GB2193355A - Machine translation - Google Patents

Abstract

A machine translation system capable of translating (S11) an input sentence by using translated words delivered from a dictionary, distinguishably displays (S13) any word not registered in said dictionary among the words contained in the input sentence, without executing the translation process, and an executing circuit erases the distinctive display (S3) on the basis of a subsequent translation command and executes the translation process at the same time. <IMAGE>

Description

SPECIFICATION Machine translation system BACKGROUND OF THE INVENTION This invention relates to a machine translation system, having a dictionary to deliver a translated word of a word contained in an input sentence, capable of translating said input sentence by using the translated words delivered from said dictionary, and more particularly to an automatic translation system which does not execute translation process if a word contained in the input sentence is not registered in said dictionary, distinguishes such word from other words, urges the operator to process said unregistered word, cancels this distinguished state according to the translation command after the distinctive display is made, and starts the execution of translation process.

A machine translation system, in order to convert an ariginal language into an intended language, requires at least a parallel dictionary of the original langauge and the intended language.

That is, if any one of the plural words in an fed original, that is, an input sentence is not registered in the dictionary, translation is disabled, and therefore it is required, before starting translation process, to tell the operator that there is a word not listed in the dictionary.

The conventional machine translation system only told the presence of an unregistered word in the original (input sentence), but it was impossible to know which word is unregistered if the presence of unregistered word in the input sentence is known, and in order to find the unregistered word, the operator must investigate every word contained in the input sentence.

SUMMARY OF THE INVENTION In the light of such situation, this invention is intended to be convenient for the operator by composing a constitution wherein translation process is not immediately started if a word not registered in the dictionary is found in an input sentence, and this word is temporarily distinguished from the other words so that the operator can see which word is unregistered at a glance, and by the subsequent translation command, such distinctive display is automatically canceled without the manipulation of the operator.

Briefly described, in accordance with the present invention, is a machine translation system, having a dictionary to deliver a translated word. of a word contained in an input sentence, capable of translating said input sentence by using the translated words delivered from said dictionary, comprising a display means for, if there is any word not registered in said dictionary among the words contained in the input sentence, distinguishing such unregistered word from other words without executing the translation process, and an executing means for erasing said distinctive display on the basis of the translation command after the distinctive display is formed by said display means, and executing the translation process at the same time.

BRIEF DESCRIPTION OF THE DRA WINGS The present invention will become more fully understood from the detailed description given hereinbelow and the accompanying drawings which are given by way of illustration only, and thus are not limitative of the present invention and wherein:: Fig. 1 is a block diagram of a machine translation system according to this invention, Fog. 2 is a structural diagram of translation module 5 of said machine translation system, Fig. 3 is a diagram showing the stored state of an input sentence in a buffer, Fig. 4 and Fig. 5 are diagrams to explain the steps of translation process of said translation module, Fig. 6 is a processing flow chart for distinctive display of an unregistered word of said machine translation system, Fig. 7 (a) to (d) are diagrams to show the display state in the processing of distinctive display of said unregistered word, Fig. 8 and Fig. 9 are diagrams to show the display state of distinctive display, Fig. 10 is a processing flow chart of registration of word, and Fig. 11 and Fig. 12 are diagrams to show the display state in said processing of registration of word.

DESCRIPTION OF THE PREFERRED EMBODIMENT Referring now to the drawings, the constitution of the invention is described below.

Fig. 1 is a block diagram of a translation system in one of the embodiments of this invention.

In the diagram, numeral 1 is central processing unit CPU responsible for translation processing and others, and to this CPU 1 are connected a main memory 2, a display device CRT 3, a keyboard (input unit) 4, and a translation module 5.

In the main memory 2, a program for system processing in the CPU 1 is stored, and it has a plurality of buffers. On the CRT 3, displays necessary for translation process are shown, and inputs necessary for translation process are made from the keyboard 4.

Said translation module 5 is to store the information necessary for translation steps such as translation program, and it has a table 6 which comprises a dictionally referred to in translation, grammatical rules, and tree structure conversion rules.

The structure of the translation module 5 is shown in Fig. 2, in which 5a is a dictionary lookup morpheme analyzing part, 5b is a syntax analyzing part, 5c is a language converting part (converter), and 5d is a target language generating part (generator).

Dictional lookup means to search the dictionary to obtain information such as morphological attributes and grammatical attributes of each word in the input sentence.

The morpheme analysis is to divide an input sentence into word units by using the information obtained by dictionary lookup, and to provide each word with morphological attributes such as singular/plural, person (first/second/third), tense, and case (nominative/objective/possessive), and grammatical attributes such as intransitive/transitive.

The syntax analysis is to determine the syntactic structure by using the morphological attributes and grammatical attributes of words obtained by the morpheme analysis, while eliminating the modifying/modified relations not allowed grammatically between words.

Conversion is to convert into the structure of the target language.

Generation is to arrange the word order in the target language, and make up a morpheme synthesis.

The operation of the translation module 5 is explained below by referring to a practical sentence.

For example, to translate an English sentence This is a book.

into a Japanese sentence, the operation is as follows.

This original sentence is fed into the buffer (not shown) of said translation module 5 through said keyboard 4. The state of input in the buffer is as shown in Fig. 3.

After the entry of input sentence, the operator manipulates the translation key (not shown) on the keyboard 4. By this translation (start) command, the translation module 5 begins to execute the translation process by using the table 6.

The procedure of translation process is approximately as follows.

That is, the dictionary lookup morpheme analyzing part 5a receives the source language fed in said buffer, and picks up information necessary for each word. As a result, the information on part of speech is obtained as shown in Table 1.

Word Part of speech This Pronoun is Verb is Verb a Article book Noun Table 1 Consequently, the relation is determined as shown in Fig. 4 by the syntax analyzing part. At this time, the analysis shown in Table 2 is executed according to the grammatical rules on the table 6.

Sentence + Subject, predicate Subject + Noun phrase Predicate ) Verb, noun phrase Noun phrase + Pronoun Noun phrase B .Article, noun Table 2 This rule indicates as follows.

(a) A sentence is composed of the subject and the predicate.

(b) The subject is made of a noun phrase.

(c) The predicate is made of a verb and a noun phrase.

(d) A noun #hrase is made of a pronoun.

(e) A noun phrase is made of an article and a noun.

Next, by the language converting part 5c, an analysis is effected according to the tree structure conversion, same as in the syntax analysis, and the result is determined as shown in Fig. 5. Later, in the target language generating part 5d, proper postpositional word and auxiliary verb are added, and the garget language is delivered, and r#zrzt;;z##;Pbo is displayed on the CRT 3.

In this embodiment, in translation process, if information of a word in an input sentence, such as morphalogical attribute and grammatical attribute, is not obtained by searching the dictionary in the dictionary lookup step, that is, if a certain word in the input sentence is not found in the dictionary, such word is destinguished in display from the other words registered in the dictionary.

At this time, the translation process is not executed.

After the unregistered word is distinguished in display from the other words registered in the dictionary, the operator processes this unregistered word properly and makes a translation instruction again, that is, when there is no unregistered word in the input sentence, the translation process is executed.

The processing action about distinctive display of an unregistered word is explained according to the flow chart in Fig. 6.

After the input of an original sentence (an input sentence) by the Qperator, a translation instruction from the operator is awaited, and if this translation instruction is given, the next process is executed (step S1).

First, the presence or absence of unknown word display is judged (step S2). The presence of unknown word display means there is an unregistered word among the words in the input sentence, as a result of dictionary lookup, in processing after the translation instruction in the previous step. If there is no unknown word display, it means all the corresponding information is obtained for the wqrds in the input sentence by dictionary lookup in processing after the translation instruction in the previous step.

If there is an unknown word display, this display is erased (step S3).

Next, the word No. showing the order of each word contained in the input sentence is set in the CPU 1 (step S4), and after setting, the presence or absence of the first word included in the input sentence is judged (step S5), and when the first word is present, the operation advances to the dictionary lookup process (step S6).

If successful as a result of dictionary lookup (step S7), the word No. is increased by one in order to process the second word contained in the input sentence in succession (step Sg), thereby returning to step S5.

If, however, unsuccessful as a result of dictionary lookup (step S7), the word No. at that time is added to the unregistered word list (step S7), and then the operation goes to step Sg, thereby increasing the word No. After the processing of step Sg, the operation returns to step S5.

Such process is repeated for all the words contained in the input sentence.

When dictional lookup process is finished for all the words, said unregistered word list is checked, and the presence or absence of unregistered word is judged (step S10).

In step S10, if the unregistered word list is empty, that is, if there is no unregistered word in the input sentence, the result of translation is displayed (step S,2) after the translation process of one sentence (step Six)' If, however, an unregistered word No. in the unregistered word list is stored in step S10, the unregistered word is displayed (step Sir3).

In step S13, if a certain word is desplayed to be an unregistered word, the translation process of the input sentence is not executed.

Therefore, the operator processes as specified as for the word found to be an unregistered word in the dictionary by the unknown word display, to make this word not an unregistered word, and instructs translation in step S,.

The practical unknown owrd display is explained while referring to Fig. 7.

As shown in Fig. 7 (a), suppose an original sentence is entered as follows.

This is a boook.

The word No, of each word in the original (input sentence) is as shown in Talbe 3.

Word Word No.

This 1 is 2 a 3 boook 4 Table 3 If the word "boook" of word No. 4 is an unregistered word, word No. 4 is stored in the unregistered word list in step S8, and as a result the display is made as shown in Fig. 7 (b) by the processing of step 513 As shown in the illustration, the mark "*" indicating an unregistered word is attached to the word "boook" of word No. 4, so that the operator knows there is an unknown word display.

When there is an unknown word display, the operator processes as required. The required process means to correct the input sentence or register a word additionally.

In this example, since the word "boook" is understood to be a misspelling of "book" and the operator corrects the word "boook" into "book" by operating the keyboard 4.

The display after correction is as shown in Fig. 7 (c), and the """ mark is still attached, but when the process of step S3 is executed according to the instruction of re-translation, this mark is isautomatically erased, and the translation process is put in action.

Incidentally, the mark is erased by clearing the mark memory region of the display memory of the CRT 3.

Here, the symbol "*'' was used as the mark to distinguish the unregistered word, but this is not limitative, and ally symbol that will rarely appear in the input sentences may be used.

Of course, aside from attaching a mark, it may be also distinguished by other means, such as underscoring the entire word in question as shown in Fig. 8, and displaying distinctively by using the function of black-and-white reversal display, highlight display or color display if the CRT 3 has such function.

Or, when adding a mark, in order to distinguish from the other word with the same symbol appearing in the input sentence, it may be also indicated numerally to show which number word as counted from the beginning of the input sentence is an unregistered word as shown in Fig. 9.

In the case of distinctive display in Fig. 9, if plural words included in the inptit sentence are unregistered, the word Nos. may be separated by the comma (,), and the plural word Nos. are displayed in batch.

Relating to the additional registration of a word mentioned above, the process is described below while referring to the flow chart in Fig. 10.

If a registration instruction is given by the operator (step n1), the word input to be registered is instructed (step n2).

According to said input instruction, the opera#tor enters the word to be registered through the keyboard 4 (step n3).

Then the input of the part of speech of the entered word is instruction (step n4).

According to said input instruction, the operator feed the part of speech of the entered word from the keyboard 4 (step n5).

Futhermore, the input of the translated word of the above entered word is instruction (step n6).

According to this input sintruction, the operator enters the translated word of the entered word from the keyboard (step n6).

After the above steps, the entered word, its part of speech, and the translated word are registered in the dictionary on the table 6 (step n8).

In step n4, a plurality of parts of speech, such as noun and adjective are listed together on the CRT 3, so that the operator can select, in the next step n5, a specific part of speech by using the cursor.

A screen display of CRT 3 urging a practical input instruction is shown in Fig. 11, and the screen display of CRT3 after the input according to said input instruction is shown in Fig. 12.

Thus, according to this invention, in the machine translation system having a dictionary to deliver a translated word of a word contained in an input sentence, capable of translating the input sentence by using the translated words delivered form the dictionary, there are a display means for distinguishing, if there is any word not registered in the dictionary is found among words included in the input sentence, such word from other words without executing the translation process, and an execution means for erasing said distinctive display and executing the translation process according to the translation instruction after the distinctive display is made by said display means. Hence it brings about the following effects.

(a) Which word in the input sentence is unregistered is directly displayed, the operator can immediately plan a measure for processing of such word, and it is efficient.

(b) If the distinctive display still remains alghough the distinctive display is no longer necessary after a proper measure for such unregistered word, it is hard to read the input sentence. To avoid this, when the translation is instructed, this distinctive display is automatically erased, and it is convenient because it is not necessary for the operator to make any action to erase the distinctive display by key operation or other manipulation.

While only certain embodiments of the present invention have been described, it will be apparent to those skilled in the art that various changes and modifications may be made therein without departing from the spirit and scope of the present invention as claimed.

Claims (10)

1. A machine translation system, having dictionary means for delivering one or more translated words of one or more words contained in an input sentence, capable of translating said input sentence by using the one or more translated words delivered from said dictionary,comprising: display means for, if there is any word or words not registed in the words contained in said dictionary means, displaying the distingushing of such unregisted words from other words without executing the translation process; and executing means for executing the erasing of the distinctive display on the basis of the translation command after the distinctive display is carried out in said display means and simultaneously executing the translation process.

2. The machine translation system of claim 1, wherein said display means is operated in response to the detection that there is no information of the word inclusive of morphalogical attribute and grammatical attribute.

3. A machine translation system comprising: buffer means for storing source words dictionary lookup morphame analyzing means for analyzing the source words to pick up the information necessary for each word inclusive bf part of speech of the words said dictionary lookup morphaphemp analyzing means further comprising:: display means if there is any word not registered in a dictionary table, displaying the distingushing of such unregistered word from other words without executing the translation process and executing means for erasing said disctinctive display of the basis of the translation command after the distinctive display is formed by said display means and for executing the translation process in response to an additional translation command; syntax analyzing means responsive to said dictionary lookup morphapheme analyzing means for picking up a grammatical relation with reference to a grammatical table; language converting means for analyzing the structure of the language of the words and target lanuage generation means for generating the necessary propositional language and axliliary verb to ontain a target language.

4. The machine transiation of claim 3, wherein said dictinctive display of said display means is carried out by adding an additional mark to the unregistered word.

5. A machine translation system for translating one or more words from a first language to a second language, characterized in that there are provided dictionary means for storing a number of standard words, detection means for detecting whether said one or more words are stored within said dictionary means and display means responsive to said detection means for displaying one or more words unregistered in said dictionary means

6. The machine translation means of claim 5, wherein there is provided erase means for erasing the additional mark in response to a further translation operation command.

7. The machine translation system of claim 6, wherein said further translation operation command is generated in response to the detection that said one or more words are modified or one or more words are added to said old one or more words

8. A machine translation system comprising input means for inputting a group of words, such as a sentence, in a first language; dictionary means for storing translation data for a dictionary table of words from the first to a second language; translation means for translating said input group of words in accordance with translation data from said dictionary means display means for displaying the input group of words of said first language; detection means for detecting which, if any, of said input group of words is or are absent from said dictionary table; and control means operable to control the display means so as to indicate visually said absent word or words, and to inhibit the translation of a said input group of words which includes a said absent word or words by said translation means.

9. A system according to claim 8 wherein said control means is operable to delete the visual indication when corrective action has been taken with respect to said absent word or words and a retranslation command has been entered.

10. A machine translation system substantially as hereinbefore described with reference to the accompanying drawings.