JP2000148756A - Translation error detecting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the efficiency of operation for correcting a translation by calculating the error rate of the translation from the original text and its translation. SOLUTION: A translation error detecting device is equipped with a sentence input part 101 which inputs the original text and translation, an information extraction part 102 which extracts original sentence information from the original text and extracts translation information from the translation, a comparison part 105 which compares the original text information and translation information with each other, an error rate calculation part 106 which calculates the error rate of the translation on the basis of the comparison result of the comparison part 105, and a notification part 107 which notifies a user of the translation having a high error rate. Consequently, the original text and its translation are correlated with each other, the translation is compared to calculate its error rate, and the extent of errors of the translation is explicitly displayed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a bilingual sentence error detecting device for improving the efficiency of searching for a document correction portion and improving the quality of a translated sentence.

[0002]

2. Description of the Related Art Conventionally, when a first language is translated into a second language by using a machine translation system, the first language sentence is translated into a format which is easy to handle in order to obtain a high quality translation. Before rewriting, and rewriting, etc., in which a skilled person replaces a translated sentence with language-like expressions.

However, there is no established technique for mechanically and efficiently performing the pre-editing and rewriting as described above, and the work requires knowledge, experience and a great deal of time of a skilled person.

When translating a first language into a second language, there are problems such as missing translations, inconsistent translations, and mistranslations.
This is especially true when there are multiple translators.

[0005] For this reason, a skilled person must check the translated sentence anew, collect translated sentences at the end, and unify the translated words. Is required.

Japanese Patent Application Laid-Open No. 5-298360 discloses a "translation sentence evaluation method, translation sentence evaluation device, machine translation system with translation sentence evaluation function and machine translation system evaluation device". A first translation of a second language for the original and a second translation of a second language for the original
Are prepared, the degree of coincidence between the first translated sentence and the second translated sentence is calculated, and the machine translated sentence is evaluated.

[0007]

However, the above-mentioned "translation sentence evaluation method, translation sentence evaluation device, machine translation system with translation sentence evaluation function, and machine translation system evaluation device"
Is intended to evaluate a translated sentence, and the above-described device cannot improve the work efficiency when correcting an error in the translated sentence or improve the quality of the translated sentence.

Further, in order to use the above-mentioned apparatus, two or more translations of the second language are required, and in order to perform an even more accurate evaluation, a sentence translated by an expert as a second translation is required. Alternatively, it is necessary to prepare a translated sentence edited by a skilled person.

That is, the above system is effective for a document in which an ideal translation is prepared in advance, or for a user who has several machine translation systems, but is not good at translation work. Alternatively, those who cannot own multiple machine translation systems cannot use it. Further, there is no method of displaying a portion to be corrected to the user in an easy-to-understand manner.

In order to solve the above-mentioned problems, the present invention provides an original translation in a first language and a translation in a second language without preparing an ideal translation or a plurality of translations. A bilingual sentence error detection device that improves the work efficiency of correction and the quality of translated sentences by mechanically comparing and presenting sentences that are likely to be incorrect and the degree of error in an easy-to-understand format to the user. The purpose is to provide.

[0011]

According to the present invention, there is provided a sentence input unit for inputting an original sentence expressed in a first language and a translated sentence obtained by translating the original sentence into a second language, and an original sentence information from the original sentence. An information extracting unit that performs extraction and extraction of translated sentence information from the translated sentence; a comparing unit that compares the original sentence information with the translated sentence information; and a comparing unit that compares the translated sentence based on the comparison result in the comparing unit. An error rate calculation unit that calculates an error rate, and a notification unit that notifies a user of a bilingual sentence of the translated sentence having a high error rate are provided.

With such a configuration, the user can easily determine whether a sentence has a high error rate, and therefore, the work efficiency of correction can be improved.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a bilingual sentence error detecting apparatus according to the present invention will be described below with reference to the drawings.

(Embodiment 1) FIG. 1 is a block diagram of Embodiment 1 of a bilingual sentence error detecting apparatus according to the present invention. This bilingual sentence error detection device includes a sentence input unit 101, an information extraction unit 102, a search unit 103, a dictionary 104, a comparison unit 10
5, an error rate calculation unit 106, and a notification unit 107.

The sentence input unit 101 includes a keyboard for inputting an original sentence expressed in a first language and a translated sentence obtained by translating the original sentence into a second language, a voice input device, and a translated sentence of the original sentence and the original sentence. And a reading device for reading a sentence from a recording medium on which is recorded.

The information extraction unit 102 performs a morphological analysis on the original sentence and the translated sentence notified from the sentence input unit 101 through the search unit 103 with reference to the dictionary 104, divides the words into words, and obtains information of each word. Get. Then, information necessary for comparison is extracted from the information, and notified to the comparing unit 105.

In the dictionary 104, morphological information, syntax information, and semantic information of each of the first language and the second language are registered.

The search unit 103 searches the dictionary 104 using the words of the original sentence notified from the information extraction unit 102 as keywords, and notifies the information extraction unit 102 of the search results. The dictionary 104 is searched using the word of the translation sentence notified from the information extracting unit 102 as a keyword, and the search result is stored in the information extracting unit 1.
02 is notified.

The comparing unit 105 compares the original sentence information and the translated sentence information notified from the information extracting unit 102, and notifies the error rate calculating unit 106 of the comparison result.

The error rate calculator 106 calculates the error rate of the translated sentence based on the error rate notified from the comparator 105.

The notifying unit 107 comprises a CRT for displaying the original sentence, the translated sentence, and the degree of error, an audio output device for presenting the degree of error by sound, and a liquid crystal display, and a means for notifying the translation of the original sentence and the translated sentence. And the degree of error.

For example, it is composed of a screen displaying a bilingual sentence and a screen displaying a setting of a hatching pattern according to an error rate. As a means of reporting the degree of error, the error rate is divided into several stages, different shading patterns are set for each stage, and the shading pattern is determined according to the error rate. Add a shaded pattern corresponding to the background. As a result, the user can discover at a glance the degree of error in the translation. In this example, the degree of error is expressed using a hatched pattern, but may be expressed using a color. Moreover, you may express with the loudness of a sound.

Next, the operation of this embodiment will be described with reference to the flowchart of FIG. When an original sentence and a translated sentence of the original sentence are input from the sentence input unit 101 (S101), the information extraction unit 102 passes through the search unit 103 (S103).
While referring to the dictionary 104 (S104), a morphological analysis is performed on the original sentence and the translated sentence, the words are divided into words, and information on each word is obtained (S102). Then, based on the obtained information, the comparing unit 105 compares the original sentence information and the translated sentence information (S105). Then, based on the error rate notified from the comparing section 105, the error rate calculating section 106 calculates the error rate of the translated sentence (S106), and the notification section 107 notifies the user of the degree of the error rate of the translated sentence. A notification is issued (S107).

According to a sixth aspect of the present invention, there is provided a bilingual sentence error detecting apparatus.
The comparison unit has a configuration as shown in FIG.

According to a third aspect of the present invention, there is provided a bilingual sentence error detecting apparatus.
The comparison unit includes a sentence length comparison unit 601.

According to a fourth aspect of the present invention, there is provided a bilingual sentence error detecting apparatus.
The comparison unit is a morphological comparison unit 602, and in the bilingual sentence error detection device according to claim 5, the comparison unit is a syntactic / semantic analysis unit 60.
3 and a syntax / semantics comparison unit 604.

The sentence length comparing unit 601 compares the number of morphemes consisting only of elements corresponding to content words, that is, the number of morphemes excluding functional words (prepositions, conjunctions, articles, particles, auxiliary verbs, etc.).

The morpheme comparison unit 602 compares the morpheme information of the original sentence with the morpheme information of the translated sentence, and calculates an error rate.

The syntax / semantic analysis unit 603 creates a syntax structure according to the grammar rules, and generates a syntax tree in which semantic information is added to the structure.

The syntactic / semantic comparison unit 604 performs a syntactic comparison, a comparison of morpheme target areas, a stylistic level comparison, and a tense comparison based on the result of the syntactic / semantic analysis, and calculates an error rate. .

Next, the sentence length comparing unit 601 according to the present embodiment.
Will be described with reference to the flowchart of FIG.

The sentence length comparing unit 601 is provided for the morphological analysis unit 10
When the morpheme information of the original sentence and the translated sentence is received from No. 3, the sentence length of the original sentence is written into the storage area Slen1 and the sentence length of the translated sentence is written into the storage area Slen2 (C001, C002).
Next, the sentence lengths Slen1 and Slen2 are compared (C00
3) If the length of the translated sentence is longer than the length of the original sentence,
Since the amount of information has increased due to translation, the error rate is set to 0% (C004). If the length of the translated sentence is shorter than the length of the original sentence, the error rate is reduced to 100 * because the amount of information is reduced.
(1−Slen2 / Slen1)
5).

For example, when the morpheme information shown in FIGS. 5A and 5B is received, the sentence length Slen1 of the original sentence is 8, and the sentence length Slen2 of the translated sentence is 4, and the information amount is reduced by translation. Therefore, the error rate 50
% Can be obtained. The sentence length comparing unit is particularly effective for missing translations in multiple sentences and compound sentences as in the above example.

If it is known that the number of content words between languages is extremely different, the equation for calculating the error rate using the ratio r of the content words between languages is expressed as 100 * (1-Sle).
n2 '/ Slen1), Slen2' = Slen2 *
r.

In the parallel sentence error detecting device according to the third aspect, a dictionary in which only morpheme information of each of the first language and the second language may be registered.

Next, the morphological comparison unit 602 of this embodiment
Will be described with reference to the flowchart of FIG.

The sentence length comparing unit 602 includes the morphological analysis unit 10
3 or the morpheme information of the original sentence and the translated sentence from the sentence length comparing unit 601, the sentence length of the original sentence is stored in the storage area Sle.
The sentence length of the translated sentence is written into the storage area Slen2 in n1 (C001, C002). Next, the number of nouns in the original text,
The number of verbs, the number of adjectives, and the number of adverbs are respectively stored in the storage areas cntN1, cntV1, cntADJ1, and cntAD.
V1 is written (C010), and the number of nouns, the number of verbs, the number of adjectives, and the number of adverbs in the translation are respectively stored in the storage area cn.
tN2, cntV2, cntADJ2, cntADV2
(C011). Next, the numbers of parts of speech in the original sentence and the translated sentence are compared (C012, C015, C018, C
021), if the number of parts of speech is greater in the original sentence, the error rate of the part of speech is 1- (cnt2 / Slen2) / (cn
t1 / Slen1) (C013, C016, C019, C022) If smaller, 1- (cnt1 / Slen1) / (cnt
2 / Slen2) (C014, C017, C020, C023). Finally, considering the error rate of each part of speech, the error rate of the entire sentence is calculated (C024).

For example, when the morpheme information shown in FIGS. 7A and 7B is received, the sentence length Slen1 of the original sentence and the sentence length Slen2 of the translated sentence are both 5, and the number of nouns in the original sentence is 3, the number of verbs is 2, the number of nouns in the translation is 4, and the number of verbs is 1.

When the error rate of each part of speech is calculated by the algorithm shown in FIG. 6, errN is 1- (3/5) / (4/5) = 0.25 errV is 1- (1/5) / (2 /5)=0.5 errADJ and errADV become 0. When the overall error rate is calculated using the error rate of each part of speech, 100 * max
(0.25, 0.5, 0, 0) = 50%.

Next, the syntax / semantics comparison section 6 of the present embodiment.
The operation of Step 04 will be described with reference to the flowchart of FIG.

Upon receiving the English syntax tree and the Japanese syntax tree from the syntax / semantic analysis unit 603, the syntax / semantics comparison unit 604 sequentially extracts parent-child trees having verbs as parents in the English syntax tree, and stores them in the storage area. From EngSubTree (0) to EngS
ubTree (n1), and sequentially extracts parent-child trees having the verb as a parent in the Japanese syntax tree, and stores the extracted data in the storage area JapS.
ubTree (0) to JapSubTree (n2)
(C031). Then, the English parent-child syntax tree is associated with the Japanese parent-child syntax tree (C032 to C039). The associated syntax tree is stored in the storage area Pa
stored in ir (0) to Pair (k) (C0
35).

Next, an area errCntN for storing the number of errors when focusing on nouns and an area errCntV for storing the number of errors when focusing on verbs are initialized (C040). (C04
1) Compare the parent and child trees of English and Japanese (C042), and determine the number of errors errCntN when focusing on nouns and the number of errors errCntV when focusing on verbs (C042). Finally, the error rate in the syntax / semantic comparison unit is calculated from the number of errors errCntN when focusing on nouns and the number of errors errCntV when focusing on verbs.

Next, the detailed operation of the parent-child tree comparison unit (C042) will be described with reference to FIG.

The parent-child tree comparison unit (C042) calculates the number of errors when focusing on nouns (from C050 to C05).
5) and calculation of the number of errors when focusing on verbs (C0
56 to C063).

First, the calculation of the number of errors when focusing on nouns will be described. The noun node of the child of the English parent-child tree is extracted and written to the storage area EngChild (C0
50). If there is no child noun node (C05
1) The processing for calculating the number of errors when focusing on verbs is started. If there is a child noun node, the number of noun nodes c
ntN is incremented (C052), the noun node of the child of the Japanese parent-child tree is extracted, and the storage area JapChi is extracted.
Write to ld (C053). And EngChil
The semantic restriction of d is compared with the semantic restriction of JapanChild (C054). If the semantic restrictions match, the processing from C050 is repeated for the next child. If they do not match, the number of errors errCntN when focusing on the noun is incremented (C055), and C0 is set for the next child.
The processing from step 50 is repeated.

Next, calculation of the number of errors when focusing on verbs will be described. A dictionary search is performed using the parent node of the Japanese parent-child tree as a key, and the result is stored in a storage area Dic.
Writing is performed from tInfo (0) to DictInfo (k) (C056). A storage area err for indicating whether or not the input parent-child tree has an error when focusing on verbs.
Tmp is initialized (C057). It is checked whether or not k is 0 (C058). If k is 0, all the corresponding dictionary information is checked and tracked, so that the storage area errTmp for indicating whether or not there is an error when focusing on the verb is 1 (C062), and if errTmp is 1,
The number of errors when focusing on the verb is incremented (C063), and the process ends. If errTmp is 0, the process ends. If k is not 0, Di
It is checked whether the semantic restriction of the case of ctInfo (k) matches the semantic restriction of the core case of the Japanese parent-child tree (C05).
9) If coincident, decrement k and C058
Is repeated. If they do not match, the storage area errTmp for indicating whether or not there is an error when focusing on the verb is set to 1 (C060), k is decremented, and the processing from C058 is repeated.

For example, a case where the syntax trees shown in FIGS. 10A and 10B are received will be described. However, the dictionaries used include a verb dictionary and a noun dictionary having the information shown in FIGS. 10C and 10D. Since the syntax trees shown in FIGS. 10A and 10B are already corresponding parent-child trees,
Description of the processing before 039 is omitted.

The parent-child tree comparison unit (C042) shown in FIG.
When the corresponding parent-child tree is input, the semantic restriction of the noun node of the English syntax tree and the semantic restriction of the noun node of the Japanese syntax tree match, so the number of errors errCntN when focusing on the noun is: It becomes 0. Next, a dictionary search is performed using "eating" as a key (C056), and information is obtained that the semantic restriction of the nominative case is "living" and the semantic restriction of the target case is "thing". Since the semantic restriction of the nominative case of the Japanese syntax tree is “thing”, the number of errors errCntV when the verb is mainly viewed is one. When the error rate in the syntax / semantic comparison unit is calculated based on the number of errors when focusing on nouns and the number of errors when focusing on verbs, the error rate is 100 * max
(1, 0) = 100%.

Next, a case where the syntax trees shown in FIGS. 10E and 10F are received will be described. FIG. 10 (e) (f)
Since the syntax tree shown in is already a corresponding parent-child tree,
Description of the processing before C039 is omitted. FIG.
When this syntax tree pair is input to the parent-child tree comparison unit (C042) shown in (1), since the semantic restriction of “crane” is “thing” and the semantic restriction of “vine” is “creature”, The error number errCntN of is 1.

Next, a dictionary search is performed using "is" as a key (C056). The information that the semantic restriction of the nominative case is "thing" and the semantic restriction of the target case is "thing", and the semantic restriction of the nominative case is " Information on "creatures" and the meaning of the purpose are obtained as "creatures".

The semantic restriction of the nominative case of the Japanese parse tree is "living", and the semantic restriction of the object case is "thing", which does not match any of the information obtained from the dictionary. The number errCntV becomes 1.

When the error rate in the syntactic / semantic comparison unit is calculated based on the number of errors when focusing on nouns and the number of errors when focusing on verbs, the error rate is 100 * ma
x (0,1) = 100%.

The search unit 104 may search the dictionary 105 using a sequence of two or more words as a keyword. The dictionary 105 may be composed of separate dictionaries of a first language of the second language and a second language of the first language.

(Embodiment 2) FIG. 11 is a configuration diagram of a bilingual sentence error detecting apparatus according to Embodiment 2 of the present invention. This bilingual sentence error detection device includes a document input unit 1010, a sentence correspondence determination unit 1011, an information extraction unit 102, a search unit 10
3, the dictionary 104, the comparison unit 105, and the error rate calculation unit 1
06 and a notifying unit 1070.

Parts having the same configuration as in the first embodiment are denoted by the same reference numerals, and description thereof will be omitted.
Only the components unique to the present embodiment will be described.

A document input unit 1010 includes a keyboard for inputting an original document as a sentence in a first language and a translated document as a result of translating the original document into a second language, a voice input device, and an original document. And a reading device for reading a sentence from a recording medium on which a translated document of the original document is recorded.

The sentence correspondence determination unit 1020 is provided in the document input unit 10
10 is extracted in units of one sentence from the original document notified from the document 10 and the translated document of the original document.
The correspondence between the original sentence and the translated sentence of the original sentence is determined by using the bilingual correspondence determination method described in the publication.

However, if the correspondence is clear as in the result of machine translation, nothing is performed. If the entire sentence is missing at the time of association, “empty sentence” is set, a temporary association is performed, the error rate is set to 100%, and the information is notified to the notification unit 107. . In cases other than the above, a bilingual sentence between the original sentence and the translated sentence of the original sentence is notified to the information extracting unit 102.

The notification unit 1070 is composed of an original document and a translated document, a CRT for displaying the degree of error, a voice output device for presenting the degree of error in a loud volume, a liquid crystal display, and a keyboard. It has means for notifying a bilingual sentence, means for notifying an error location, and means for notifying the degree of error.

For example, the display screen is as shown in FIG.
A screen for displaying a document in a first language, a screen for displaying a document in a second language linked so that correspondence can be understood, and a screen for displaying a setting of a hatching pattern according to an error rate Have been.

As means for informing the degree of error, the error rate is divided into several stages, different shading patterns are set for each stage, and the shading pattern is determined according to the error rate.
A shaded pattern corresponding to the background of the erroneous translation is added. As a result, the user can find a bilingual sentence with a high error rate at a glance.

In this example, the degree of error is represented by using a hatched pattern, but may be represented by using a color. Also, the translated sentences with the highest error rate may be displayed in order. Alternatively, a sentence with a high error rate may be output as speech.

Next, the operation of this embodiment will be described with reference to the flowchart of FIG. When an original document and a translated document of the original document are input from the document input unit 1010 (S1010), it is determined whether or not the document is at the end (S101).
1). If it is the end of the document, the process ends. If it is not the end of the document, the sentence correspondence determination unit 1020 associates the original sentence with the translated sentence (S1012). Then, the sentence correspondence determination unit 10
It is checked whether the bilingual sentence notified from 20 includes “empty sentence” (S1013). If it contains “empty sentence”, the error rate is set to 100% (S111), and the notification unit 1
At 07, the bilingual sentence and the error rate are reported in an easy-to-understand format (S107). If it does not include “empty sentence”, the information extracting unit 102 performs morphological analysis to divide the original sentence and the translated sentence into morphemes while referring to the dictionary 104 through the search unit 103, and extracts information of each word. (S102). Then, using the information of each word notified from the information extracting unit 102, the comparing unit 105 compares the bilingual sentences (S
105). Then, based on the error rate notified from the comparing section 105, the error rate calculating section 106 calculates the error rate of the translated sentence (S106), and the notifying section 107 notifies the bilingual sentence and the error rate in an easy-to-understand manner. (S107). Then, the above processing is repeated until the end of the document comes.

(Embodiment 3) FIG. 14 is a configuration diagram of Embodiment 3 of a bilingual sentence error detecting apparatus according to the present invention. The bilingual sentence error detection device includes a sentence input unit 101, an information extraction unit 102, a search unit 103, a dictionary 104, a sentence length comparison unit 601, a morpheme comparison unit 602, a syntax / semantic analysis unit 603, , A syntax / meaning comparison unit 604, a comparison level input unit 605, a comparison level control unit 606, an error rate calculation unit 106, and a notification unit 107.

Parts having the same configuration as those in the first and second embodiments are denoted by the same reference numerals, and the description thereof will be omitted. Only the components unique to the present embodiment will be described.

The comparison level input unit 605 includes a keyboard for inputting which combination of the sentence length comparing unit 601 and morphological comparing unit 602, and the syntactic / semantic analyzing unit 603 and the syntactic / semantic comparing unit 604 to perform, and in which combination. , A voice input device, and a pointing device such as a CRT, a liquid crystal display, and a mouse for displaying the input comparison level information. Then, the comparison level control unit 606 is notified of the input comparison level.

The comparison level control unit 606 is based on the comparison level notified from the comparison level input unit 605, and based on the sentence length comparison unit 601, the morphological comparison unit 602, the syntax / semantic analysis unit 603, and the syntax / meaning comparison unit 604. Controls which comparisons are made and in which combinations.

Next, the operation of this embodiment will be described with reference to the flowchart of FIG. In addition, parts that perform the same operations as those in the first and second embodiments are denoted by the same reference numerals, and the description thereof is omitted. Only the operation parts unique to the present embodiment will be described.

When the comparison level input unit 605 inputs which combination is to be performed in which combination (S600), the comparison level control unit 606 checks whether there is an input for performing sentence length comparison (S601). If there is input,
The sentence length comparison unit 601 compares the lengths of the original sentence and the translated sentence, and calculates the error rate (S602). If there is no input, this process is skipped. Next, the comparison level control unit 606
It is checked whether there is an input for performing morpheme comparison (S603).
Then, the morpheme information of the original sentence and the translated sentence are compared, and the error rate is calculated (S604). If there is no input, this process is skipped. Finally, the comparison level control unit 606 checks whether there is an input for syntactic comparison (S605). If there is an input, the syntax / semantic analysis unit 603 generates a syntax tree of the original sentence and the translated sentence (S606). ), Syntax and semantic comparison unit 6
04, these syntax trees are compared, and the error rate is calculated (S
607). If there is no input, these processes are skipped. Then, the error rate calculator 106 calculates the error rate of the translated sentence by summing up the error rates obtained by the above comparison (S10).
6).

(Embodiment 4) FIG. 16 is a block diagram of a bilingual sentence error detecting apparatus according to Embodiment 4 of the present invention. This bilingual sentence error detection device includes a document input unit 1010, a sentence correspondence determination unit 1020, an information extraction unit 102, a search unit 10
3, the dictionary 104, the comparison unit 105, and the error rate calculation unit 1
06, the notification unit 107, the error pattern automatic correction location determination unit 310, the error pattern automatic correction unit 311, the man-machine interface unit 312, the error pattern database control unit 313, and the error pattern database 3
14 is provided.

Parts having the same configuration as in the first, second, and third embodiments are denoted by the same reference numerals, and description thereof will be omitted. Only the components unique to the present embodiment will be described.

The man-machine interface 312
An interface portion for allowing the user to make some corrections or check according to the degree of error notified to the notifying section 107, and comprises a CRT, a liquid crystal display, a pointing device such as a keyboard and a mouse.

Error pattern database control section 313
Considers the translated sentence corrected by the user in the man-machine interface unit 312 as an error sentence, and registers the original sentence, the translated sentence, the corrected translated sentence, and the error rate in the error pattern database 314 as an error pattern. Or the error pattern database 314 is changed, or the error pattern is deleted from the error pattern database 314.

The error pattern database 314 has registered therein error patterns. The error pattern automatic correction portion determining unit 310 searches the error pattern database 314 using the bilingual sentence as a key to check whether the bilingual sentence notified from the sentence correspondence determining unit 1020 has a bilingual sentence in the error pattern database 314. Then, the search result is notified to the error pattern automatic correction unit 311.

When the search result notified from the error pattern automatic correction portion judging section 310 is an error pattern, the error pattern automatic correction section 311 replaces the translated sentence with the corrected translated sentence in the error pattern.

Next, the operation of this embodiment will be described with reference to the flowchart of FIG. In addition, portions that perform the same operations as those in the first, second, and third embodiments are denoted by the same reference numerals, and the description thereof is omitted. Only the operation portions unique to the present embodiment will be described.

The error pattern automatic correction location judging section 310 searches the error pattern database using the bilingual sentence notified from the sentence correspondence judging section 1020 as a key (S30).
0), it is checked whether or not the bilingual sentence exists in the error pattern database (S301). If there is information, the error pattern automatic correction section 311 uses the search result of the error pattern automatic correction section determination section. Then, the translation of the input translation is automatically corrected (S304), and the process proceeds to the next translation.

If there is no information, the processing from the information extracting unit 102 is continued. It is checked whether or not the user has corrected the bilingual sentence notified to the notifying unit 107 using the man-machine interface unit 312 (S30).
2) If the correction has been made, the error pattern database control unit 313 registers the original sentence, the translated sentence, the corrected translated sentence, and the error pattern such as the error rate of the translated sentence in the error pattern database 314 ( S303) The process proceeds to the next bilingual sentence. If no correction has been made, the process proceeds to the next parallel sentence. Then, the above processing is repeated until the end of the document comes.

For example, when the error pattern database shown in FIG. 18 is searched using the third parallel sentence shown in FIG. 12 as a key, the corrected translation in that database is already registered in the error pattern database. Using the sentence, the bilingual sentence is automatically corrected (FIG. 19).

Further, when the error pattern database shown in FIG. 18 is searched using the fourth parallel sentence shown in FIG. 19 as a key, the error determination processing is performed because the parallel sentence has not been registered in the error pattern database yet. , An error rate of 100% is obtained, and the display is updated as shown in FIG.

When the user corrects the fourth translated sentence with a hatched pattern shown in FIG. 20 as “acid corrodes metal”, the translated sentence, the corrected translated sentence, and an error rate of 100% are incorrect. It is registered in the pattern database (FIG. 21).

(Embodiment 5) FIG. 22 is a configuration diagram of a bilingual sentence error detecting apparatus according to Embodiment 5 of the present invention. This bilingual sentence error detection device includes a document input unit 1010, a sentence correspondence determination unit 1020, an information extraction unit 102, a search unit 10
3, the dictionary 104, the comparison unit 105, and the error rate calculation unit 1
06, the notification unit 107, the error pattern automatic correction location determination unit 310, the error pattern automatic correction unit 311, the man-machine interface unit 312, the error pattern database control unit 313, and the error pattern database 3
14 and an automatic correction inquiry unit 315.

Parts having the same configuration as in the first, second, third, and fourth embodiments are denoted by the same reference numerals, and description thereof will be omitted. Only the components unique to the present embodiment will be described. I do.

The automatic correction inquiry section 315 inquires of the user whether or not to make automatic correction using the search result notified from the error pattern automatic correction location judging section 310, determines whether or not to make automatic correction, and makes an automatic correction. Is performed, the error pattern and the bilingual sentence notified from the sentence correspondence determining unit 1020 are notified to the error pattern automatic correcting unit 311. If automatic correction is not performed, the sentence correspondence determining unit 1011 is used.
The bilingual sentence notified from 20 is notified to the information extracting unit 102.

Next, the operation of this embodiment will be described with reference to the flowchart of FIG. In addition, parts that perform the same operations as those in the first, second, third, and fourth embodiments are denoted by the same reference numerals, and the description thereof is omitted. Only the operation parts unique to the present embodiment will be described.

Error pattern automatic correction point search section 310
Then, it is checked whether or not the bilingual sentence notified from the sentence correspondence determining unit 1020 is in the error pattern database 314 (S300, S301). If the bilingual sentence is in the error pattern database 314, the automatic correction inquiry unit 315 Then, an inquiry is made to the user as to whether or not to perform automatic correction (S305). If the user selects automatic correction, the error pattern automatic correction unit 311 performs automatic correction (S304), and the user selects not to perform automatic correction. If so, the processing from the information extraction unit 102 is continued.

(Embodiment 6) FIG. 24 is a block diagram of a bilingual sentence error detecting apparatus according to Embodiment 6 of the present invention. This bilingual sentence error detection device includes a document input unit 1010, a sentence correspondence determination unit 1020, an information extraction unit 102, a search unit 10
3, the dictionary 104, the comparison unit 105, and the error rate calculation unit 1
06, a notification unit 107, an error pattern automatic correction location determination unit 310, an error pattern automatic correction unit 311, a man-machine interface unit 312, an error pattern control unit 313, an error pattern database 314, and an automatic correction inquiry unit. 315 and the correct answer pattern determination unit 400
And a correct answer pattern database control unit 401 and a correct answer pattern database 402.

In the first, second, third, fourth and fifth embodiments,
Portions having the same configuration as those described above are denoted by the same reference numerals, and description thereof will be omitted. Only components specific to the present embodiment will be described.

The correct pattern judging section 400 searches the correct pattern database 402 using the bilingual sentence notified from the sentence correspondence judging section 102 as a key. Move on to parallel translation processing.

Correct answer pattern database control section 401
Considers a bilingual sentence specified by the user as a correct bilingual sentence in the man-machine interface unit 311 as a correct pattern, registers the bilingual sentence in the correct pattern database 402, updates / changes the correct pattern database 402, The correct pattern is deleted from the correct pattern database 402.

The correct answer pattern database 402 has registered therein correct answer patterns. Next, the operation of the present embodiment will be described with reference to the flowchart of FIG. Note that the same reference numerals are given to portions that perform the same operations as those in the first, second, third, fourth, and fifth embodiments, and description thereof is omitted. Only the operation portions unique to the present embodiment will be described. .

The correct pattern judging section 400 checks whether or not the bilingual sentence notified from the sentence correspondence judging section 102 exists in the correct pattern database 402 (S400, S40).
1) If the bilingual sentence is in the correct pattern database 402, the bilingual sentence is regarded as a correct bilingual sentence, and the process proceeds to the next bilingual sentence. If the bilingual sentence is not in the correct answer pattern database 402, an error automatic correction process or an error detection process is performed. Then, the above processing is repeated until the end of the document comes.

For example, as shown in FIG. 26, the error rate of the sixth translated sentence is displayed as 31% or more and 50% or less, but the user thinks that the translated sentence is correct, and designates the correct answer. In this case, the original sentence and the translated sentence are registered as the correct answer pattern in the correct answer pattern database (FIG. 27).

Thereafter, when the above-mentioned translated text is input, since the translated text has already been registered in the correct answer pattern database, the error determination processing is not performed. Note that the configuration may be such that the automatic correction inquiry unit 315 is omitted.

Further, the bilingual sentence error detecting apparatus according to the first to sixth embodiments is realized by a program, which is recorded on a recording medium such as a floppy disk and transferred, or transferred through a communication line. By doing so, it can be easily implemented by another independent computer system.

[0096]

As described above, according to the present invention, a sentence input unit for inputting an original sentence expressed in a first language and a translated sentence obtained by translating the original sentence into a second language, an original sentence from the original sentence An information extracting unit that extracts information and extracts translated sentence information from the translated sentence; a comparing unit that compares the original sentence information with the translated sentence information; and An error rate calculator for calculating an error rate of the translation,
A notification unit for transmitting a bilingual sentence of the translated sentence having a high error rate to the user is provided, so that the original sentence of the first language and the translated sentence can be prepared without preparing an ideal translated sentence or a plurality of translated sentences. Improve the efficiency of correction work by mechanically comparing the original sentence into a second language and presenting the sentence that is likely to be incorrect and the degree of error in a user-friendly format. Can be.

According to the ninth aspect of the present invention, there is further provided a man-machine interface unit for correcting an erroneous part by a user, and a bilingual sentence corrected by the user.
A sentence in the first language, a sentence in the second language translated from the sentence in the first language, a corrected translated sentence in which the user has modified the sentence in the second language, and an error rate An error pattern database that holds the error pattern database, an error pattern database control unit that controls access to the error pattern database, an error pattern automatic correction location determination unit that determines whether a translation is to be automatically corrected, and the error pattern automatic An error pattern automatic correction unit that automatically corrects a bilingual sentence determined to be automatically corrected by the correction part determination unit by using information of the error pattern database is provided. For
Once the error is corrected, the information is registered in the error pattern database, and the correction is automatically performed from the next time, so that the burden of the user on the correction work can be reduced.

[0098] In the invention of claim 11, further, a man-machine interface unit for inputting information that a user will treat the translated sentence displayed as having an error as a correct sentence in the future, A correct pattern database that holds the translated text input as a correct pattern when the user treats it as a correct sentence, a correct pattern database controller that controls access to the correct pattern database, and determines a correct pattern in the input document. Since a correct pattern determination unit is provided, an error display can be prevented from being performed for a sentence that the user wants to treat as a correct translated sentence.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a bilingual sentence error detection device according to a first embodiment of the present invention;

FIG. 2 is a flowchart illustrating an operation of the embodiment.

FIG. 3 is a diagram showing components of a comparison unit according to the embodiment.

FIG. 4 is a flowchart for explaining the operation of a sentence length comparing unit according to the embodiment;

FIG. 5 is a diagram showing a part of the morphological information notified from the morphological analyzer of the embodiment.

FIG. 6 is a flowchart illustrating an operation of a morphological comparison unit according to the embodiment.

FIG. 7 is a diagram showing a part of the morpheme information notified from the morpheme analysis unit or the sentence length comparison unit of the embodiment.

FIG. 8 is a flowchart for explaining the operation of the syntax / semantics comparison unit of the embodiment.

FIG. 9 is a flowchart for explaining the operation of the syntax / semantics comparison unit of the embodiment.

FIG. 10 is a diagram illustrating a part of a syntax tree notified from the syntax and semantic analysis unit according to the embodiment;

FIG. 11 is a configuration diagram of a bilingual sentence error detection apparatus according to a second embodiment of the present invention.

FIG. 12 is a diagram illustrating an example of a configuration of a screen of a display unit in the embodiment.

FIG. 13 is a flowchart illustrating the operation of the embodiment.

FIG. 14 is a configuration diagram of a bilingual sentence error detection device according to a third embodiment of the present invention.

FIG. 15 is a flowchart illustrating an operation of the embodiment.

FIG. 16 is a configuration diagram of a bilingual sentence error detection device according to a third embodiment of the present invention.

FIG. 17 is a flowchart illustrating an operation of the embodiment.

FIG. 18 is a diagram showing an example of the content example of an error pattern database according to the embodiment.

FIG. 19 is a diagram illustrating an example of a screen configuration of a display unit according to the above embodiment.

FIG. 20 is a diagram illustrating an example of a screen configuration of a display unit according to the above embodiment.

FIG. 21 is a diagram showing an example of the content example of an error pattern database according to the embodiment.

FIG. 22 is a configuration diagram of a bilingual sentence error detection apparatus according to a fourth embodiment of the present invention.

FIG. 23 is a flowchart illustrating the operation of the above embodiment.

FIG. 24 is a configuration diagram of a bilingual sentence error detection apparatus according to a fifth embodiment of the present invention.

FIG. 25 is a flowchart illustrating the operation of the above embodiment.

FIG. 26 illustrates an example of a screen configuration of a display unit in the above embodiment.

FIG. 27 is a diagram showing an example of the content example of a correct answer pattern database according to the embodiment.

[Explanation of symbols]

 Reference Signs List 101 sentence input unit 102 information extraction unit 103 search unit 104 dictionary 105 comparison unit 106 error rate calculation unit 107 notification unit 1010 document input unit 1020 sentence correspondence determination unit 601 sentence length comparison unit 602 morpheme comparison unit 603 syntax / meaning analysis unit 604 Syntax / meaning comparison section 605 Comparison level input section 606 Comparison level control section 310 Error pattern automatic correction section determination section 311 Error pattern automatic correction section 312 Man-machine interface section 313 Error pattern database control section 314 Error pattern database 315 Automatic correction inquiry section 400 Correct answer pattern determination unit 401 Correct answer pattern database control unit 400 Correct answer pattern database

Claims (11)

[Claims] 1. A sentence input unit for inputting an original sentence expressed in a first language and a translated sentence obtained by translating the original sentence into a second language, extraction of original sentence information from the original sentence, and translation from the translated sentence An information extracting unit for extracting sentence information; a comparing unit for comparing the original sentence information and the translated sentence information; and an error rate for calculating an error rate of the translated sentence based on a comparison result in the comparing unit. A bilingual sentence error detection device comprising: a calculating unit; and a notifying unit that notifies a user of a bilingual sentence of the translated sentence having a high error rate. 2. A document input unit for inputting an original document as a set of original sentences expressed in a first language and a translated document as a set of translations obtained by translating the original sentence into a second language; A sentence correspondence determining unit that associates the original sentence with the translated sentence from the translated document to obtain a translated sentence; and extracts original sentence information from the original sentence and extracts translated sentence information from the translated sentence. An information extraction unit to perform; a comparison unit that compares the original text information with the translation text information; an error rate calculation unit that calculates an error rate of the translated text based on a comparison result in the comparison unit; 2. A bilingual sentence error detecting apparatus according to claim 1, further comprising: a notifying unit that notifies a user of a bilingual sentence of the translated sentence with a high level. 3. The information extracting unit is realized by a sentence length extracting unit that extracts a sentence length of the original sentence and the translated sentence, and the comparing unit is configured to extract the sentence length of the original sentence extracted by the sentence length extracting unit. 2. The bilingual sentence error detecting apparatus according to claim 1, wherein the sentence length comparing unit compares the sentence length with the sentence length of the translated sentence. 4. A dictionary that stores information in a first language and information in a second language, and a search unit that searches the dictionary to extract information, wherein the information extraction unit A morphological analysis unit that performs morphological analysis of the original sentence and the translated sentence using information is implemented. The comparing unit compares morphological information of the original sentence obtained from the morphological analysis unit with morphological information of the translated sentence. 2. The bilingual sentence error detection device according to claim 1, wherein the bilingual sentence error detection device is implemented by a morphological comparison unit. 5. A dictionary that stores information of a first language and information of a second language, and a search unit that searches the dictionary and retrieves information. A morphological analysis unit that performs morphological analysis of the original sentence and the translated sentence using information and a syntactic and semantic analysis unit that performs syntactic and semantic analysis using morphological information obtained by the morphological analysis unit, A comparing unit configured to compare a syntax of the original sentence obtained from the syntactic / semantic analysis unit with a syntax of the translated sentence; and a target area of a morpheme of the original sentence and a target area of a morpheme of the translated sentence. 2. The bilingual sentence error detecting apparatus according to claim 1, wherein the bilingual sentence error detecting device is realized by a semantic comparison unit that performs comparison and compares the style of the original sentence with the style of the translated sentence. 6. The information extracting unit is realized by a sentence length extracting unit, a morphological analyzing unit, and a syntactic / semantic analyzing unit, and the comparing unit is a sentence length comparing unit, a morphological comparing unit, 2. The bilingual sentence error detecting device according to claim 1, wherein the bilingual sentence error detecting device is realized by a comparing unit and a meaning comparing unit. 7. A comparison level input section for inputting a level of a combination of a sentence length comparison section, a morphological comparison section, a syntax comparison section, and a semantic comparison section, and a comparison for selecting a process according to the comparison level. 2. The apparatus according to claim 1, further comprising a level determination unit.
A bilingual sentence error detection device described. 8. The bilingual sentence error detection apparatus according to claim 1, wherein the notification unit changes the notification mode according to the degree of the error. 9. A man-machine interface unit for a user to correct a translated sentence of the translated sentence for a translated sentence with a high error rate notified by the notifying unit, an original sentence of the translated sentence, and An error pattern database that retains the translated sentence before correction, the translated sentence of the translated sentence, and the error rate as an error pattern; an error pattern database control unit that controls access to the error pattern database; An error pattern automatic correction location determination unit that determines whether the translation is to be performed, and the bilingual text determined that the automatic correction should be performed by the error pattern automatic correction location determination unit, based on information in the error pattern database. 2. The bilingual sentence error detecting apparatus according to claim 1, further comprising: an error pattern automatic correcting unit that automatically corrects the error using the error pattern. 10. A man-machine interface unit for a user to correct a translated sentence of the translated sentence with respect to a translated sentence with a high error rate notified by the notifying unit, an original sentence of the translated sentence, A translated sentence before correction, a translated sentence of the translated text, and an error pattern database that holds the error rate as an error pattern; an error pattern database control unit that controls access to the error pattern database; An error pattern automatic correction location determining unit that determines whether the translation is to be performed, and whether or not the parallel text determined to be automatically corrected by the error pattern automatic correction location determination unit may be automatically corrected. An automatic correction inquiry section for inquiring of the user whether or not the bilingual sentence is automatically sent in response to a user response to the automatic correction inquiry section. Translated sentence error detecting apparatus according to claim 1, further comprising a error pattern automatic correction unit for correcting manner. 11. A man-machine interface unit for inputting information that the user will treat the translated sentence as a correct sentence in response to the translated sentence in which the notification unit indicates that there is an error, and A correct pattern database that holds the translated text input as a correct pattern to be handled, a correct pattern database control unit that controls access to the correct pattern database, and a correct pattern determination unit that determines the correct pattern in the input document. The bilingual sentence error detection device according to claim 1 provided.