translated word that can be used as a word base, store the translated word in the memory unit (step 9).Next, check whether the suffix stored in the memory unit and the word base can be connected by referring to the word base suffix connection table. (stoplO). If the connection is possible, the base node is moved to the suffix node to form a noun phrase node consisting of continuous nouns (step H).

Further, the preposition node is deleted (stopl2), and the above process is repeated until there are no more nodes.

Next, a specific example will be explained.

(1) Input English sentence t: Ha has a strong
feeling of superiority to
his classmates.

This English sentence is the noun node “He n (( in Table 4-1)
a)) H ell does not have a suffix as a translated word according to the dictionary information in Table 5-1, so it is converted into a suffix.

Compound word processing is not performed.

The next node is a noun node consisting of "feeling" ((b) in Table 4-1), and it can be seen from the dictionary information section 5-2 table that there is a suffix "r-kan" as a translation of "feeling". , the "feeling" is stored in the storage section.

Table 4-1 Table 5-2 Next, this noun node has a child node, so check whether there is a preposition node in the child node. 71o f
” ((C) in Table 4-1), so check whether this node has any child nodes.a Since there is a child node called “5superiority” ((d) in Table 4-1) )) 5periorit
y'' dictionary information (Table 5-3), and it is found that the translated word ``Yuyoshi'' can be a word that can be a base word, so ``Yyyoshi'' is stored in the storage section.

The node moves beyond the preposition node consisting of "feeling" to the noun node consisting of "feeling" and places it before "feeling" to form a noun phrase node "5superiority feeling". The prepositional node consisting of “of” which is a child node of this noun phrase is deleted (No. 4-2
table), where there is a noun node consisting of “feeling”, a preposition node consisting of “of”, and “5superior”.
The zero-order noun node that completes the tonal ratio processing is "classmates" ((e) in Table 4-1), but the dictionary information (Table 5-4) (table), it does not have a translated word as a suffix, so compounding processing is not performed.

Next, refer to the word base suffix connection table (Table 6).

Since we can see that [feeling] and "superiority" can be connected, we set the noun node "5superiority" to the fifth parent node.
-4 Table 4-2 From this result, we get the following translation: ``He has a strong sense of superiority over his classmates.''

As mentioned above, the second
In the figure, we explained how to translate into a compound word with a prefix, and in Figure 3, how to translate into a compound word with a suffix, but by providing both of these at the same time, it is possible to translate into a more appropriate Japanese expression. It is completely understandable that this is possible.

Effects As is clear from the above explanation, according to the present invention, Japanese words generated by translating into compounds with the structure of prefix gentleman base and/or compound words with the structure of word base + suffix. It has the effect of reducing the monotony and redundancy of sentences.

[Brief explanation of the drawing]

FIG. 1 is a schematic diagram for explaining one embodiment of the present invention,
FIG. 2 is a partial detailed view of step 4 in FIG. 1. FIG. 3 is a partial detailed diagram of another embodiment of 5tep4 in FIG. 1. 1...Dictionary, 11-15...Dictionary information file, 2
... Grammar rules, 21.22... Parsing rules. Diagram output text (Japanese)

Claims (1)

[Scope of Claims] 1. A dictionary that describes words of a language to be translated and their parts of speech, Japanese as a translation word and parts of speech of the Japanese, and the results of a parsing function performed using the dictionary; Translated word compound noun processing in a machine translation device characterized by transforming into Japanese noun phrases in the conversion part from the source language to the target language using a connection table between prefixes and/or suffixes and word bases. method. 2. A dictionary that describes the words of the target language and their parts of speech, the translated Japanese words and their parts of speech, the results of the parsing function performed using the dictionary, the prefix table, and the word base. Using a table, in the conversion part from the source language to the target language, by moving the nodes of the tree structure that is the analysis result, when a noun phrase with the composition of adjective + noun appears, that phrase can be changed to prefix + Chinese. A translated compound noun processing method in a machine translation device, which is characterized by transforming a translated word into a Japanese noun phrase with the following structure. 3. A dictionary that describes the words of the target language and their parts of speech, the translated Japanese word and its parts of speech, the results of the parsing function performed using the dictionary, and the word base suffix connection table. In the conversion part from the source language to the target language using
By moving and deleting nodes in the tree structure that is the result of the analysis, when a noun phrase with the structure of noun (phrase) + prefix + noun (phrase) appears, that phrase is called Chinese + suffix. A translated word compound noun processing method in a machine translation device, which is characterized by transforming a Japanese noun phrase into a Japanese noun phrase.