GB2211641A - Language translation machine - Google Patents

Abstract

A machine translation system has a word/idiom equivalent display key. When this key is depressed, idioms formed of words and prepositional words in the source language are displayed in whole together with translation equivalents thereof. <IMAGE>

Description

TITLE OF THE INVENTION Machine Translation System BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a machine translation system which is able to produce a correct translation result through interaction between translation machine and operator.

2. Description of the Prior Art A machine translation system as is known from the prior art is described as follows. In the prior art machine translation system, the source language, for example, "They consider him a good teacher," is entered from the keyboard and translation processing is performed and then, when the cursor is positioned at the word "good" in the original (source language) as shown in Fig. 8, the word "good" and the equivalent "YOI" in Japanese (target language) will be highlighted together. If the word equivalent display key is depressed while the word "good" is highlighted, a window shows word equivalents selected from the dictionaries, as shown in Fig. 8.

In the prior art machine translation system, however, the equivalents are to be displayed such that the window shows only the word in the source language and a corresponding word in the target language. It is therefore difficult to determine from the displayed equivalent(s) alone how the source word is related to what prepositional words or adverbial phrases, and whether the source word is part of an idiom. It is therefore difficult to obtain a good grasp of the results of syntactic analysis.

SUMMARY OF THE INVENTION The present invention has been developed with a view to substantially solving the said disadvantages and has as its aim the provision of a machine translation system which, when a source word is identified as part of an idiom formed by the word and prepositional words, displays the source word with the prepositional words forming the idiom and equivalents of the idiom, and thereby makes it easier and faster to understand the results of syntactic analysis.

In accomplishing these and other objects, a machine translation system comprises: a morpheme analyzer having a dictionary for analyzing a source sentence into morphemes and for obtaining part of speech and other information using the said dictionary; a syntax analyzer having grammatical rules for analyzing the structure of the morphemes identified by the said morpheme analyzer using the said dictionary and grammatical rules, a transfer mechanism for converting the structure obtained by the said syntax analyzer to a structure of the target language and a translated sentence generator for generating a translated sentence according to the structure of the target language obtained by the said transfer mechanism.The machine translation system further comprises: a word/idiom equivalent display command means which instructs that words and idioms formed of words and prepositional words in the source language be displayed with translation equivalents; and a word/idiom equivalent display control means which, in addition to displaying words with translation equivalents, displays idioms with translation equivalents when the morpheme analyzer selects an equivalent for an idiom when selecting translations of words from dictionaries and when so instructed by the word/idiom equivalent display command means.

In operation, when the source language for translation is input, the text is separated into morphemes on the basis of dictionaries by the morpheme analyzer, and translation equivalents and other information are obtained. The structure of the morpheme string is analyzed by the syntax analyzer, and the syntactic structure is then converted to the syntax of the target language by a transfer mechanism.

The translated text is then generated by a translated sentence generator according to the syntax of the target language and then output.

In the translation process, if an equivalent for an idiom formed by a word and prepositional words is selected by the morpheme analyzer during the selection of word equivalents from dictionaries by the morpheme analyzer, translation processing is performed. When a word/idiom equivalent display command means is operated to display a word/idiom equivalent display in which the word or idiom in the original is displayed with equivalents, the idiom will be displayed with equivalents by the word/idiom equivalent display control means. Conversely, if a display equivalents command is entered by the word/idiom equivalent display command means when the morpheme analyzer has not selected an idiomatic equivalent, the word/idiom equivalent display control means displays the specified word and its equivalent.

Therefore, when word equivalents are displayed and the displayed word has been identified as forming an idiom with a prepositional word, the prepositional word can be displayed together with idiomatic equivalent.

BRIEF DESCRIPTION OF THE DRAWINGS These and other objects and features of the present invention will become apparent from the following description taken in conjunction with preferred embodiments thereof with reference to the accompanying drawings, throughout which like parts are designated by like reference numerals, and in which:: Fig. 1 is a block diagram of a preferred embodiment of a machine translation system according to the present invention; Fig. 2 is an operating structure diagram of the translation module according to the preferred embodiment; Fig. 3 is a diagram showing a sample of the contents of a dictionary look-up buffer; Fig. 4 is a diagram showing a sample of the information stored in a dictionary for the word "come"; Figs. 5 is a flow chart of the translation process according to the preferred embodiment; Fig. 6 is a diagram showing a sample display of word/idiom equivalent display processing; Fig. 7 is a diagram showing an alternative display sample of Fig. 6; and Fig. 8 is a diagram showing a sample of conventional word equivalent display processing.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Referring to Fig. 1, a block diagram of a first preferred embodiment of a machine translation system according to the present invention is shown. In the drawings, reference number 1 is a CPU (central processing unit), 2 is a main buffer, 3 is a CRT display, 4 is a keyboard, 5 is a translation module, and 6 is a buffer storing dictionaries, grammatical rules, and tree structure conversion rules for translation.

When the source language text is input to the translation module 5, translation module 5 translates this, and outputs the target language. Specifically, the source language entered from keyboard 4 is sent to translation module 5 as controlled by CPU 1. Translation module 5 translates the input source language to the target language as will be described hereafter using the dictionaries, grammatical rules, and tree structure conversion rules stored in buffer 6. The result of this translation is temporarily stored in main buffer 2 and is displayed on CRT display 3.

A specific construction of translation module 5 of the preferred embodiment is shown in Fig. 2.

The operation of the translation module will be described hereinafter with reference to Figs. 1 and 2.

First, the source language is separated into individual morpheme strings (word strings) using the dictionaries in buffer 6 by morpheme analyzer 11. Then, grammatical information such as part of speech and word equivalents are obtained for each word. Also, tense, person, number, and other factors are analyzed. This information is then stored in a dictionary look-up buffer. In syntax analyzer 12, a syntax analysis tree showing the relationship between each word is determined according to the dictionaries and grammatical rules in buffer 6 as will be described hereinafter.

At transfer mechanism 13, the structure of the syntax analysis tree for the input original is converted to the structure of the syntax analysis tree of the translated sentence using the tree structure conversion rules in buffer 6. Next, at translated sentence generator 14, auxiliary words and auxiliary verbs appropriate to the translated sentence in the target language are added, the obtained translated sentence is output from translation module 5, and stored in main buffer 2 and displayed on CRT display 3.

If a word equivalent display key 4a provided in keyboard 4 is depressed, word/idiom equivalent display processing occurs so that the specified word or idiom composed of the word and prepositional words is displayed together with translation equivalents.

Fig. 3 shows part of the contents of the dictionary look-up buffer which stores the information obtained as the result of dictionary look-up by morpheme analyzer 11.

Fig. 4 shows part of the information stored in the dictionaries of buffer 6 relating to the verb "come" which has been entered in the dictionary look-up buffer of Fig. 3.

This information defines the following. Specifically, to output the translation equivalent "KURU (meaning 'come')," equivalent No. 1, the minimum predicate required is only one verb. Furthermore, to output the translation equivalent "SOUGU SURU (meaning 'to meet' or 'to encounter')," equivalent No. 2, the minimum predicate required consists of a verb, the preposition "on", and a noun. In this case, the grammatical rules give priority to equivalent No. 2, "a predicate consists of a verb, preposition, and noun," over equivalent No. 1, "a predicate is a verb." Fig. 5 is a flow chart showing the translation process, which will be described hereinbelow also with reference to Figs. 3 and 4.

At step S1, the source language to be translated is input. In operation, two different cases are considered.

One is a case when the original sentence: (1) "He came here." is entered, and the other is a case when the original sentence: (2) "He came on a friend there." is entered.

At step S2, the dictionary look-up process by morpheme analyzer 11 is performed, part of speech and other grammatical information is obtained, and tense, person, and number are analyzed.

At step S3, syntactic analysis is performed by syntax analyzer 12, and the structural analysis tree is determined.

At step S4, the structure of the structural analysis tree of the source language is converted to the structural analysis tree of the target language.

At step S5, an appropriate translation is generated by translated sentence generator 14.

At step S6, the translated sentence generated by translated sentence generator 14 is displayed on CRT display 3.

If in this translation process translation occurs using the dictionary information as shown in Fig. 4, when source language (1) is input the grammatical rule "a predicate is a verb" is applicable, and the translated sentence "KARE WA KOKO NI KITA. (He came here.)" is output in Japanese.

However, when source text (2) is input, the grammatical rule "a predicate consists of a verb, preposition, and noun" is applied, and the output text is "KARE WA SOKO DE WJIN NI SOUGU SHITA. (He came on a friend there.)" This result is obtained because, as the result of dictionary look-up by morpheme analyzer 11, the word "on" in source language (2) is identified as a preposition as shown in Fig. 3, and further identified as the preposition "on" shown in Fig. 4 which is needed to satisfy the grammatical rule stating that "a predicate consists of a verb, preposition, and noun" as stored in such a way as to indicate a relation with the verb "to come" in the dictionary.

At step S7, it is determined whether or not the word equivalent display key is depressed. If the key is depressed, processing passes to step S8 and word/idiom equivalent display processing occurs; if not, processing passes to step 511.

At steps S8, S9 and S10, the single word "come" specified in the source language, the preposition "on" forming an idiom with the word "come," and equivalents for the idiom "come on" are displayed in a window such as shown in Fig. 6, and processing returns to step S7. If at this time the single specified word does not form an idiom, only the specified word and its equivalents are displayed in the window.

Here, equivalent #0 "SOUGU SHITA" displayed on CRT display 3 in Fig. 6 is the result of translation processing of the phrase "come on," and has been conjugated from equivalent #1 "SOUGU SURU" to agree with the tense and case of the original.

At step S11, it is determined whether or not this is the last sentence. If there is another sentence, processing returns to step S1; if not, processing terminates.

Thus, when a source language is translated in a machine translation system according to a preferred embodiment of the present invention, it is possible to clearly show the relationship between a word and prepositional words which form an idiom when word/idiom equivalents are displayed because the verb "to come", preposition "on," and a group of idiomatic equivalents are displayed when analysis identifies as shown in source language (2) that there is an idiomatic relationship between the verb "to come" and the preposition "on." Therefore, it becomes possible to better understand the results of syntactic analysis, and the burden placed on the operator is reduced.

Furthermore, when a word string is erroneously identified as an idiomatic expression, it is necessary to reject the idiomatic relationship of the words and prepositional words forming the erroneous idiom. In this event, if the prepositional words are displayed as heretofore described, it is possible to easily reject the erroneous idiomatic relationship between the words and prepositional words using a pre-edit support function, and rapid, correct translation processing can be performed.

In the aforementioned preferred embodiment, the selected word(s), prepositional word(s), and idiomatic equivalent(s) may be displayed as shown in Fig. 6, but the selected word "come' in the source language, the preposition "on" forming an idiom with "come", and the equivalent translation "SOUGU SHITA" corresponding to the idiom "come on" may be highlighted in a display as shown in Fig. 7.

As will become clear from the description of the present invention hereinbefore provided, because the present invention provides a machine translation system provided with a morpheme analyzer, syntax analyzer, transfer mechanism, and translated sentence generator with a word/idiom equivalent display command means and a word/idiom equivalent display control means in such a manner that when the morpheme analyzer selects an idiomatic equivalent for words during the selection of translation equivalents, the word/idiom equivalent display control means may display the idiom with equivalents if the word/idiom equivalent display command means instructs that the word or idiom of words be displayed with its equivalents. The prepositional words can be displayed with the idiom in the event the word for which equivalents are to be displayed has been determined to form an idiom with prepositional words. Therefore, it is possible to quickly and correctly grasp the result of syntactic analysis.

Although the present invention has been fully described in connection with the preferred embodiments thereof with reference to the accompanying drawings, it is to be noted that various changes and modifications are apparent to those skilled in the art. Such changes and modifications are to be understood as included within the scope of the present invention as defined by the appended claims unless they depart therefrom.

Claims (6)

CLAIMS:

1. A machine translation system comprising: a morpheme analyzer having a dictionary for analyzing a source language into morphemes and for obtaining part of speech and other information using said dictionary; a syntax analyzer having grammatical rules for analyzing the structure of the morphemes identified by said morpheme analyzer using said dictionary and grammatical rules; a transfer mechanism for converting the structure obtained by said syntax analyzer to a structure of the target language; a translated sentence generator for generating a translated sentence according to the structure of the target language obtained by said transfer mechanism; a word/idiom equivalent display command means for instructing that words and idioms formed of words and prepositional words in the source language be displayed together with translation equivalents thereof; and a word/idiom equivalent display control means which displays idioms with translation equivalents when said morpheme analyzer selects an equivalent for an idiom when selecting translations of words from dictionaries and when so instructed by said word/idiom equivalent display command means.

2. A machine translation system as claimed in Claim 1, wherein said idioms and translation equivalents are displayed in such a manner as to be distinguishable from the rest of the text.

3. A machine translation system comprising means for translating text input in a source language to equivalent text in a target language and display means which is selectively operable so that when the translation means has established an idiomatic association between a plurality of words in the input text and has provided an equivalent in the target language to the associated words, said display means displays said associated words and the target language equivalent or equivalents thereof.

4. A machine translation system according to claim 3 including means operable in an equivalent display mode for permitting the user to select a word in the input text as displayed by the display means, the system being operable to determine whether the translation means has established a said idiomatic association including the selected word, and if so to cause the display means to so display the associated words, including the selected word, and the target language equivalent or equivalents.

5. A machine translation system according to claim 4 further including user-operable command means for selecting said equivalent display mode.

6. A machine translation system substantially as herein described with reference to the accompanying drawings.