JP2003330924A - Translation support device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an easy-to-understand support on translation of an input sentence to a second language sentence by emulating a relationship between a first language sentence existing in a database and a second language sentence which is a translated result of the first language sentence. <P>SOLUTION: A translation support device comprises input means 2, similar translation example retrieving means 3 for retrieving a similar translation example from a translation example database, similarity calculation method selecting means 4 for selecting a similarity calculation method, ranking means 6 for indicating a plurality of sentences in an order of the similarity, retrieval result indicating means 7 for indicating a translation result as a combination of three sentences of a first language sentence translation object, a first language sentence translation example, and a second language sentence translation example, corresponding information retrieving means 11 for indicating with highlights a word/phrase made to correspond among the combination of the three sentences of the translation object, the first language sentence translation example, and the second language sentence translation example, highlight control means 12, when selecting one of the highlighted words/phrases, for highlighting the one more strongly than others, and translation example selecting means 8. <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a translation support device, and more particularly, to a relationship between a first language sentence (for example, English sentence) already existing in a database and a second language sentence (for example, Japanese sentence) translated from the first language sentence. The present invention relates to a translation support device that supports translation of an input sentence in a first language into a second language sentence in an easy-to-understand manner.

[0002]

2. Description of the Related Art For example, when a revised manual is translated into one of the translations using a computer, the document before the revision and the document after the revision (translation target sentence) are compared and similar to the translation target sentence. There is a translation support device that searches for a translation example and presents it to the user.

For example, as shown in (A) of FIG.
"Ihave a ba written in a language (eg English)
g. When translating the translation target sentence "," the translation support device (not shown) searches the database for an already created English translation example and its second language sentence (for example, Japanese) similar to this translation target sentence. , Display to the user. At this time, as shown in FIG. 5A, a plurality of search results are displayed to the user. In this case, the user cannot change the calculation method of the similarity that determines the display order. Figure 5 (A)
In the case of, since not only the number of matching words but also the similarity is calculated by including the number of consecutively matching words,
Will be displayed first.

In the conventional translation support device, the translation target sentence, which is the three elements of the search result, the first language sentence of the translation example, and the second language sentence of the translation example are “search target sentence” and “translation example”. First
It was displayed separately in two areas, "language sentence and second language sentence".

In the conventional translation support device, among the translation target sentence, the first language sentence of the translation example, and the second language sentence of the translation example, only the translation target sentence and the first language sentence of the translation example correspond to each other in the sentence. I highlighted the words / phrases I want to do.

In the conventional translation support apparatus, the correspondence information between the translation target sentence and the first language sentence of the translation example is used for the similarity calculation, but the correspondence information of the second language sentence of the translation example is not used.

[0007]

When a plurality of search results are displayed, the similarity calculation method is based on the number of consecutively matched words rather than being fixed based on the number of matched words, for example. It may be preferable to judge. Therefore, it is desired that the user can select the similarity calculation method and change the order of displaying the search results.

It is easier to judge the display of the search results by displaying them together rather than by dividing them into two areas, that is, the "search target sentence" and the "first language sentence and the second language sentence of the translation example". Therefore, it is desirable to display them without separating them.

Further, the first sentence of the translation target sentence and the translation example of the search result
Rather than highlighting the corresponding words and phrases in each other's sentences only for language sentences, words in all sentences of the translation target sentence in the search results, the first language sentence of the translation example, and the second language sentence of the translation example It is required to highlight the phrase. It is desired that when a highlighted word / phrase in any sentence is selected, the corresponding word / phrase in the remaining two sentences is further highlighted.

Further, it is desired to calculate the degree of similarity by using information of corresponding words / phrases in all the sentences of the translation target sentence, the first language sentence of the translation example, and the second language sentence of the translation example.

Therefore, an object of the present invention is to provide a translation support device that enables these things.

[0012]

The principle of the present invention is shown in FIG. In FIG. 1, 1 is a translation example database, 2 is input means, 3 is similar translation example retrieval means, 4 is similarity degree calculation method selection means, 5 is similarity degree calculation means, 6 is ranking means, 7
Is a search result display unit, 8 is a translation example selection unit, 9 is an output unit, 11 is a corresponding information search unit, and 12 is a highlight control unit.

The above object of the present invention can be achieved by the following (1) and (2).

(1) Input means 2 for inputting a sentence to be translated
And a similar translation example retrieval means 3 for retrieving a similar translation example from the translation example database 1 using the translation target sentence as a search key.
And a similarity calculation method selection unit 4 for selecting a plurality of similarity calculation methods prepared to determine the order of displaying a plurality of search results, and the search results in order according to the similarity calculated by the similarity calculation method. Ranking means 6 for displaying a plurality of items
And a search result display unit 7 for displaying the search results as a set of a translation target sentence of the first language sentence, a first language sentence of the translation example, and a second language sentence of the translation example, and a translation target sentence, First translation example
Correspondence information search means 11 for highlighting and displaying the words / phrases associated between the three sets of the language sentence and the second language sentence of the translation example, and the highlighted word / phrase. A highlight control unit 12 that highlights and displays the other two by selecting one of them, and a sentence / phrase in the second language of the translation example selected from a plurality of search results.
A translation support device, comprising: a translation example selection unit 8 for outputting a word.

(2) The similarity calculation method associates three sets of the translation target sentence, the first language sentence of the translation example, and the second language sentence of the translation example in the similar translation example retrieval result. 2. The translation support device according to claim 1, wherein the number of words / phrases that have been added are weighted and added together.

As a result, the following effects can be obtained.

(1) According to the present invention, the result of searching the similar translation example database with the input sentence to be translated as a search key is used to change the ranking by selecting an appropriate similarity calculation method or to translate the sentence to be translated. , The first language sentence and the second language sentence in the translation example are highlighted by displaying the word / phrase correspondence of the triplet, or ranked by using the similarity using the triplet information. The translation example search result can be effectively used, and it becomes possible to support the translation work with higher accuracy and speed.

(2) By weighting the number to be added according to conditions such as the part of speech and the number of consecutively matching words, it is possible to finely adjust the degree of similarity, to finely determine the degree of similarity, and to improve accuracy. High translation support can be provided.

[0019]

BEST MODE FOR CARRYING OUT THE INVENTION The outline of the present invention will be described with reference to FIG. In the following description, an example in which English is used as the first language and Japanese is used as the second language is shown, but the language of the present invention is not limited to these.

As a sentence to be translated, that is, a sentence to be translated, as shown in FIG. 5B, "I have a ba"
g. Is entered as a search key, a similar translation example database (not shown) is searched, and the first language sentence of the translation example of this translation target sentence, which is shown from among the translation examples that have been input in advance, is displayed. I have a pe
n. The second language sentence “I have a pen.” Is displayed as a triplet. At this time, the corresponding words “have”, “have”, and “have” are highlighted brighter than other words, for example. In addition, “pen” and “pen” corresponding to each other are also highlighted by a method such as changing the display color so that they can be distinguished from “have” and “hold”.

Therefore, as shown in FIG. 5 (B), the user looks at the sentence displayed in three sets and says "b".
Recognizing that "ag" and "pen" are different,
Since it recognizes that the "n" part is the "pen" of,
It is understood that it is sufficient to translate the “pen” of “to” into the “bag”, which is the Japanese translation of “bag” to, so this is output. In this way, extremely effective translation support can be provided.

An embodiment of the present invention will be described with reference to FIGS. FIG. 1 is an embodiment of the present invention, FIG. 2 is a translation example database explanatory diagram, FIG. 3 is a ranking display example of similar translation search results, and FIG. 4 is a search result display state explanatory diagram.

In FIG. 1, the translation example database 1 is
The translation example to be searched is registered, and as shown in FIG. 2, the search target sentence expressed in the first language sentence and the translated sentence expressed in the second language sentence are registered in association with each other. Therefore, for example, many English sentences and Japanese sentences are associated and registered.

At this time, as will be described later with reference to FIG. 2, the English sentence and the Japanese sentence are divided into words and phrases, respectively.
The words and phrases are registered in association with each other using language information that shows English words and phrases and Japanese words and phrases as correspondence information.

The input means 2 is for inputting a sentence to be translated, that is, an input sentence, and is composed of, for example, a keyboard of a terminal device such as a personal computer or a mail receiving unit, and the first language sentence such as English or Japanese. The input sentence of is input.

The similar translation example retrieval means 3 retrieves the translation example database 1 using the translation target sentence as a search key to obtain a plurality of similar translation examples. For example, as shown in FIG. 5A, a translation target sentence “I have a bag.” Is searched with a search key, and a translation example “I have a pe” is obtained.
n. , "I have a bag." And their Japanese sentences.

The similarity calculation method selection means 4 determines the similarity calculation method used for ranking display when the user selects it. For example, as shown in FIG.
CORE "," RATE "," WORDS "," LIN "
When one of the “KS” is selected with the mouse, for example, it is determined which similarity calculation method is selected.

Here, "SCORE" is the sentence to be translated (first
Shows how many consecutive word matches between the (language sentence) and the translation example (first language sentence) appear (rank 1
Is 4 consecutive, rank 2 is 2 consecutive, for example 4 ² ÷
2 + 2 = 10, ^{2 2} ÷ 2 + 2 = 4)),
“RATE” indicates what percentage of the sum of the numbers of both words the words that match the translation target sentence and the translation example correspond to. “WORDS” indicates the number of matching words in the translation target sentence and the translation example. And "LINKS" is
It shows the number of matches between the English of the translation target sentence and the English and Japanese of the translation example. In rank 1 shown in FIG. 3, (hav
There are two links (e-have-hold) and (pen-pen-pen), but rank 2 shows that there is only one link (pen-pen-pen).

The similarity calculation means 5 calculates the similarity for each of the plurality of search results based on the selected similarity calculation method. For example, when "WORDS" is selected as the similarity calculation method, as shown in FIG. 3, the translation target sentence is "I have a pen."
Is input, and the translation example database 1 is searched using this as a search key to search for “I have a pen while”.
I love. : I have my favorite pen. "
And "I am a pen .: I am a pen.", The similarity calculation means 5 causes the translation target sentence "I h.
ave a pen. , ”And the first language sentence“ I
have a pen which I hav
e. , "I am a pen." And calculated that the number of matching words was 4 in the former and 3 in the latter,
As shown in FIG. 3, the former is ranked 1 and the latter is ranked 2.
Will be displayed as.

The ranking means 6 sequentially processes a plurality of translation examples obtained as a result of searching the translation example database 1 based on the respective similarities calculated by the similarity calculation means 5, as shown in FIG. It is something to display.

As shown in FIG. 3, the search result display means 7 sets each of the plurality of search results as three sets of the translation target sentence, the first language sentence of the translation example, and the second language sentence of the translation sentence, as shown in FIG. It is something to display. At this time, the corresponding words and phrases in all the sentences of the translation target sentence, the first language sentence of the translation example, and the second language sentence of the translation example are highlighted. When a highlighted word / phrase in any sentence is selected, the corresponding word / phrase in the remaining two sentences is further highlighted. Therefore, it has a correspondence information retrieval means 11 and a highlight control means 12 which will be described later.

The translation example selecting means 8 outputs the sentence / phrase / word of the second language sentence of the translation example selected by the user from the plurality of search results. The user selects the second language sentence of the translation example by an instruction means such as a mouse.

The output means 9 outputs the translation result of the translation target sentence, that is, the input sentence. For example, FIG. 5 (B)
When "I have a bag." Is input as the translation target sentence and "I have a pen." Is selected as the second language sentence of the translation example, this is output as the translation support result. . Then, the user converts the pen into a bag and the translation result is "I have a bag."

The structure of the translation example database 1 will be further described. As shown in FIG. 2, the translation example database 1 includes
Search target sentences in the first language (English in this example), translated sentences in the second language (Japanese in this example), and these English sentences 21,
Correspondence information 23 indicating the corresponding English words / phrases obtained by dividing the Japanese sentence 22 into words / phrases and the Japanese words / phrases is registered.

Corresponding information 23 includes corresponding words / phrases and corresponding parts indicating the positions of these words / phrases. No. 2 shown in FIG. In the example of 1, the correspondence information is that "This" starting from the first character of the English sentence corresponds to "this" starting from the first character of the Japanese sentence, and the eleventh character of the English sentence (the number of characters in the blank part is also counted. It is registered that “pen” beginning with) and “pen” beginning with the fourth character of the Japanese sentence correspond to each other. In the example of FIG. 3, “=” indicates that they correspond, and the number indicates the starting character position in each sentence of that character.

FIG. 3 shows an example in which the search result display means 7 rank-displays the similar translation example search results in accordance with the ranking order sent from the ranking means 6.

In FIG. 3, the translation target sentence 31 is 3 together with the first language sentence 33 of the translation example and the second language sentence 34 of the translation example.
The search results are displayed in the area 32 displaying one search result.

In this area 32, there is a calculation method selection area 35 for selecting a similarity calculation method, and the "SCORE",
The user selects a similarity calculation method such as “RATE”, “WORDS”, and “LINKS” with, for example, a mouse and uses it for ranking display.

The area 32 is displayed in ranking according to the similarity selected in the calculation method selection area 35.
The above shows the search results with high similarity.

As shown in FIG. 3, when "I have a pen." Is input as the translation target sentence 31 and "WORDS" is selected as the similarity calculation method, "I have a pen" is selected as a translation example. whi
ch I love. , "And“ I am a pen. ”Are searched, the number of words that match the sentence to be translated is
"I have a pen while I lov
e. 4 "in the case of""and 3 in the case of" I am a pen. "Therefore, the ranking as shown in FIG. 3 is performed and displayed.

Next, the highlight will be described with reference to FIG. FIG. 4 is an enlarged view of a region 32 displaying one search result. The translation target sentence 41 is displayed as a triplet of the first language sentence 42 of the translation example and the second language sentence 43 of the translation example as a result of the search. That is, these are arranged close to the vertical position.

In the case of FIG. 4, the translation target sentence "pen" indicated by 44, the first language sentence "pen" of the translation example, and the second language sentence "pen" of the translation example are these translations. Target sentence,
The first language sentence of the translation example and the second language sentence of the translation example are all corresponding words / phrases, and "have" of the translation target sentence indicated by 45 and the first language of the translation example Sentence "ha
“Ve” and “holding” of the second language sentence of the translation example are also one of the corresponding words / phrases in all the sentences. Therefore, "pen-pen-pen" is highlighted in the same color in a light color, and "have-have-hold" is highlighted in a different color or the same color as the "pen".

From these, it can be seen that the above-mentioned "pen-pen-pen" corresponds, and that the above-mentioned "have-have-hold" also corresponds. In this case, the correspondence between the word / phrase of the first language sentence of the translation example and the word / phrase of the second language sentence of the translation example is obtained by the correspondence information search means 11 recognizing the correspondence information shown in FIG. The highlight control means 12 performs the highlight display based on this correspondence information. Then, the correspondence between the word / phrase of the first language sentence of the translation target sentence and the word / phrase of the first language sentence of the translation example is performed by the similar translation example search means 3.

Then, when displayed as shown in FIG. 4, the user inputs the word / phrase "pen" of the translation target sentence 41 into
For example, when the mouse is selected, the corresponding words / phrases “pen” and “pen” 47 in the remaining two sentences are highlighted more strongly. The highlight control is performed by the correspondence information search means 11 and the highlight control means 12. This makes it easy to understand the correspondence between words and phrases between languages, which is effective for translation work.

In FIG. 4, “a” and “a” indicated by 48 indicate the correspondence between the translation target sentence 41 and the words / phrases of the first language sentence 42 of the translation example, and “love” and “a” indicated by 49. "I love you" indicates the correspondence between the first language sentence 42 of the translation example and the second language sentence of the translation example.
“N”, “have”, etc. are highlighted so that they can be distinguished by different colors. In this way, it helps the user to recognize the corresponding portions and non-corresponding portions between the translation target sentence and the translation example.

With reference to FIG. 4, the first sentence of the translation target sentence and the translation example
A description will be given of a similarity calculation method based on "LINKS" that obtains the similarity using information of corresponding words / phrases in all sentences of the language sentence and the second language sentence of the translation example.

Only the word correspondence will be described. First, the corresponding words in all the sentences of the translation target sentence 41, the first language sentence 42 of the translation example, and the second language sentence 43 of the translation example in the search result are counted. In FIG. 4, since “pen-pen-pen” indicated by 44 and “have-have-hold” indicated by 45 correspond to this, the total is 2, which is the similarity. Actually, finely adjusting the degree of similarity in advance so as to weight the number to be added according to conditions such as part-of-speech, for example, if noun 1, verb is 0.7, auxiliary verb 0.5 is weighted, In this example, the similarity is 1.7. By weighting in this way, it is possible to finely adjust the degree of similarity in which part of speech having a matched part is to be emphasized.

The operation of the present invention will be described.

(1) The user uses the input means 2 shown in FIG. 1 to input, for example, "Ihave a pe" as a translation target sentence.
n. Enter ".

(2) The similar translation example search means 3 searches the translation example database 1 using this input sentence as a search key,
As shown in FIG. 3, “I have a pen whi”
chI love. I have my favorite pen. "
And "I am a pen. I am a pen."

(3) The user selects "WOR" as the similarity calculation method from the calculation method selection area 35 of the display screen of the terminal device.
D ”is selected and sent to the similarity calculation means 5, and the number of matching words between the translation target sentence and the first language sentence of the first translation example is 4 and the first language sentence of the second translation example. It is recognized that the number of word matches is three. This recognition result is transmitted to the ranking means 6, and the ranking means 6 outputs to the search result display means 7 to display the first translation example as rank 1, and displays the next translation example as rank 2. Output to.

(4) As a result, in the search result display means 7, translation examples of translation target sentences and ranks 1 and 2 are shown in FIG.
Display as shown in. In this case, in each rank, the sentence to be translated is placed on top, and the first language sentence of the translation example and the second language sentence of translation example are placed below each other. Then, the translation target sentence, the first language sentence of the translation example, and the second language sentence of the translation example are arranged in this order. With such a triple display, the translation target sentence and its search result are displayed in close proximity, so that the user can easily understand the correspondence between the translation target sentence and the translation example.

Further, as shown in FIG. 4, all the sentences of the translation target sentence, the first language sentence of the translation example, and the second language sentence of the translation example have corresponding pale words / phrases (have-have-hold etc.). Words / phrases that are highlighted and correspond only to the translation target sentence and the first language sentence of the translation example (aa, etc.), and words / phrases that correspond only to the first language sentence and the second language sentence of the translation example ( (love-love) is also identifiably highlighted in a different color, for example.

When the user selects one of the words / phrases with, for example, the mouse, the other corresponding words / phrases are displayed brighter and highlighted. For example, when the user selects “pen” of the first language sentence 42 of the translation example shown in FIG. 4, “pen” of the translation target sentence 41 and “pen” of the second language sentence 43 of the translation example are brighter and higher at the same time. Light is displayed.

When the user selects another similarity calculation method from the calculation method selection area 35, the similarity is calculated according to this and the ranking display is performed.

(5) When the user selects the second language sentence of the translation example with the mouse, for example, by looking at the search results displayed in the ranks as described above, this is output by the translation example selecting means 8 and selectively displayed on the display screen. To be done. For example, if "I have a pen that I love" is selected, this is selected and displayed by the output means 9.

(6) The user corrects this output sentence to "I have a pen."
a pen. You can make a translation of ".

In the above description, the present invention has been described for the case of translating English into Japanese, but the present invention is not limited to this, of course, and can be applied to other languages.

[0059]

According to the present invention, the following effects can be achieved.

(1) According to the present invention, the result of searching the similar translation example database using the input sentence to be translated as a search key is used to change the ranking by selecting an appropriate similarity calculation method or to translate the sentence to be translated. , The first language sentence and the second language sentence in the translation example are highlighted by displaying the word / phrase correspondence of the triplet, or ranked by using the similarity using the triplet information. The translation example search result can be effectively used, and it becomes possible to support the translation work with higher accuracy and speed.

(2) By weighting the number of additions according to conditions such as the part of speech and the number of consecutively matching words, it is possible to finely adjust the degree of similarity, to finely determine the degree of similarity, and to improve accuracy. High translation support can be provided.

(3) By highlighting corresponding words / phrases, it is possible to effectively use translation examples that are hardly similar to conventional ones and which are not similar to each other, as the equivalent of the translation / phrase dictionary search considering the context. it can. As a result, more translation examples than before can be used, and efficient support for translation work can be provided.

Further, in such a use method, a set of three words, that is, a translation target sentence, a first language sentence of a translation example, and a second language sentence
A similarity calculation method that adds weights to the number of phrase correspondences is effective. This is because such a similarity calculation method can obtain a lot of translated word / phrase information in consideration of the context.

As described above, the optimum similarity calculation method for a translator is different when a very similar translation example can be retrieved and when only a very similar translation example can be retrieved.

In order to quickly find the translation example that the translator really wants, it is necessary for the translator to be able to freely select the ranking criterion, that is, the similarity calculation method, when displaying the ranking. According to the present invention, the translator can select the similarity calculation method, and efficient translation support can be performed.

[Brief description of drawings]

FIG. 1 is an embodiment of the present invention.

FIG. 2 is an explanatory diagram of a translation example database.

FIG. 3 is an example of ranking display of similar translation search results.

FIG. 4 is an explanatory diagram of a search result display state.

FIG. 5 is a schematic comparison diagram of a conventional example and the present invention.

[Explanation of symbols]

1 Translation example database 2 input means 3 Similar translation example search means 4 Similarity calculation method selection means 5 Similarity calculation means 6 ranking means 7 Search result display means 8 Translation example selection means 9 Output means

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Hide Fuji             4-1, Kamiodanaka, Nakahara-ku, Kawasaki-shi, Kanagawa             No. 1 within Fujitsu Limited (72) Inventor Kiyoshi Okura             4-1, Kamiodanaka, Nakahara-ku, Kawasaki-shi, Kanagawa             No. 1 within Fujitsu Limited F term (reference) 5B075 ND03 NK32 PP24 PQ02 PQ36                       PQ46 PR06 QM08 UU06                 5B091 AA06 CB24 CC15

Claims (2)

[Claims] 1. An input unit for inputting a translation target sentence, a similar translation example retrieval unit for retrieving a similar translation example from a translation example database using the translation target sentence as a search key, and an order of displaying a plurality of retrieval results is determined. A similarity calculation method selection means for selecting a plurality of prepared similarity calculation methods, a ranking means for displaying a plurality of search results in order according to the similarity calculated by the similarity calculation method, and a search result The translation target sentence of one language sentence and the first translation example
Between the search result display means for displaying the language sentence and the second language sentence of the translation example as a triplet, and the triplet of the translation target sentence, the first language sentence of the translation example, and the second language sentence of the translation example. Correspondence information search means for highlighting and displaying the associated words / phrases, and selecting one of the highlighted words / phrases to further highlight the other two. And a translation example selecting means for outputting a sentence / phrase / word in a second language of a translation example selected from a plurality of search results. . 2. The similarity calculation method is associated with three sets of a translation target sentence, a first language sentence of a translation example, and a second language sentence of a translation example in a similar translation example search result. 2. The translation support device according to claim 1, wherein the number of words / phrases that are added are weighted and added together.