TW200530856A - System, method and program for identifying corresponding translation - Google Patents

Abstract

Suitable translations of phrases to be translated in original sentences having a high probability of obtaining natural translated sentences as sentences in an object language from the original sentences in an original language are obtained. A plurality of natural sentences, in the original language composed of a plurality of words, correlated with translated sentences in the object language are registered in corresponding translation database. When a phrase to be translated in natural sentences in the original language is specified, the corresponding translation database is searched for natural sentences containing the phrase to be translated (102). When no natural sentence completely coinciding with the original sentence is found (negative in 104), a degree of coincidence between each of the extracted natural sentences and the original sentence is evaluated, and a translation having the greatest degree of coincidence is displayed (108-112). When a plurality of natural sentences having the largest degree of coincidence is found (positive in 110), words having a high probability (correlation) of existing in the same sentence as the phrase to be translated are identified and translated sentences of natural sentences containing words having high correlation are displayed (114-118). When no natural sentences containing words having a high correlation are found (negative in 116), the original sentences are searched for words replaceable with targeted words (120-128), and the translated sentences of the natural sentences of which targeted words are replaced with the alternative words in the original sentence are displayed.

Description

200530856 V. Description of the invention α) [Technical field to which the invention belongs] The present invention relates to a system, a device, and a program for identifying a corresponding translation, and particularly to at least one translation in an original sentence written in an original language. A translation recognition system for the translation of vocabulary, a device for identifying a translation applied to a translation recognition system, and a program for allowing a computer to have the functions of the above translation recognition system. [Prior art] It is expected to understand that a machine translation system translates a sentence (an original sentence) narrated in a natural language into a language (an object language: an object language) sentence (translated sentence) has been a long time, and many techniques related to machine translation improvement have been mentioned. For example, Japanese Patent Application Laid-Open (JP-A) Patent No. 1 1-3 2 8 1 7 8 discloses a technology in which a group of expressions expressed in one phrase (expressi 〇η forms) are stored in one of the English-Japanese phrase dictionaries installed on a hardware device, and retrieve the coordinating conjunction's vocabulary-containing expressions in the original English text of the grammatical analysis process, and when The search expression form is regarded as a morpheme when it is stored in the English-Japanese phrase dictionary, or when a prefix or suffix containing a vocabulary of the search expression is identified

Page 8 200530856 V. Description of the Invention (2)) to analyze the structure of a sentence without the need to separate the vocabulary. In addition, Japanese Patent Application Publication-Publication (JP-A) No. 0 6-3 1 4 2 9 4 discloses a technology in which a characteristic table (characteristictab) of each mu 1 tiv 〇ca 1 vocabulary is stored in advance. 1 e) Relevant meanings and features, and generate a feature record in the original sentence written in the first language, and based on the features processed by the original sentence compared with the generated feature table and the stored feature table Choose and display the meaning of a multi-meaning word. [Summary of the Invention] The work of translating an original sentence narrated in an original language into a translated sentence narrated in an object language (〇bject 1 anguage) is accomplished through a translation process, in which the narration is described in the original language The vocabulary of the original sentence in is converted into the vocabulary in the target language, and the converted vocabulary is rearranged based on the understanding that the sentence usually contains vocabulary and grammar and according to the grammar rules of the target language. The translation process uses machine translation using a computer to perform translation, and the corresponding translation system is registered in the dictionary according to the vocabulary unit (word unit), and the vocabulary is sequentially selected from the original sentence, and the selected vocabulary is repeatedly retrieved. Corresponding translations to replace the original sentence with the corresponding vocabulary unit, and identify each vocabulary of a part of speech to analyze the sentence structure, based on the analysis results of frequently used sentence structures and the grammar of the target language Rules, rearrange the corresponding translations according to the vocabulary units to get the commonly used 200530856 V. Invention Description (3) (g en e ra 1 1 yused) translation sentences (corresponding translations). In machine translation, even if one of the stereotyped phrases in the original sentence can be applied to the technology described in the Japanese Patent No. 1 1-3 2 8 1 7 8 and has the possibility of obtaining the above phrase Appropriate translation, even if a polysemy exists in the original sentence, the specific translation of the polysemy can be obtained using the technology described in Japanese Patent Application No. 0 6-3 1 4 2 9 4 (narrated in Japan) The technology of the patent application No. 0 6-3 1 4 2 9 4 requires complicated work, such as analyzing the characteristics of selecting polysemes and meanings and selecting polysemes to prepare characteristic tables in advance). However, in machine translation, even if the translation obtained is free of grammatical errors and translation according to vocabulary units, it is still unnatural (u η natura 1) to find the translated sentence as a sentence of the target language, even if narrative For example, the technologies of Japanese Patent Nos. 1 1-3 2 8 1 7 8 and 0 6-3 1 4 2 9 4 are actually machine translation technologies that still cannot obtain the appropriate translation accuracy for practical use. It is estimated that the cause of these facts is the translation process using machine translation. When people speak or write a sentence in their native language, it is obviously different from the process of creating a sentence. That is, when people speak or write a sentence in their native language, they use grammar rules to retrieve unused vocabulary, but in fact, based on the context of the context-dependent vocabulary and background knowledge about the phrase itself After that, from the number of huge chunks (meaning quality:

Page 10 200530856 V. Description of the invention (4) The vocabulary, phrases, and stereotyped expressions (stereotyped express i ο arrangement, sentence structure, and sentences) are connected to the appropriate vocabulary and phrases searched (and sometimes selected) As well as sentences to create sentences. Therefore, to understand the background knowledge of the phrase itself related to related words before context-dependent vocabulary and when people speak in their mother tongue or both use the same method as when people speak or write in their mother tongue The unit (also known as chunks) searches for the translation of the original sentence and obtains the natural translation as a sentence in the target language. The actual situation is based on the translation of the chunk to find the corresponding translation. The implementation in the translation system is because the above-mentioned translation creation is not easy to identify the original text or the generation of background knowledge chunks related to the phrase itself will also be ambiguous. The present invention was established by taking the above facts into consideration, and the purpose is to obtain a corresponding translation recognition system, a corresponding translation recognition method, and a program. In order to achieve the above purpose, the translation recognition system of the scope of patent application related to the present invention includes a storage device for storing a plurality of natural sentences in a plurality of vocabularies of the original language, and the above natural sentences are related to the translation in the language; the retrieval device is used for Retrieve natural sentences containing translated phrases from the original original sentence from the plurality of natural sentences stored in the original language in the aforementioned center; the corresponding translation device selects ns according to the above retrieval device and the above original sentence retrieval ), According to love. After rendezvous, at the time of writing, the search terms are missing / η, the translated sentence of the sentence, and the corresponding 1 item contains the nature of the target language storage language translation recognition 200530856 V. Description of the invention (5) Between the sentences To determine the degree of agreement, and to identify the above translation phrases included in at least the corresponding translations of natural sentences selected based on the degree of agreement determination, as at least the above translation phrases in the original sentence. According to the first patent application scope of the present invention, a natural sentence containing a plurality of words in the original language is stored in a storage device related to the translation of the natural sentence and the target language. In the dictionary of the traditional machine translation system, the natural sentences related to the present invention are at least unedited or processed, such as the selection of sentences, clauses, phrases, stereotyped words, and idiomatic phrases that are divided by the vocabulary unit and the multi-word vocabulary. (C ο 1 1 〇cati ο η), and in the example registered in the dictionary that stores the vocabulary units of natural sentences and sentences related to translation in the target language, the exclusion of polysemes provides a lot of correspondence The need for translation (multi-word vocabulary is related to many translations, because suitable translations must be selected, so there may be incorrect choices), and translations corresponding to a natural sentence can be converted into natural sentences in the target language. In addition, translations of vocabulary corresponding to other natural sentences themselves can be stored in the storage device of the present invention. According to item 1 of the scope of patent application of the present invention, the retrieval device retrieves from the plurality of natural sentences stored in the original language of the storage device the narration contained in the original language (sentences, clauses, phrases, stereotyped words, and idiomatic expressions). Any one of the languages), the natural sentence of the translated phrase of the original sentence. Translation phrase means a phrase translated especially in the original sentence and can be a word

Page 12 200530856 V. Description of the invention (6) Or it may contain plural words. For example, when a user (translator) uses a translation recognition system related to the present invention as an electronic dictionary for translation work, the translator can specify the translation phrase. In addition, for example, when using a machine translation system or an automatic interpretation system as a part of the translation recognition system related to the present invention, a machine translation system that performs machine translation using a corresponding translation identified by the translation recognition system related to the present invention Or the automatic interpretation system can clearly explain the translation phrase. Alternatively, the user (translator, etc.) can explicitly specify the original sentence containing the translation phrase, or can automatically specify it (for example, automatically recognize a sentence or include a clause of the translation phrase as the original sentence). Since the retrieval device retrieves a natural sentence containing a translation phrase, the retrieval result is a selection of a natural sentence translated by the translation phrase included in the corresponding translation. According to the first translation recognition device of the first patent application scope, the degree of agreement between the natural sentences selected by the search of the retrieval device is determined, and the corresponding translation of the natural sentence selected based on at least the degree of agreement is identified The translation of the translated phrase in the translation of at least the original sentence in the original sentence. Therefore, between the high-probability natural sentences picked by the retrieval device (for example, using a natural sentence with the same meaning as the polysemous word in the original sentence), the natural sentence is similar to the original sentence as Natural sentences with a high degree of consistency, and at least the translation phrase in the translation is recognized as the translation phrase in the original sentence (as a matter of course, the translation of the phrase is in addition to identifying the translation phrase in the relative translation, and according to

Page 13 200530856 V. Description of the invention (7) According to the degree of consistency with the original sentence, it is used as the translation of the phrase in the original sentence. Therefore, according to item 1 of the scope of patent application of the present invention, consider the punctuation marks in the original sentence. The biggest technical difficulty of (punctuati ο η) is to store a plurality of natural sentences in the original language corresponding to the translation in the original language, and to select from the stored natural sentences with a high degree of consistency with the original sentences in the original language Natural sentences and identifying translations in the natural translations corresponding to the selected natural sentences as translation phrases in at least the original language. Therefore, the present invention enables a natural translation to be equivalent to a translation corresponding to an original sentence often performed in accordance with a chunk, as a result of not needing to recognize punctuation marks in the original sentence, and enables the original sentence to be obtained. Appropriate translation of the translation phrase (highly probable translation obtained from the original sentence as a sentence in the target language). According to item 1 of the scope of patent application of the present invention, it is possible to understand the sentences corresponding to the natural sentences stored in the storage device, without the need to perform complicated editing and procedures, such as dividing natural sentences by vocabulary units, and selecting multiple ideas. The vocabulary and when storing translations corresponding to the natural sentence itself, at least when storing the device, list as many translations of the multi-sense vocabulary as possible for each related vocabulary. In addition, according to the first patent application scope of the present invention, the degree of agreement between the natural sentence selected by the retrieval device and the original sentence is determined. Based on the degree of agreement, a natural sentence can be selected. 200530856 Description of the Invention (8) A proper translation is obtained, and it is not necessary to execute a complicated program, such as analyzing a part of speech and sentence structure analysis in a traditional machine translation system, and the program can be simplified. According to the invention, the storage device of item 1 of the patent application stores the translation of the natural sentence corresponding to itself, and the natural sentence may be completely consistent with the original sentence stored in the storage device. Taking this into consideration, as described above, for example, in the scope of patent application No. 2, the preferred translation recognition device also retrieves natural sentences that are completely consistent with the original sentence from the plurality of natural sentences stored in the original language of the storage device, and should pick up When selecting a natural sentence that is identical to the original sentence retrieved by the retrieval device, the first translation recognition device recognizes the translation of the natural sentence that is identical to the original sentence. Therefore, when a natural sentence that is completely consistent with the original sentence is stored in the storage device, a translation of the original sentence can be obtained. According to item 1 of the scope of patent application of the present invention, the degree of agreement between the retrieved natural sentence and the original sentence is retrieved, such as the following method. That is, according to item 3 of the scope of patent application of the present invention, there is a characteristic. According to item 1 of the scope of patent application of the present invention, between the natural sentence and the original sentence selected by the retrieval device, the first recognition device Calculate the number of words that are consistent with the original sentence and evaluate the degree of agreement, so that as the number of words that coincide with each other increases, the degree of agreement with the original sentence is also raised. Between natural sentences and original sentences, the number of words that express one degree of consistency is an important indicator. By assessing a consistent process

Page 15 200530856 V. Description of the invention (9) degree, when the number of consistent vocabulary increases, the degree of consistency with the original sentence also increases, and based on the high accuracy of the degree of consistency, the selection is similar to the original sentence One of the natural sentences. According to item 3 of the scope of patent application of the present invention, although it is possible to directly use the calculated number of consistent vocabularies as the degree of consistency, as described above, for example, for item 4 of the scope of patent application, it is better to use the calculated number of consistent vocabularies divided The number of vocabularies containing translated phrases is taken as the degree of consistency. Therefore, from this procedure, the number of vocabularies containing translation phrases can be normalized-the value of the number of vocabulary values is used to obtain a degree of agreement with the present invention to select the degree of consistency with use The highly accurate original sentence is similar to the natural sentence and is not related to the number of words containing the translated phrase. In order to calculate a consistent vocabulary, the number of vocabularies containing translation phrases (the value of the number of consistent vocabularies evaluated) and other vocabulary can be distinguished, except that different weights (weights set) are used to make the weights include slices. The word vocabulary of the language is higher (heav i er)), and the number of vocabulary including the translation phrase is multiplied by the number of the vocabulary of the consistent vocabulary, and the value obtained by adding the total value is used as the word number. According to item 3 of the scope of patent application of the present invention, the first translation recognition system is allowed to calculate the number of mismatched words between the natural sentence and the original sentence selected by the retrieval device and to evaluate the original sentence. The degree of consistency, so that when the number of words calculated is reduced, and

Page 16 200530856 V. Description of the invention (10) The degree of consistency of the original sentence has also increased, as described in item 5 of the scope of patent application. Therefore, it is possible to improve the accuracy of the degree of consistency between the evaluation of the natural sentence and the original sentence, so that when the number of non-consistent words decreases, the degree of consistency also increases. The number of consistent vocabulary, and the number of consistent vocabulary as described in item 3 of the scope of patent application. Chu Zhongming's original text was in the original language. In the current item 3, the Circumvention of the Scriptures and Fan Li only specializes, please subordinate φ-sentence as it should be. You can save this deposit due to the debit and exchange of e) In the middle of h, 子 子 丨 the beginning of the opposite sentence is a for V. The word cn for the word J) Now • 1 out of Γ Changzhi > d is for ο sentence example when the first one is not high, the sentence has a beginning and the original sink and the word, the actual reality is often chosen to think that it can be done in the cause, so that the can, but the sentence can have its own First, at that time, I would like to ask you to consider the application of the testimony from the example of the description, such as the question, the question of the question in the question. Of course, the definition of the number of degrees in the white S and the degree of the process of the capsule is to make the device identification available in advance so that it can be, therefore, due to the accent. The one in the shadow of the word suffix above the hui hui is now called: river outbound fKr-# f {子 经 的} present, de output term η 6 i often ef d d e e except Fan pr row profit {to The clause between the finer ones begins with the original.子 X) 6 sentences d η ran. 1 White (indicated as the meaning of the degree of the course of the degree of the convergent response should be based on the correct application of the invention according to the scope of the patent application item 3 or 5 , Preferably as described above, for example, the 7th in the scope of patent application, when calculating consistent or non-consistent vocabulary

Page 17 200530856 V. Description of the invention (11) When considering the singular / plural (singu 1 ar / p 1 ura 1) or tense as a consistent vocabulary, it has a zigzag form (inf 1 ecti ο η) Vocabulary, the first translation recognition device calculates a vocabulary as a consistent vocabulary. Register the relevant vocabulary in advance in the form, and study whether to display only the inconsistent suffixes registered in the form. You can also find that the vocabulary has a zigzag form because of the singular / plural or tense. Therefore, fundamentally (fundamentally) consideration is to exclude the influence of vocabulary in zigzag form due to singular / plural or tense. It is also possible to use the number of consistent or non-consistent vocabulary on the number of consistent or inconsistent vocabulary Used as an indicator of the degree of more consistent response between natural sentences and original sentences. In addition, it is also preferable to use the vocabulary in the English sentence of the letter above or below as a consistent vocabulary. According to item 3 of the scope of patent application of the present invention, preferably as described above, for example, item 8 of the scope of patent application, when calculating a consistent vocabulary, the first translation recognition device does not repeatedly calculate the vocabulary used two or more times. Therefore, the effect of using two or more consistent vocabularies can be ruled out on the consistent vocabulary, and the number of consistent vocabularies can be used as an indicator of the degree of more consistent response between the natural sentence and the original sentence. Instead of avoiding double counting as described above, the maximum number of times that the number of consistent vocabularies using two or more predefined numbers η (η-2) can be calculated can be limited. According to item 3 or item 5 of the scope of patent application of the present invention, preferably as described above, for example, item 9 of the scope of patent application, the first translation recognition device searches for

Page 18 200530856 V. Description of the invention (12) Search the natural sentence selected and the original sentence to evaluate the similarity of the order of the vocabulary, and evaluate the degree of consistency with the original sentence, so that the order of the vocabulary As the degree of similarity increases, the degree of agreement with the original sentence also increases. Even if the sentences contain the same vocabulary, a natural sentence sometimes has different meanings (translations) depending on the arranging order of the vocabulary. Based on the evaluation of the degree of consistency from the arranging order of the vocabulary as described above, you can choose the original Sentences are similar to a natural sentence and the degree of consistency is evaluated, so that when the similarity of the arrangement order of the vocabulary increases, the degree of consistency also increases. According to item 3 or item 5 of the scope of patent application of the present invention, preferably, as described above, for example, item 10 of the scope of patent application, the first translation recognition device is consistent with the natural sentence selected by the retrieval device. In the vocabulary, the number of inconsistent vocabularies is calculated, and the degree of agreement with the original sentence is evaluated, so that when the number of inconsistent vocabularies in the consistent vocabulary is reduced, the degree of agreement with the original sentence increases. Therefore, the accuracy of the degree of consistency is improved, and a natural sentence close to the original sentence with high accuracy can be selected. Therefore, according to item 1 of the scope of patent application of the present invention, sometimes based on the degree of consistency, it is difficult to identify (select) natural sentences that are similar to the original sentence, such as selecting a plurality of natural sentences with the same degree of consistency); Although, the above identification depends on the kind of natural sentence (translation) stored in the storage device. Considering the problems described above, according to the application of the present invention

Page 19, 200530856 V. Description of the invention (13) Please refer to item 1 of the patent range, preferably as described above, for example, for item 11 of the patent application range. The above search device is used to translate phrases and words that often appear in the same article in the original language provided separately. With reference to specific frequently-occurring phrases found in the search device and existing sentences, the first translation recognition device searches Regarding the frequently occurring translations of natural phrases containing each translated phrase and specific frequently occurring phrases, as in the case of including each translated phrase between natural sentences selected by the retrieval device search Translation of natural sentences, and the first translation recognition device recognizes frequently occurring translations as translations of translation phrases in the original sentence. When a translated phrase is the same as the original sentence in the original sentence, which is often found in the original sentence (referred to as idiomatic phrase), the translation of the highly possible translated phrase includes each translated phrase and the specific regular phrase. The translations of the corresponding translations in the natural sentences of the emerging phrases are consistent. However, although the translated phrase is likely to be a proper translation of the translated phrase in the corresponding translation of the natural sentence, the above-mentioned natural sentence contains each specific frequently occurring phrase, and has an inappropriate translation of the translated phrase. Natural sentences may also be mixed in the above-mentioned natural sentences stored in the storage device. Based on the above, according to item 11 of the scope of patent application of the present invention, when searching for highly relevant and frequently occurring phrases, between understanding frequently occurring phrases and the translation of natural sentences containing each translated phrase, refer to Specific frequently-occurring phrases in the original sentence, and frequently-occurring ones found in natural sentences used to translate phrases and include each specific frequently-occurring phrase

On page 20, if you have n’t done the original, please ask the translation of the idiom, consider turning t) Shen Li ’s review of the special and cause of the X, suitable for the present language, Guo Hefa issued the final text of the film As usual. Capsules According to the reference translation, the meaning of the reference will be explained in the root and review. The chapters of the post-Chinese-Israeli translation should be listed in the same sentence, which is difficult to translate in the plain text. The relative selection is based on the rough selection of Ying Ziyuan's choice. Sentences of the sentence can be taken into consideration, and then the time is taken into account. The self-sentence and the sentence are included in the same predecessor of the C language. The original item is from the original text of the film, and the original word is translated. According to the enclosing term, it is necessary to be the same as 4 as the post-normative 1J? ≫ Yin Yudi has a text translation 200530856 V. Description of the invention (14) Translation. Therefore, appropriate translations of the translated phrases in the original sentence can be obtained based on the frequently occurring phrases (the corresponding translation of the natural translation from the original sentence is highly likely to be used as the sentence in the target language), even if only It is also difficult to identify (select) a suitable natural sentence based on the degree of consistency. Register the phrases that often appear in the same original text as the original language and refer to the table to search for the frequently occurring phrases mentioned above. According to item 11 of the scope of patent application of the present invention, there are two advantages of frequently occurring phrases found based on the search of natural sentences selected by the retrieval device. Among them, time and manpower for preparing the above-mentioned forms can be saved, and Requires storage capacity for the form. According to item 1 of the scope of patent application of the present invention, it is preferred to identify a selection

Page 21 200530856 V. Description of the invention (15) Alternative phrase recognition device can replace the target phrase in the original sentence, but it does not include the search phrase The natural sentence in the selected original sentence, and the first translation recognition means recognizes the translation of the selective phrase recognized by the recognition means, and each translation phrase including at least the natural sentence selected by the retrieval means by the retrieval means The translation phrase in the corresponding translation sentence of one natural sentence is used as the translation of at least the translation sentence in the original sentence, as described in item 12 of the scope of patent application. According to item 1 of the scope of patent application of the present invention, a natural sentence is sometimes selected by a retrieval device, and in a natural sentence, a specific phrase of the original sentence replaces another phrase. Here, if the meaning of the original sentence is similar to that of the natural sentence, consider that the specific phrase has the substitution of other phrases. Even if, for example, a plurality of natural sentences having the same degree of consistency as the natural sentences picked up by the retrieval means of the retrieval device, when a natural sentence used for a specific phrase in one of the original sentences replaces a specific phrase in the natural sentence, it can be substituted. In other phrases (optional phrases), because natural sentences are likely to have the same meaning as the original sentences, it is better to choose natural sentences. On the other hand, according to item 12 of the scope of patent application of the present invention, the optional phrase is recognized by the recognition device, and the optional phrase replaces the selected target phrase by the search of the retrieval device, and does not include the existence of the original Among the natural sentences in the sentence, the first translation recognition device recognizes the translation of the optional phrase recognized by the recognition device, and at least the retrieval selected by the retrieval device 200530856 V. Description of the natural sentence of (16) One of each of the translated phrases is translated from the iron translation sentence 1 ^ ^…, the corresponding 4 phrases of Xizi, as a translation of at least the private 4 2 mountain. Therefore, even if σ Α + f starts with the phrase "a p make, the base & cause is difficult to recognize the existence of j and get one of the original sentences in the original sentence ^ < _ 1 彳Mask potential 丨 Ancient ^ Γ Translation of mouth fit (from a sentence in a sentence). His Majesty is a natural sentence suitable for the target language, which can be based on; = Selection to get one of the original sentences Translation: presence or absence of mouth ”T I 2 ° appropriate translation (from the scope of the patent application related to the invention f identification of optional phrases can have each other in the form 夂:;, line-

: ^ ΐ 夕; For example, from a plurality of magic sentences stored in the storage dress, you can retrieve a phrase containing the subject phrase from the bran sentence, "quote since the lack of a spoon early item f, search and retrieve far away Since ..., the sentence has a natural sentence with the same sentence structure, and retrieves the identified phrase in the selected natural sentence as an alternative language, as described in the item i 3 of the patent application. In addition, the time and manpower required for the above table can be saved, and the storage capacity of the required storage table can also be saved. A large amount of background knowledge about the phrase itself (immeasurable amount) has many characteristics based on the existence of the phrase, and it is difficult to select all and classify them according to the type (ca t e g ο r i e s). According to the application of the present invention, the items 12 and 13 of the patent scope identify a replaceable phrase in the same sentence structure as a phrase of the same kind, based on the relevant phrase itself.

Page 23 200530856 V. Description of the invention (17) The background knowledge can be roughly translated without the need to obtain the background knowledge of all the phrases classified according to categories. The translation recognition system for item 14 of the scope of patent application for the present invention includes a storage device for storing a plurality of natural sentences including a plurality of words in the original language, and the natural sentence system is associated with the translation in the original language; A retrieval device for retrieving translation phrases in the original sentence including the original language from the plurality of natural sentences stored in the storage device; a retrieval device for retrieving the retrieval of the retrieval device for the translation phrases The selected natural sentences are frequently-occurring phrases that are the same sentences that often appear in the original language; and the second translation recognition device is used to search for the above-mentioned translated phrases and natural sentences that include each specific frequently-occurring phrase The frequently occurring translation of the translated phrase in the translated sentence, and used to identify the frequently found phrase found as the translation of at least the above translated phrase in the original language, and the second translation recognition device refers to the above search device Certain frequently-occurring phrases found and present in the original sentence, as well as references Each of the above-mentioned retrieval means retrieves the translation sentence corresponding to the natural sentence of the above-mentioned translation phrase between the selected natural sentences, and the frequently occurring translation of the translation phrase in the translation sentence of the natural sentence. According to item 14 of the scope of patented application of the present invention, a storage device and a retrieval device provided with one of the scope of patented scope of application similar to the present invention. The search device searches for natural sentences picked up by the search device used to translate the phrase, and frequently appears in the same sentence that often appears in the original language.

Page 24 200530856 V. The phrase of invention description (18). The second translation recognition device searches for frequently occurring translations of the translated phrase and the translated phrase in the natural sentence of each particular frequently occurring phrase, and recognizes the frequently occurring phrase found as At least the translation of the translation phrase in the original language, and the second translation recognition device refers to the specific frequently-occurring phrase found by the search device and existing in the original sentence, and refers to the search that includes each of the search devices. The translated sentences corresponding to the natural sentences of the above-mentioned translation phrases between natural sentences and the frequently occurring translations of the translated phrases in the translated sentences of natural sentences. Therefore, according to the same method as the item 11 of the scope of patent application, even if the natural sentences stored in the storage device are mixed with inappropriate translated natural sentences with translation phrases, and each translation phrase is included with a specific frequently-occurring phrase You do not need to be affected by this problem to get a proper translation of the translated phrase in the original sentence (one of the high-potential natural translations is obtained from the original sentence as a sentence in the original sentence). The scope of patent application for the invention is item 1 or item 14. Although the translation phrase can include multiple words, when the number of words included in the word phrase increases, even after the retrieval device performs the search, the natural sentence still contains and cannot The possibility to pick all the translated phrases. Considering this problem, as described above, for example, in the case of the 15th patent application, when the translation phrase contains a plurality of words, it is better to retrieve a plurality of natural sentences from the original language stored in the storage device. A natural sentence that contains at least one of the plural words of the translation phrase (that is, the search contains all translations

Page 25 200530856 V. Description of the invention (19) Natural sentences of translated phrases and natural sentences containing a part of translated phrases). Therefore, it is clearly stated that the phrase contains a lot of vocabulary as a translation phrase, and even if a natural sentence including all the translated phrases is not stored in the storage device, it can be retrieved by at least the retrieval of the device from the corresponding translation of the translated phrase to obtain the natural Sentences (corresponding translated sentences). The translation recognition method according to item 16 of the scope of patent application of the present invention includes a first step, from a natural sentence containing a plurality of vocabularies stored in a storage device, to retrieve a translation phrase contained in the original sentence of the original language A natural sentence, which is related to the translation of the target language; and a second step, in the first step described above, to determine the degree of agreement between the retrieved natural sentence and the original sentence, and To identify the translation of the above translation phrase in the corresponding translation of at least the natural sentence selected based on the degree of consistency of the above decisions, as the translation of the above translation phrase in the original sentence. Therefore, according to the same method of the first patent application scope of the present invention, a proper translation of the translated phrase in the original sentence can be obtained (a natural translation with a high probability of translation is obtained from the original sentence as the target language. A sentence). Regarding the translation recognition method according to item 17 of the scope of patent application of the present invention, the method includes a first step, from a natural sentence containing an original language of a plurality of vocabularies stored in a storage device, and the original sentence used to retrieve the original language includes Translate the natural sentences of the phrases. The natural sentences are related to the translation of the target language. The second step is to search for the phrases used in the translation as well as the original.

Page 26 200530856 V. Description of the invention (20) The first step in the first language that often appears in the same sentence often appears in the first step of the retrieval device to retrieve the selected natural sentence; and the third step is to search for the above translation Phrases and frequently used translations of the translated sentences used in the corresponding translations of the natural sentences contained in each of the above-mentioned specific frequently-occurring phrases, and identify the frequently-occurring translations found as the original sentences The translation of the above-mentioned translation phrase, and the third step refers to the above-mentioned original sentence between the specific frequently-occurring phrase found in the above-mentioned second step and the natural sentence existing in the retrieval and selection in the above-mentioned first step, And refer to the corresponding translation of each natural sentence containing the above translation phrase. Therefore, according to the same method as in item 14 of the scope of the patent application, a suitable translation of the translated phrase in the original sentence can be obtained (a natural translation with a high probability of translation is obtained from the original sentence as a sentence in the target language ). Regarding the program according to item 18 of the scope of patent application according to the present invention, a computer is allowed to be connected to a storage device for storing natural sentences in an original language containing a plurality of vocabularies, and from the plurality of original languages stored in the above storage device A natural sentence is used to store a plurality of natural sentences in an original language including a plurality of vocabularies, and the above vocabulary has a corresponding translation in the target language, such as a natural sentence used to retrieve the translated phrase contained in the original sentence of the original language One of the functions of a retrieval device, and the first translation recognition device is used to determine the degree of agreement between the natural sentence selected by the retrieval of the retrieval device and the original sentence, and to identify at least the degree of agreement based on the decision Translation of the above translations in the translation of natural sentences of choice

Page 27 200530856 V. Description of the invention (21) Translation as a translation of at least the above-mentioned translation phrase in the original sentence. Regarding the program according to item 18 of the scope of patent application according to the present invention, a computer is connected to a program for storing a natural sentence containing a plurality of vocabularies in the original language. Translation, such as the functions of the retrieval device and the first translation recognition device. Therefore, when the above-mentioned computer executes the program according to item 18 of the scope of patent application for the present invention, the computer has the function of a translation recognition device according to the item 1 of scope of patent application. Therefore, according to the first patent application scope of the present invention, a proper translation of the translated phrase in the original sentence can be obtained (a natural translation with a high possibility of translation is obtained from the original sentence as one sentence in the target language). Regarding the program according to item 19 of the scope of patent application of the present invention, a computer is connected to a storage device for storing natural sentences in an original language including a plurality of words, and the plurality of original languages stored in the storage device are Natural sentences are used to store plural natural sentences in the original language containing plural words. The above words have corresponding translations in the target language, for example, natural sentences used to retrieve the translated phrases contained in the original sentence of the original language. A function of a retrieval device, and a search device for searching for a natural sentence selected by the retrieval of the retrieval device used for the above-mentioned translated phrase and a phrase frequently appearing in the same sentence that often appears in the original language, And a second translation recognition device is used to search for the translation of the above-mentioned translated phrases and corresponding to each natural sentence containing the above-mentioned frequently-occurring phrases.

Page 28 200530856 V. Description of the invention (22) Frequently occurring translations of translated phrases in translated sentences, and used to identify the frequently occurring translations found as translations of the aforementioned translated phrases in the original sentence, and The second translation recognition device refers to a specific frequently occurring phrase of the original sentence that is found between the search device and the natural sentence that exists in the search and retrieval of the search device, and refers to each natural sentence that includes the translated phrase. Corresponding translation. Regarding the program according to item 19 of the scope of patent application of the present invention, a program for connecting a computer to a storage device for storing a natural sentence containing a plurality of vocabulary in the original language, the above vocabulary has the corresponding Translation (creating a computer with a storage device, or connecting to another computer through an electrical cable), such as the functions of the above-mentioned retrieval device, search device, and second translation identification device. Therefore, when the above-mentioned computer executes the program concerning the item 19 of the scope of patent application according to the present invention, the computer has a function of a translation recognition device according to the item 14 of the scope of patent application of the present invention, for example. Therefore, according to the same method as in item 14 of the scope of patent application of the present invention, a proper translation of the translated phrase in the original sentence can be obtained (a highly probable translation of natural translation is obtained from the original sentence as the target language. A sentence). As described above, the present invention retrieves natural sentences containing translation phrases from the original sentences in the original language from the plurality of natural sentences stored in the storage device to determine the search between the selected natural sentence and the original sentence. The degree of agreement, and at least identify the choice based on the degree of agreement determined

Page 29 200530856 V. The translation of the translation phrase in the translation of the natural sentence of the description of the invention (23), as a translation of the translation phrase in at least the original language. Therefore, the present invention has a special utility (e X c e 1 1 e n t e f f e c t), in which it is possible to obtain a naturally translated sentence, and from the original sentence of the original language, an appropriate translation of the translated phrase in the original sentence can be obtained from the original sentence of the original language. In addition, the present invention searches for a translation phrase in each corresponding translation including a certain frequently occurring phrase and a frequently occurring translated natural sentence, and from the plurality of natural sentences stored in the original language stored in the storage device, In the original sentence, the original sentence of the original language is searched for natural sentences containing the translation phrase to identify the frequently occurring translations found as the translation of the above translation phrase in the original sentence, and the search is used to translate the phrase and often appear in the original Retrieve selected natural sentences that are frequently found in the same sentence of the language, and refer to the corresponding translation of each natural sentence that contains the specific frequently-occurring phrase and the translated phrase in the original sentence. Therefore, the present invention has an excellent effect, in which a natural translation sentence is likely to be obtained, and a proper translation of the translation phrase in the original sentence can be obtained from the original sentence in the original language. [Embodiment] A preferred embodiment of the present invention will be described in detail with reference to the drawings. First, a device is described that uses the present invention in an electronic dictionary system to provide a user with a translation that translates a designated phrase when the user translates the designated phrase.

Page 30 200530856 V. Description of the invention (24) Figure 1 shows that a personal computer (PC) 10 can perform the functions of the electronic dictionary described above. The personal electrical month direction 10 includes a central processing unit (CPU) 1〇 Α, read-only memory (Reacj-〇nly memory: ROM) 10B, I1 remote machine access memory (random-access memory: RAM) ioc and input / output port (port) i〇D, and through the convergence The bus 10E is connected to each other to assemble a personal computer, and the bus includes a data bus, an address bus, and a control bus. The display includes a cathode ray tube (CRT), a liquid crystal display (LCD), and the like. 'Input data to the keyboard through the user's keyboard 14' 'mouse' 16, hard disk Device (hard disk drive: HDD) 18. CD read-only memory (CD-ROM) 2 4 CD-only read-only device (CD-R 0 M drive) 2 0 and read data The scanner 22 is connected to the input / output unit 100D, for example, many input / output devices. A translation recognition program (corresponding to the program described in the patent application range 丨 8 and 丨 9) enables the personal computer 10 to have the function of an electronic dictionary system equipped with a hardware device of the personal computer 10. A corresponding translation database is used to store the corresponding translation identification program data, which is used to identify the appropriate translation stored in the hardware device 18

Page 31, 200530856 V. Description of the invention (25) Several devices that install (transfer) translation recognition programs to personal computer 10, such as recording translation recognition programs and setting programs in CD read-only memory 24 and installing translation recognition in advance Program 0: Insert CD CD-ROM memory 2 4 into CD read-only memory device 2 0 and instruct CPU 1 () A to execute the above-mentioned setting procedure to obtain translation recognition from CD read-only memory 2 4 in sequence. Programs and sequentially write the read translation recognition programs in the hardware device 3. The corresponding translation database can be stored in the hardware device 18, such as the corresponding translation database recorded in the CD random access memory 24 in advance, and When the translation recognition program is installed, the setting program is changed so that the corresponding translation database is written in the hardware device 18 at the same time. Lack of sentences and a large amount of original materials narrated in the original language (sentences, phrase phrases, stereotyped words, and any of the idiomatic phrases that have not been edited or processed. Specific examples as described later such as word division, polysemous word picking (Selection, etc.) Translations of the corresponding source materials described in the target language (white sentences such as translations that have not been edited or processed, such as the selection of vocabulary, multi-word vocabulary, etc., as described in specific examples, etc.) Registered in the corresponding translation database. 0 In this preferred embodiment, the hard disk device 18 stores the corresponding translation database corresponding to the storage device of the present invention. 0 On a recording medium (recording med i.um). The corresponding translation database is used to use the corresponding translation database, such as CD 5 MB memory 2 4 and DVD Read memory | (DVD-ROM) and discipline record from the media and in direct

Page 32 200530856 V. Description of the invention (26) ^-Read the corresponding translation database. Therefore, in this preferred embodiment, the function of the corresponding translation database is stored in the media, for example, about this two _, +% storage of clothes. In addition, the words in the translation corresponding to the original language and the target language itself are registered in the corresponding translation database of this preferred embodiment. Increasing the translation of natural sentences and its own translation can create a corresponding translation database for the ^ most ^ embodiment, for example, it needs to exist in the dictionary = ^ 二, for example, the translation of the vocabulary in the original language corresponding to itself. The effect of the preferred embodiment is described here. In one embodiment, the translation of Narada in the original sentence of the original language and the translation of the narrative in the target language—the phrase exists (the translation phrase can be a vocabulary, or the original sentence contains multiple consecutively arranged vocabularies Phrase), the user uses the personal computer to perform specific operations to control the translation of the translated phrase in the original sentence. Can read the original text in personal computer 10, such as the original text data (such as the original text entered by the user through the keyboard 14, the original text created using word processing software, the original text browsed on the Internet through the Internet, scanner origin, records The original text obtained from the character database and the data processed by the Optical Character Recognition (OCR) system, etc.) as the original sentence. The operation of the preferred embodiment, for example, correctly selects a content menu displayed in a translation phrase, and after displaying the translation phrase (reverse 1 y), selects a translation phrase in this case and selects a corresponding item using "Corresponding translation output"

Page 33 200530856 V. Operation of invention description (27), for example, the original sentence is displayed on the display 1 2. It is also possible to perform an operation to specify (spe c c fy) the output of the translation of the translation phrase, such as selecting to display the desired image (i con), for example, using the toolbar in the content menu as described above. When the above operation is performed to clearly explain the translation output of the translation phrase, the central processing unit 10 A of the personal computer 10 executes the translation recognition program to execute the translation recognition program, as shown in FIGS. 2A and 2B. The translation recognition program deals with the translation recognition methods described in items 16 and 17 of the scope of patent application. By executing this procedure, the personal computer 10 has the function of an electronic dictionary system (the translation identification system described in the patent application scopes 1 and 14). In the translation recognition program, the original data (processed original sentence) in a single original sentence containing one of the specific translation phrases is given priority in step 100, and it is needed to identify the translation piece in the original sentence required for processing Language information. The processed original sentence can be a sentence containing a translation phrase, or a clause, a phrase, a stereotype, or a phrase containing a translation phrase. When it is assumed that any one of a clause, a phrase, a stereotyped phrase, or a idiom containing a translated phrase is a processed original sentence, it is automatically determined by the user or through a translation recognition program, so that the clause, the phrase Words, stereotyped words, or idiomatic phrases as clear explanations of the original sentence processed. In the next step 102, from the natural sentences registered in the corresponding translation database, use the original text of the original sentence processed in step 100.

Page 34 200530856 V. Description of the invention (28) material, retrieve a natural sentence completely consistent with the processed original sentence as a key, and use it from the natural sentence registered in the corresponding translation database. The original data of the original sentence that needs to be processed in step 100 is used as a key, and a natural sentence containing the translated phrase (including a natural sentence partially consistent with the original sentence 'is included because it contains at least the translated phrase ). When the translation phrase contains plural words, the natural sentence contains at least one of the plural words containing the translated phrase to be retrieved at the same time. Step 102 corresponds to the search devices described in items 1, 14 and 18 of the scope of patent application, and item 19, and also corresponds to the items 16 and 17 of the scope of patent application. The first step of the narrative. Steps 104 and subsequent steps are equivalent to the first translation recognition device described in items 1 and 18 of the scope of patent application. In step 104, it is asked whether the natural sentence is exactly the same as the original sentence of the translation selected in the corresponding translation database retrieved in step 102. If the answer is positive, the program moves to step 106 to read the corresponding translation in the target language associated with the natural sentence consistent with the processed original language, and register it in the corresponding translation database, and The corresponding translation is displayed on the display 12 to end the procedure. When these corresponding translations are displayed, reading translation is searching for translations of translation phrases, and making translations of translations found to be important (h 1 gh 1 1 gh t e d). In this example, the user can understand the appropriate translation of a specific translation phrase (getting a natural translation from the original sentence with a high probability of translation as a sentence in the target language), and can understand the inclusion of special features

Page 35 200530856 V. Description of the invention (29) Appropriate translation of the processed original sentences (naturally translated sentences as sentences in the original language). The above steps 104 and 106 are equivalent to the first translation recognition device described in item 2 of the scope of patent application. (Example 1) The above program is described using real examples. Suppose an original language is English and an object language is Japanese. When included in a processed original sentence, "For your safety" clearly states "For your safety, don't rush into the train ( Foryour safety, don't rush into the train) "as the translation phrase, because the original language and the corresponding translation are registered by the vocabulary in the general translation program, and the comparison contains the above translation phrase "Safety" and choose to translate "anzen". However, in the translation recognition program of this embodiment, since the natural sentences in the original language and the corresponding translation systems in the target language are registered in the corresponding translation database, the natural sentence systems that are completely consistent with the original sentences have registrations. The corresponding translation database. When the natural sentence is exactly the same as the original sentence, the translation registered in the translation database corresponding to the natural sentence is displayed with the translation of the important translation phrase, as shown below: r Kiken desu kara? Kakekomi- josha wa oyame kudasa i ". It ’s difficult to choose "ki i k e n" as

P.36 200530856 V. Description of the invention (30) The corresponding translation of the word "safety", and you can only get a disapproval, but you can turn 5 translations. kakekomanaide kudasai ". On the other hand, the present invention provides a natural translation such as a native language (na t i v e 1 a n g u ag e). When the translation is displayed in step 106, in addition to the natural sentences that are completely consistent with the processed original sentence, it is also possible to display the natural sentences that are consistent with the processed original sentence portion selected in the search in step 102. . A plurality of natural sentences that are consistent with the processed original sentence part are generally selected by retrieval. When these are displayed, in order to show the translation of each natural sentence in a descending order of consistency on the display, it is preferred to use individual natural sentences that are consistent with the processed original sentence portion. The degree of agreement is calculated in the narrative. On the other hand, when a natural sentence completely identical to the processed original sentence is not found in the search of the corresponding translation database in step 102, the answer to the question in step 104 is negative. Then, the program moves to step 1 q 8 to calculate the number of sub-sinks (the number of consistent vocabularies) consistent with any of the vocabulary of the original sentence between the sub-sinks of the natural sentence for the retrieval in step 102 Corresponding natural sentences are selected in the corresponding translation, and the degree of agreement with the original sentence is calculated based on the result of calculating the number of different consistent words. In the present embodiment, when the translation recognition program is assembled, it often appears in the original

Page 37 200530856 V. Description of the invention (31) The vocabulary of the original sentence in the language (for example, "a", "the", "to") in the natural English sentence, and " (I η) ″) in the frequently occurring vocabulary table, and also stored in the hard disk device 18. Step 108 calculates the number of coincidences while referring to the frequently occurring vocabulary table to exclude the words appearing in the frequently occurring vocabulary table among the calculated number of consistent vocabularies. Therefore, the influence of the vocabulary that often appears in the number of consistent vocabulary can be avoided. The above-mentioned procedure in step 108 is equivalent to the first translation recognition device described in item 6 of the scope of patent application. In this preferred embodiment, when a translation recognition program is assembled, a word form in a zigzag form having a suffix having a different suffix due to differences in registered singular / plural or tense is also stored in the hard disk device 18. When the number of consistent suffixes is calculated in step 108 and only suffixes with inconsistent suffixes appear, regardless of whether the reason for the inconsistent suffixes in the singular / plural difference or the tense difference is assumed, the calculation is assumed Words with different suffixes are used as consistent words due to differences in singular / plural or tense. Instead of this difference, only the system of the irregular zigzag form is registered in the zigzag form form and the regular zigzag form, and the consistent vocabulary is automatically assumed to be acceptable. Therefore, the influence of suffixes with different suffixes on the number of consistent vocabularies due to singular / plural or temporal differences (the vocabulary should be fundamentally understood as a consistent vocabulary) can be avoided. The above procedure in step 108 is equivalent to the first translation recognition device described in item 7 of the scope of patent application.

Page 38 200530856 V. Description of the invention (32) In this preferred embodiment, even if the vocabulary reappears in the natural sentence, the period during which the number of consistent vocabulary is calculated in step 108 is once in order to match with some of the original sentences. The word vocabulary is calculated for the reason that the word vocabulary is consistent. To avoid repeated calculations, the word vocabulary is not calculated twice or more. Therefore, even if many parts of a natural sentence have the same consistent vocabulary, the number of consistent vocabularies can be prevented from being affected by the consistent vocabulary. The above procedure in step 108 is equivalent to applying for the first translation recognition device described in item 8 of the patent scope. A formula is defined to calculate the degree of agreement, so that as the word of agreement increases, the degree of agreement also increases. For example, based on the number of vocabularies containing translated phrases, a calculation formula that normalizes the number of consistent vocabularies (the degree of consistency is equal to the number of consistent vocabularies divided by the number of vocabularies containing translated phrases) can be used. The above-mentioned procedure in step 108 is equivalent to the first translation recognition device described in item 3 of the patent application scope (more specifically, as described in item 4 of the patent application scope in the next step 1 10), The degree of agreement in each natural formula is calculated to ask whether there are plural natural sentences with the greatest degree of agreement. If the answer is negative, the program moves to step 1 12 to read the phase registered in the corresponding translation database. Corresponding translation sentences, the above translation database has one of the largest degree of consistency, corresponding natural sentences, and translation sentences used to search for translations for translation of translated phrases, and is used for 200530856 V. Description of the invention (33) Display the translated sentence read on the display 12 together with the translation that found the important (hi gh 1 i gh ted) translation phrase, and end this process. In this example, the user can understand the specific translation piece A proper translation of the language (getting a natural translation of the highest possible translation from the original sentence as a sentence in the target language). Step 11 In step 2, The natural sentences with the greatest degree of agreement between the natural sentences selected in the search in step 102 (the number of natural sentences in a decreasing order of the degree of agreement, or the degree of agreement is high In addition to all the natural sentences of a specific value), it is also read from the corresponding translation database in a decreasing order of the degree of consistency, and displayed on the display 12 at the same time. If there is a The degree of plural natural sentences counts the number of words that are inconsistent with the original sentence in each natural sentence. Corresponding to the translated sentences of natural sentences with the same degree of consistency, it is based on the inconsistent words in the corresponding natural sentences. Decreasing (smaller) order is displayed to show translated sentences at the same time. (Example 2) The above program uses a real example to describe. Assume that the original language is English and the target language is Japanese. One word in the processed original sentence, "operat i on", specifically states "the system suddenly entered the operation (Th e system suddenly came into o p e r a t i〇n) ", and the natural sentences (1) to

Page 40 200530856 V. Description of the invention (34) (8) is related to the translation sentences (1) to (8), and is registered in the corresponding translation database as the translation operation "operation" Natural sentences, except for the natural sentence (3) shown in Figure 3, which has a degree of agreement with the number of consistent vocabulary of the original sentence with "1" 100% (also calculates the "operation" in the natural sentence As the consistent vocabulary as above), and the natural sentence (3) has the same number of vocabulary as the original sentence with "4 (the vocabulary" the "selected from the calculation of the consistent vocabulary as frequently appear Vocabulary) of 4%. Therefore, the translated sentence (3) is related to the natural sentence (3) and registered in the corresponding translation database. It is displayed together with important translation phrases, as described below: r Sisutemu ga kadosh i hajimeta ”. (Example 3) Assume that an original language is English and an object language is Japanese. When it contains a word "operation" in a processed original sentence, specify "a small business is exempt from the implementation of the new law (A sma 1 1 business is exempt from the operation of the new law) ", and the natural sentences (1) to (8) shown in Figure 3 are provided from the corresponding translated sentences (1) to (8), And registered in the corresponding translation database as a natural sentence containing the translation phrase "operation", except that the natural sentence (2) shown in Figure 3 has the same vocabulary as the original sentence with "1" The degree of agreement of the number 100% (also calculates "operation" in natural sentences as a consistent vocabulary as described above), and since

Page 41 200530856 V. Description of the invention (35) Natural sentence (3) has the same number of vocabulary as the original sentence with "4" (the vocabulary "the" selected from the calculated vocabulary is used as a regular The vocabulary that appears) is consistent with 400%. Therefore, the translated sentence (2) is associated with the natural sentence (2) and registered in the corresponding translation database, and displayed with important translation phrases, as described below: r Sh i mpo no tek i yo ”in The vocabulary "operation" in English sentences is used for a polysemous word that is difficult to identify a suitable translation. When it is easy to understand from the above examples 2 and 3, the corresponding translation database is registered in the translation sentence related to the natural sentence, and you can get a suitable natural sentence that uses one more word in the same meaning. (Translation sentence) as the original sentence, and a natural sentence is selected based on the degree of consistency (translated sentence (example 4) Chinese sentence example, because it is difficult to distinguish a large number of meanings without sufficient understanding of the Chinese language, it is often difficult to identify what should be obtained A unit in translation. In order to solve the problem in the translation recognition program of the present invention, a natural sentence containing a specific translation phrase is retrieved from the natural sentence registered in the corresponding translation database (when the translation phrase is When plural vocabularies are included, natural sentences containing at least one of the plural vocabularies including translation phrases are also retrieved, and translated sentences (including translated sentences including natural sentences) are displayed with a degree of consistency with all the original sentences , Except for natural sentences with a high degree of consistency). Therefore, users Can learn from the original

Page 42 200530856 V. Description of the invention (36) — In the sentence and the proper translation itself, the translated sentence with reference to the display should also obtain the unit in translation (u n i t). For example, when one of the original sentences of Tian Chinese is "developing country (developing country)" as an original predicate, it is translated into Japanese as a target language. If you specify "fa ()", "development (exhibition)" , "Developing (Exhibiting)", "Developing China (Exhibiting China)", or "Developing China (Exhibiting Countries)" as translation phrases, arrange the following based on the degree of consistency in the following order Translated sentences of natural sentences. (When the description of "developing countries (developing countries)" is used as a translation phrase, the degree of consistency is also shown in the following. However, it is a matter of course that the degree of agreement among the specific phrases is different. Translation phrases): 1. "Developing countries (developing countries)" = "hatten tojokoku, (developing country)" (the number of consistent vocabulary "5", and the degree of consistency 100% ). 2. "Chugokuwo hattensaseru" and "to develop China" "(the number of consistent vocabulary is" 4 ", and the degree of consistency is 80%). 3. "Under development (in development)" = "hatten-chu no, (developing)" (the number of consistent vocabulary is "3" and the degree of consistency is 60%). 4 · "Development (Exhibition)" 2 "Exhibition, Exhibition (hattensuru, hattensaseru), (Develop)"

Page 43 Licensing, factory reference. , The mid-level translation of the order of the Chinese translation of the order of the Nakamoto in the translation {body to the child to get the sentence in time to start with the same original} should be interpreted from J should be / the family sentence can be interpreted by the use of the { Line up the ambassadors, as stated in the State Council because of the exhibition 200530856_ V. Invention Description (37) (the number of consistent vocabulary is "2 丨, and the degree of consistency is 40%). 5." Fa () "=", out (Hassuru, dasu), (start, put out) "(the number of consistent vocabulary is" 1 ", and the degree of consistency is 20%). When an original sentence and a translated phrase are both "fa ()", each natural spoken sentence has the same number of vocabulary "1", and the degree of consistency is 100%. However, when the number of inconsistent vocabularies is not the same, the original sentence system is arranged in the following order, which is different from the example "Developing Country (Exhibiting Country)" in the original sentence.

1. Factory issued () j (the number of inconsistent vocabulary "0") 2. Factory development (exhibition) "(the number of inconsistent vocabulary" 1 ") 3. Factory development (in exhibition) j (of the inconsistent vocabulary The number "2") 4. Factory development China (exhibiting China) "(the number of inconsistent vocabulary S factory 3") 5. Factory developing countries (developing country) "(the number of inconsistent vocabulary factory 4") in similar In the method 3, both the original field of Yitian and a translated phrase are "developed (exhibited) j. Based on the degree of consistency and inconsistent vocabulary, the original sentences are also arranged in the following inhabitants. 1 in the original sentence. Examples in the order of "developing countries (developing countries)" or "developing ()" page 44 200530856_ V. Description of invention (38) 1 · "Development (exhibition) j (number of consistent vocabulary" 2 ", The degree of consistency is 100%, and the number of inconsistent words is" 0 "). 2. "In development (in development)" (the number of consistent vocabularies is "2", the degree of consistency is 100%, and the number of inconsistent vocabularies is "1"). 3. "Developing China (Exhibiting China)" (the number of consistent vocabularies is "2", the degree of consistency is 100%, and the number of inconsistent vocabularies is "2"). 4. "Developing countries (developing countries)" (the number of consistent vocabularies is "2", the degree of consistency is 100%, and the number of inconsistent vocabularies is "3") ° 5 · "Fa ()" (consistent The number of vocabularies is "1", the degree of consistency is 50%, and the number of inconsistent vocabularies is "0"). Even when the translation phrase system is the same, the translation sentence system is arranged in a proper order according to the natural sentence which is displayed as described above if it is different from the original sentence. Therefore, the user can refer to the arrangement result of the translated sentences to understand a suitable translation relative to one of the original sentences. When a plurality of natural sentences with the greatest degree of consistency exist, the answer in step 1 10 is yes. Then, the program moves to steps 1 to 4 to execute a sentence based on the translation phrase and the same sentence that often appears in the original language (with a vocabulary highly related to the translation phrase: corresponding to the first paragraph of the scope of patent application). Item 1, item 14, and item 17 that are frequently encountered), and select the analysis of the relevance of natural sentences in steps 1 1 4 and 1 16.

Page 45 200530856 V. Description of the invention (39) That is, in step 1 1 4 whether the translated phrase and the same sentence often appearing in the original language are included in the processed original sentence. Highly relevant vocabulary), no investigation was conducted to determine whether each vocabulary except the translated phrase in the original sentence contains the translation database retrieved in step 102 and checked each natural sentence. In this program, for example, the number of occurrences of each natural sentence between each vocabulary other than the translated phrase in the processed original sentence is recognized, and the vocabulary greater than a specific number of times (once, twice) , Or multiple times) as a vocabulary of high relevance to the translation phrase. The above steps 1 1 4 are equivalent to the search device described in item 11 of the patent application scope, item 14 and item 19, and also the second step described in the patent application scope item 17. In the next step 1 1 6, ask whether the vocabulary that has a high correlation with the translated phrase was found by searching in step 1 1 4. When the answer is affirmative, the program moves to step 1 18, between each natural sentence selected from the corresponding translation database retrieved in step 102, to refer to the translated sentence of the translation phrase, and Each natural sentence containing the vocabulary (words in the processed original language) that are highly relevant to the determined translation phrase in step 1 1 4 and is used to identify the often translated phrases in the natural sentences described above Appear translation. Often translations of translation phrases. Specifically, when a translation phrase and a single natural sentence are selected as a natural sentence containing each vocabulary that has a high relevance to the translated phrase, the translation in the natural sentence is identified. sheet

Page 46 200530856 V. Description of the invention (40) The translation of the language is used as a frequent translation. When plural natural sentences are selected as the above-mentioned natural sentences, the most frequently occurring translations of translations of translation phrases in natural sentences are identified as frequently occurring translations. As mentioned above, when a natural sentence containing a translated phrase and a vocabulary of a recognized original sentence is used as a vocabulary that is highly relevant to the existence of the translated phrase, the natural sentence may use the sentence of the translated phrase as having the same meaning Sentence, such as the original sentence. A natural sentence with different translations of the translated phrase may also be mixed in the natural sentence described above. On the other hand, in step 1 1 8 referring to the translated sentence containing each translated phrase and a natural sentence with a highly relevant vocabulary of the translated phrase, the frequently occurring translation of the translated phrase in the above-mentioned natural sentence is identified. Therefore, a proper translation of the translated phrase in the processed original sentence can be obtained. In step 1 19, from the translated sentences containing each translated phrase and a natural sentence with a high vocabulary related to the translated phrase, read the target language registered in the corresponding translation database related to the natural sentence. Translate sentences. In the natural sentence, the translation phrase series is related to the translation that is often found in step 1 1 8. Reading the translation sentence series is related to the translation of the translation phrase. In reading the translation sentence, it is emphasized that the translation phrase has The highly relevant vocabulary is displayed on the display 12 together to end the process. In this example, the user can also know the appropriate translation for a specific translation phrase (getting a high-probability translation from the original sentence as a sentence in the target language). The above steps 1 16 to 1 1 9 are repeated corresponding to

Page 47 200530856 V. Description of the invention (41) The first translation recognition device described in item 11 is the second translation recognition device described in items 14 and 19 of the scope of patent application, and the scope of patent application The third step described in item 17. Step 1 1 9 also includes each translated phrase and a natural sentence with a high vocabulary related to the translated phrase, and the above-mentioned translated phrases are often related to the natural sentence selected in step 102. The phrases are related, and the translations of plural natural sentences are read from the corresponding translation database in a decreasing order of consistency (the number of natural sentences in a decreasing order of consistency, or all have Sentences with specific values higher than the degree of consistency), and at the same time arranged on the display 12 according to the degree of consistency or the order of the inconsistent vocabulary. (Example 5) The above-mentioned program further describes the use of a real example. Suppose an original language is English and an object language is Japanese. When it contains a word "operation" in a processed original sentence, it specifies "I accept the hand to remove rectal cancer # ί (I had an operation to remove a rectal cancer) ", and the natural sentences (1) to (8) shown in Figure 3 are registered in the corresponding translation database with other natural sentence systems as a translation phrase containing" operative (〇perati ο η) ", because both natural sentences (6) to (8) have the same number of vocabularies" 2 ", it is difficult to choose the most suitable natural sentence (translated sentence) only from the degree of consistency. When searching from natural sentences (1) to (8)

Page 48 200530856 V. Description of the invention (42) In addition to processed original sentences (for example, "remove", "rectal", or "cancer"), the translation phrase "surgery" (Operat i on) "keyword, because natural sentences (6) to (8) contain" cancer ", so" cancer "is retrieved as a translation phrase in the processed original sentence The word "cancer" is highly relevant.

Natural sentences containing the translated phrase "operative (〇 p e r a t i ο η)" and the highly relevant vocabulary "cancer" may use the same meaning as the translated phrase "operative (〇 p e r a t i ο η)". However, the translation of the translated phrase is different from the natural sentences in the translated phrase. It also has the possibility of mixing the natural sentences registered in the corresponding translation database and conforming to the above situation. In this example, the natural sentence (8) corresponds to this example. Although the natural sentence (8) contains the translation phrase "ope rat i on" and the highly relevant vocabulary "cancer", but because the translation phrase "operat i on" is " sosa ", so the natural sentence (8) is different from the processed original sentence.

Due to the influence of these natural sentences, in order to prevent incorrect recognition of the translation of the translation phrase, reference is made to the translation phrase "surgery (〇 p er a t i ο η)" and the highly relevant vocabulary "cancer". The translation phrase "operat i on" and the highly relevant vocabulary "cancer" are included in nature in addition to the natural sentence (6)

Page 49 200530856 V. Description of the invention (43)

In the sentence, and the translation phrase "operat i on" of one of the above natural sentences is associated with the translation "shu j ut su" in the corresponding translation phrase registered in the natural sentence. Therefore, it is decided that the frequently-used translation of the translation phrase "operation (〇perati ο η)" in the natural sentence containing the translation phrase "operation" and the highly relevant vocabulary "cancer" is "shujutsu" , Including the natural phrase (6) including the translation phrase "operative (〇P erati ο η)" and the highly relevant vocabulary "cancer", choose the above-mentioned natural sentence t the translation phrase and the often appearing decision Corresponding translations, as well as natural sentences related to the translation sentence (6), are registered in the corresponding translation database to emphasize the vocabulary related to the translation of the translation phrase, as shown below. Factory Jon — san no shujutsu-g〇, kare no gan wa naotta. " Therefore, in this example, we also chose to use one of the appropriate natural sentences (translated sentences) as the original sentence.

Referring to FIG. 3, although the number of natural sentences including "operation (〇 p r a t i ο η)" is registered to simplify the narrative, more natural sentences are actually registered. When processing the original sentence, for example, it was picked and used the translated phrase "operat i on", such as "inj ury" or "hospital" as a sentence, according to The content of the processed original sentence as a highly relevant vocabulary. Everywhere

Page 50 200530856 V. Explanation of the invention (44) The original sentence system that has been processed, for example, pick and use the word "kad 〇", such as "injury" or "Dian Yue I (hospital)" One sentence in the translated phrase "operative (〇pera ΐ i ο η)" is based on the content of the processed original sentence as a highly relevant vocabulary. When processing the original sentence, for example, pick and use the vocabulary "ΐ ekiy 〇", such as a sentence in the meaning of "law (1 aw)" in the translated phrase "operation", according to the processed Content of the original sentence as a highly relevant vocabulary. On the other hand, when the vocabulary related to the translation phrase is not retrieved, the answer to the question in step 1 16 is negative. Therefore, the program moves to step 120, steps 120 to 134, based on the alternative vocabulary that replaces the target vocabulary other than the translated phrase in the processed original sentence (corresponding to the patent application (Optional phrase described in item 12 of the scope) and natural sentences selected to perform scheme analysis. That is, the vocabulary of the processed original sentence that exists in step 120 (the target vocabulary) is identified, and that it does not exist in each natural sentence selected by the corresponding translation database retrieved in step 120. Zigzag. In step 1 2 2, a natural sentence containing one of the target words is retrieved from the corresponding translation database registered in the corresponding translation database. When there are a plurality of target vocabularies, the search in step 1 2 2 is performed for each word bag. Steps 1 2 4 identify the natural sentences used to retrieve the mother (such as "natural sentences with the same sentence structure as described in item 13 of the scope of the patent application, as the natural sentences containing the target phrase"

ill Page 51 200530856 V. Description of the invention (45) Retrieving selected natural sentences "), the above natural sentence recognition is only different from each natural sentence selected in the search of step 1 2 2 to retrieve from the registration The natural sentences are retrieved from the natural sentences in the corresponding translation database based on the recognition retrieval situation. In the next step 1 2 6, ask whether the search in step 1 2 4 selects a consistent natural sentence. When the same translation sentence exists in the corresponding translation database in which only the target vocabulary replaces the other vocabulary, the above other vocabulary may be used to replace one optional phrase of the target vocabulary. Therefore, when the answer is affirmative, the program moves to step 1 2 8 to understand that the vocabulary replaces the search for the target vocabulary in the selected natural sentence as an optional vocabulary for the target vocabulary. The above steps 120 to 130 are equivalent to the identification device described in item 12 of the scope of patent application (as described in more detail in item 13 of the scope of patent application). Step by step, ask whether Gukou is the mark of the hui word in the surrogate. The clause is chosen by Γ \ hui zi. The number of clauses is complex. The son of the degree makes the sentence self-contained and self-contained. Repeatedly. Degree can} Cheng Zi's generation of the sentence is taken from the most important degree. Some of the most important processes are determined to be determined, the existence of the value of the algebraic number of words, the definiteness of the selected word, the original clause of the clause that has been selected and exceeded. 2ffl 3 CD-d processing 1 step to step by step attributive sentence Ken Ken film as a translation 3J mrc Therefore, most of them have the ability to register for reading资 的 致 致 _ 1 The corresponding value of the one The logarithmic phase is related to the special super-subordinate sentence C Ranzi's self-sentence translation The translation of the speech

Page 52 200530856 V. Description of the invention (46th generation), and contains alternative words that replace the subject vocabulary; read translation sentences for searching for translation of translation phrases and translation of selective vocabulary; It is displayed on the display 12 together with the translations that are deemed important and the marked optional vocabulary, also used as an optional vocabulary, and the process ends. In this example, the user can also know the appropriate translation for a particular translation phrase (getting a natural translation with a high probability of translation as a sentence in the target language) from the original sentence. The above steps 130 and 132 are equivalent to the first translation recognition device described in item 13 of the patent application scope. Therefore, in step 1 2 2, the natural sentences with the greatest degree of consistency and the optional words containing the words that replace the target between the natural sentences selected in step 102 are searched in the descending order of the degree of consistency. Read translated sentences of plural natural sentences (the number of certain natural sentences in a decreasing order of consistency, or all sentences with a specific value higher than the degree of consistency), and at the same time according to the degree of consistency The number of words is arranged on the display 12 in order. (Example 6) The above-mentioned program further describes the use of a real example. Assume that an original language is English and an object language is Japanese. When the word "have" is specified as a translated phrase that includes "have 1 unch" in the processed original sentence, one Natural sentences containing "ha ve breakfast" (use translated phrases in the same meaning

On page 53, the self-retrieval word Γ t is chosen because it is a check 1 noon. Eating sexually Γ eating Γ is optional for each of the standard inspections huihui huihui 200530856 V. Description of the invention (47) The natural sentence "have" as the processed original sentence) is registered in the corresponding No translation phrase containing "have lunch" is registered in the translation database. Even a survey based on the degree of consistency or correlation analysis cannot identify a proper translation sentence. The translation in the processed original sentence The phrase "have" or each vocabulary other than the one that has been determined to have a high relevance to the translated phrase (for example, "lunch") is assumed to be the target vocabulary, and the search contains each The natural word of the target word. Therefore, the word "lunch" which is the target word is retrieved, and the natural sentence includes, for example, "eat lunch" or "take a late lunch". From each natural sentence picked up by the above search, a natural sentence with only the vocabulary of the successful bid in the natural sentence difference is retrieved. Then, when the natural sentence used for picking is used as the content of a natural sentence, for example, "lunch", the content of the natural sentence, for example, "eat breakfast" or "take brunch" "Breakfast" "as a natural sentence with only the difference in the target vocabulary. When determining the word" breakfast "as one of the high-possibility vocabulary" lunch ", select it in the first search office The selected natural sentences (including the search of the translated phrase have) "natural sentence search) may replace the optional word vocabulary (" have breakfast "" translation sentence) Targeted word meal (breakfast) "and registered in the corresponding translation

Page 54 200530856 V. Description of the invention (48) Translation of the display and emphasis of the translation "have < taberu >" and the optional wording "breakfast & breakfast >" ◦ Therefore, in this example, a suitable natural sentence (translated sentence) is selected using the translated phrase "have" in the same meaning as the processed original sentence. When it is specified that "have" in the processed original sentence containing "have cats" is used as a translation phrase, the natural sentence containing "have dogs" is registered in the phase The corresponding translation database does not register a natural sentence containing "have cats". Even if it is based on the degree of consistency or similarity surveys in similar methods, an appropriate translation sentence cannot be identified. Pick the natural sentences (including the search of natural sentences with the phrase "have") between the 'hypothetical word bag' such as "cats" as the target vocabulary "and decide that the word" dogs "is It has the high possibility of using the same method to replace the target word "cats", and the target word "1¾ (cat s)" replaces the natural sentence of the optional word "dogs" (including "have dogs" (Have dogs) natural sentences) related translation sentences registered in the corresponding translation database, the translation "kau" with the emphasis translation phrase "have", and the marked options Vocabulary "Dog (d〇g s)" appears along with the translation of "i nu." Therefore, in the example, one of the appropriate natural sentences (translated sentences) with the same meaning in the translated phrase "have" was also selected as the processed original sentence.

Page 55 200530856 V. Description of the invention (49) (Example: Suppose the original language is Japanese and the original language is English. When the "kakeru" in the processed sentence "o-yu wo kakeru" is specified as the translation phrase , A natural sentence "mizu wo kakeru" (a natural sentence using the translated phrase "kakeru" in the same meaning as the processed original sentence) is registered in the corresponding translation database, but "〇-yu is not registered wo kakeru ", and even a survey based on the degree of consistency or relevance analysis cannot identify a suitable translated sentence, because registering other natural sentences, such as" denwa wo kakeru ", will assume the" o- "yu" is the target vocabulary, and natural sentences containing the target vocabulary are retrieved. Therefore, pick natural sentences such as "〇- yuni tsukeru", r 〇_yu wo wakasu "," 〇_yu wo nagasu "," 〇-yu de arau "," o-yu kara hikiageru ", and" 〇-yu de atatameru ". Therefore, from each of the natural sentences selected above, the search Natural sentences in the target vocabulary of natural sentence differences. The search results can be obtained by replacing the target vocabulary "o-yu" "~ ni tsukeru": r sosu ", factory shoyu", factory mizu "," sutsukesu ", and "Ude" r ~ wakasu ": None factory ~ wo nagasu" ... "denki", "mizu", "bangumi", and r uwasa "factory ~ de arau": r shampu "," suponji ",

Page 56 200530856 V. Description of Invention (50) Factory sh i om iz U ″ Λ factory miz U ″ and r sekke η ″ factory ~ kara h ik i ageru ″; factory in iz uj, ^ Γ ka it ei ”factory ka ig ij, and "zero (j) j factory ~ de atatame: ru" • factory densh i- -renjij and factory yonetsu "between the above vocabulary 1 This vocabulary factory mi zu" often appears and the sentence factory mi zu i NO kaker u ”exists o Therefore, the vocabulary“ m: iz 1 Jj ”is determined as a possible alternative vocabulary with the target vocabulary“ 〇- -yu ”in place of the processed original sentence“ 〇 一 yuwo kakeru ”ο because it In the first search (including the translation of the Baidi sentence factory kakeruj search), the relevant word vocabulary factory oy u between the selected natural sentences was replaced by the sentence factory mi zu wo kakeruj. Translate and register in the corresponding translation database. Tone translation phrase plant kakeru "The translation plant down (pour.)", As well as selective phrase labeled the "k a k e r u" of translation "water (w a t e r)" is displayed together, as the one shown below. "隹 J『 水 "入 ～ （pour * water 氺 into ～）" In the above translation sentence, "『 ... 」" ("氺 ··· *") means "..." is an optional word. Therefore, in this example, it is also chosen to use a suitable natural sentence (translated sentence) that is one of the translated phrases "kakeru" in the same meaning of the processed original sentence. In a similar way, when specifying the processed original sentence "3 wo

Page 57 200530856 V. Description of the invention (51)

), But "3 wo kakeru" is not registered, and even based on the degree of consistency or a survey of correlation analysis in similar methods, the appropriate translated sentence cannot be identified. The "3" vocabulary of the translation and target is replaced by r and the sentence r /… ------..... J kakeru "" kakeru ", record a natural sentence" 4 wo kakeru "in the corresponding translation database a (use the translation" kakeru "in the same meaning "The natural sentence as the processed original sentence p? ^ 1 heart κλ * ra,… kakeru" is related and registered in the first search wo • A quasi-corpse bad inspection = natural sentence (including translation / The corresponding translation database of my "kakeru" natural sentence search) is related to the important translation of the seventy-five "kakeru" translation ("multiply") and the use of two private words, such as "3" is an optional word that is marked with the target word "4" = "four" is displayed together, and it is determined that the word "r 4" is used as a replacement of the target word "3", which is one of the highest possible words. , "Multiply" by 4 (氺 f0UI > 氺) ”Therefore, in this example, we also chose to use one of the translated phrases“ kakeru ”in the meaning of the processed original sentence, as appropriate. Spoken and old Mu Ran sentences (translation :: aspect 'When in the target vocabulary difference' since the search of Ran sentence wide right than k matches the natural sentence (when the answer is no in step i 2 6 ') Natural sentences that replace selective vocabulary are not: there are multiple natural sentences with a degree of succinctness or similarity (when step 丨 3 0 is answered, there is a big one-T two to understand, bu, appropriate translated sentences. Therefore, the program Go to step to retrieve all selected natural

200530856 V. Description of Invention (52) Sentences, or translated sentences related to plural natural sentences in decreasing order of consistency (the number of natural sentences in decreasing order of consistency, or all have high Sentences with a specific value for the degree of consistency), and registered in the corresponding translation database, used to search each read translation sentence for the translation of the translation phrase, in accordance with the order of the degree of consistency and each with The number of inconsistent vocabularies in the descending order of the maximum consistent degree of translation of the emphasized translation phrase and the natural sentence agreement is found, the individual translated sentences are arranged on the display 12, and the process ends. In this example, the user can learn about suitable translations for some specific translation phrases. In the above example, although the mode that uses only the number of consistent words to calculate the degree of consistency is described, the mode is not limited to this embodiment. You can identify the formula for calculating the degree of consistency. When the number of consistent vocabularies increases, the degree of consistency also increases, as a natural sentence and the original sentence because of the calculation based on the number of consistent vocabularies and the number of inconsistent vocabularies. Increase the number of consistent vocabularies by assessing the degree of consistency. This mode is equivalent to item 5 of the patent application scope of the present invention. It is also possible to first calculate the degree of consistency based only on the number of consistent vocabularies. It is used to calculate the number of inconsistencies when it is difficult to select a single natural sentence even when performing a correlation analysis or a day-to-day analysis. Natural sentences with minimum number of inconsistent words (translated sentences). When assessing the degree of agreement, a natural sentence and a primitive

Page 59 200530856 V. Description of the invention (53) The degree of similarity in the order of the sentences is similar to the natural words of the original sentence. Increasing or evaluating inconsistent words in the consensus sentence, or the number between the number of existing sinks and the number of non-words. It also reduced the number of sentence patterns, and used the sequence order of 9 sentences in the program of the present invention to mention “Nan'an”, and when it was similar to the original equivalent of a rooted application, the word vocabulary of similarity was specifically understood. The phase makes the original sentence according to the degree of vocabulary of the original word used to evaluate the degree of consistency related to the original sentence. When the number of similarities between the natural sentence and the other consistent vocabulary increases, the use of the word invention between the stored sentences There are 10 inconsistencies around the consensus. However, the evaluation of the degree of consistency between the inconsistent vocabulary in the sentence and the original vocabulary is also arranged with the vocabulary. The number of vocabulary in the patent scope can be estimated by the number of the original sentences: Items, phases, such as, and interstitial, words, so that the degree of consistency to the word vocabulary is at a level of self-similarity and basis, the following are arranged in the same order as the original vocabulary, that is, according to the translation of the original sentence in the original sentence The linguistic differences (the number of vocabulary), each vocabulary in the original sentence provides a first assessment value. Define the first estimated value so that the first estimated value of the translated phrase in the original sentence is the largest, and in addition to the translated phrase in the original sentence, the first estimated value of each word is reduced to serve as the translation in the original sentence The difference caused by the increase in phrases (as an increase in the number of words in the middle). For example, when the original sentence contains nine vocabularies A to J (original sentence = A, B, C, D, E, F, G, H, and J), specify the fourth vocabulary D in the original sentence to start with As a translation phrase,

Page 60 200530856 V. Description of the invention (54) Each vocabulary A to J in the original sentence provides the following first estimated values: A 2 1 · 0, B 2 2. 0, C = 5 · 0, D 2 1 0 · 0, E = 5.0, F = 2.0, G = 1.0, H = 0.5, and 1.2 = 0.2. Next, each natural sentence that contains the search of step 102 is selected from the corresponding translation database. The natural sentences are used to search the original sentence and any vocabulary between each vocabulary in the natural sentence. The consistent vocabulary is used to calculate the sum of the first estimated values provided to each found consistent vocabulary. The sum of the first estimated values is assumed to be the degree of agreement (the sum of the first estimated values), and in accordance with the decreasing order of the degree of agreement of each related natural sentence, the retrieval of each natural sentence selected is retrieved. translate sentence. For example, when there are natural sentences 1 to 5 arranged by the following vocabulary selected by searching (where the vocabulary X is an arbitrary vocabulary): natural sentence 1 = (D, X, X, E, F, G) Natural sentence 2 = (D, X, E, F, G) Natural sentence 3 = (D, E, F, G) Natural sentence 4 = (A, B, X, C, X, D) Natural sentence 5 2 ( D, E, F, G, H, J) The sum of the first estimated values of each natural sentence: Natural sentence 1: 1 0 · 0 (Word D) + 0 (Word X) + 0 (Word X) + 5.0 (word vocabulary E) +2.0 (word vocabulary F) + 1 · 0 (word vocabulary G) = 18.0 natural sentence 2: 1 0 · 0 (word vocabulary D) + 0 (word vocabulary X) + 5 · 0 (word vocabulary E) + 2 · 0 (word vocabulary F) + 1. 0 (word vocabulary G) = 18 · 0 200530856 V. Description of the invention (55)

Natural Sentence 3: 1 0. 0 (Word D) + 5. 0 (Word E) + 2. 0 (Word F) + 1 0 (Word G) = 18. 0 Natural Sentence 4: 1 · 0 (Word A ) + 2 · 0 (Word B) + 0 (Word X)

+ 5.0 (word vocabulary C) +0 (word vocabulary X) +10.0 (word vocabulary D) 2 18.0 Natural sentence 5: 10.0 (word vocabulary D) +5.0 (word vocabulary E) +2.0 (word vocabulary F) +1.0 (word vocabulary G) +0.5 ( The vocabulary H) +0.2 (the vocabulary J) = 18.7 Of which the sum of the first estimated values (the degree of consistency) of the natural sentence 5 is the largest. The first estimated value is defined so that as the translation phrase in the original sentence and each vocabulary other than the translation phrase in the original sentence decrease, the value also increases. Therefore, as described above, based on the sum of the first estimated values, and each natural sentence containing the translated phrase is evaluated, the above natural sentence contains many vocabularies close to the position of the translated phrase in the original sentence. In other words, the estimated natural sentence has Contains many phrases that include translated phrases from the original sentence, as well as a high probability of including words that are close to the translated phrase, as natural sentences with a high degree of consistency. In the above embodiment, the natural sentences 1 to 4 have values of the same degree of agreement (the sum of the first estimated values). When a plurality of natural sentences have the same degree of consistency based on the values existing on the first estimated value, each word of the same vocabulary has the same degree of consistency as the second estimated value provided based on the first estimated value. Any vocabulary in the original sentence is consistent, the second estimated value is based on the translation piece in each natural sentence

Page 62 200530856 V. Description of the invention (56) Differences in terms (number of words) are provided. The second estimated value is also defined so that when the increase in translation phrases in each original sentence causes a difference, the second estimated value for the translated phrases in each original sentence is the largest (as an increase in the number of words in the middle ) ◦ Therefore, based on the first estimated value, calculate the sum of the second estimated value of each consistent vocabulary in each natural sentence that provides a numerical value of the same degree of consistency, and based on the first estimated value, according to each The descending order of the sum of the second estimated values of the related natural sentences (including the degree of agreement with the present invention), and the translation of each natural sentence with a value of the same degree of agreement is shown. The numbers are as follows, each of the total values of the unity is 4 JnJ. The first sentence of the first sentence is counted to the content statement, and the value is used as the example of the number. The interpretation of Yuguchi Piangugugu should be set to false, and the value is estimated in several cases.

4 少, minus 2 of 3, also r ▲ ο, value: ο number · If count 2 examples to estimate the confluence of the G number-as usual when he set a fake + 1i ΠZ f-tf 字 f 〒 字-fci1 ο Sentence 4. ο .G) ο I capsule 3: F capsule D) sink D) word D) huizihui 彳 huizi {字. 字 汇 字 汇 字

G + word 0) 2+ G +, +) sink E \) y X word X 5 capsule + X sink word

E sink word F sink word

Page 63 200530856 V. Description of the invention (57) Natural sentence 4: 0 · 2 (word vocabulary A) + 0 · 5 (word vocabulary B) +0 (word vocabulary X) + 2.0 (word vocabulary C) +0 (word vocabulary X) +10.0 (Word D) = 12.7 Therefore, the natural sentences 1 to 4 are displayed according to the descending order of the sum of the second estimated values (translated sentences), that is, according to the order of natural sentences 3, 2, 4, and 10. Define the second The estimated value will increase as the difference between the vocabulary in each natural sentence and the consistent vocabulary of any vocabulary in the original sentence that is consistent between the translated phrases in each natural sentence decreases. Therefore, a natural sentence contains a larger number of vocabulary consistent with the original sentence, and there is as little as possible inconsistent vocabulary between the consistent vocabulary, that is, based on the sum of the second evaluation values, by evaluating each containing translation The natural sentence of the phrase, the natural sentence has a natural sentence that includes the translated phrase in the original sentence, and the phrase may contain a vocabulary close to the translated phrase as a natural sentence with a high degree of consistency. For example, the following procedures can be used to understand the evaluation of the similarity of the order of the vocabulary in the natural sentence and the original sentence, and to consider the degree of agreement between the number of inconsistent vocabulary in the natural sentence that exists between the original sentence and the consistent vocabulary. : Pick and translate phrases from one of the upstream words group that exists in the upstream side of the original phrase in the original sentence (here, a previous word is adjacent to a translation phrase ) Have 200530856 V. Description of the invention (58) The vocabulary in front of one of the shortest distances, and the downstream word group from one of the downstream sides of the translation phrase existing in the original sentence In, pick and translate phrases that have one of the shortest kerning. Next, the location where the translation phrase exists is assumed to be a reference position in each natural sentence that contains the selected translation. (When a translation phrase exists in many plural places in a natural sentence, it exists in any translation sentence. The position of the language is assumed to be the reference position), from the reference position to the top of the vocabulary for which the previous vocabulary has been selected, to search for an area with a specific number of vocabularies (for example, with three vocabularies), and from the reference position to the Below the selected vocabulary, search for the area behind the number of specific vocabularies (for example, three vocabularies). Provide the natural words in the previous and subsequent vocabulary found in the search above to define the third assessment value, so that when the reference position and the previous vocabulary and between the reference vocabulary and the subsequent vocabulary As the space between words decreases, the value increases. Picking the face, the beginning of the middle and the beginning of the sub sentence, the original C in the short distance and the short of the word to pick the middle of the group of words with the face of the group face before the word from the face is not yet with, there are optional 'After selecting the Hui face word of the character C on the search group, the word “C” has not existed and has been merged with each other.} 片} Translate the words before the Hui face word

One of the most single inspections is the first one, the first one, and the one from the group.字 的 的 的 片} Interpretation of the vocabulary and translation of the word with the confluence of the face and the word of the choice of picking the noodles with the middle and the post in the middle of the sentence, the original inspection of the original description on the 及 and the distance {with the word capsule

Page 65 200530856 V. Description of the invention (59) The position of the previous vocabulary, to the previous search and found the natural vocabulary of the previous vocabulary and the following vocabulary in each natural sentence with the number of specific vocabulary areas (such as Three vocabularies), search for the previous vocabulary of the previous search, and from the position of the previous vocabulary of the previous search to the previous search and previous natural vocabulary and the following vocabulary in each natural sentence, with a specific vocabulary The number of fields is searched for the word after the previous search. Provide this search to find natural sentences with at least the preceding vocabulary and the following vocabulary, to define the third estimated value, so that when the previous search finds the position of the preceding vocabulary and the following vocabulary, and this search finds the previous As the kerning between the vocabulary and the position of the subsequent vocabulary decreases, the value also increases. You can use one of the same definitions to estimate the value based on the previous search and the position of the next word, and the distance between the previous word and the position of the next word. Described. However, considering the gap between the translated phrase in the original sentence and the word vocabulary before or after the search, a third estimated value can be defined so that when the word vocabulary before or after the word search in the original sentence, and As the kerning between translated phrases decreases, the influence of the preceding or following vocabulary on the third estimate will increase. Instead of defining the third estimated value, when the previous search finds the position of the previous word and the subsequent word, and the current search finds that the kerning between the previous word and the position of the subsequent word decreases, the value also changes with The increase, as mentioned earlier, can define a third estimated value such that when the above reference position and this

Page 66 200530856 V. Description of the invention (60) When the kerf between the previous vocabulary and the position of the subsequent vocabulary was found to decrease, the third estimated value also increased. From the original sentence in accordance with the decreasing order of repeating the above procedure, select the previous vocabulary and the subsequent vocabulary, until there is no longer any original vocabulary and subsequent vocabulary that can be selected from the original sentence. Next, the total of the third estimated value is calculated and provided to each natural sentence and the degree to which the hypothesis is consistent, and in accordance with the decreasing order of the degree of consistency of each corresponding natural sentence (the total of the third estimated value), and Display the translation of each natural sentence retrieved. For example, when an original sentence contains five words A to E (original sentence = (A, B, C, D, E)), and the third word C in the original sentence is specified as a translation phrase, Select the first vocabulary B and D from the original sentence as an upper vocabulary and a lower vocabulary, respectively, to search the area with a specific number of vocabularies from the translation of the previous vocabulary to the upper (for example, three Vocabulary), and from the corresponding translation database t to retrieve the translation phrase of each subsequent natural vocabulary in the natural sentence containing the translation phrase to search below for a region with a specific number of vocabulary (for example, (Three vocabularies) to provide a third estimate to the preceding vocabulary and subsequent vocabulary found in natural sentences. Therefore, the first vocabulary A and E are selected from the original sentence as an upper vocabulary and a lower vocabulary, respectively, from the translation phrase of the previous vocabulary A to the upper to search for a region with a specific number of vocabularies (for example, the Three words), and from the natural sentence

Page 67 200530856 V. Description of the invention (61) The preceding word vocabulary B and the following word vocabulary D are followed by the word vocabulary E and the word vocabulary D behind is used to search for a region with a specific number of word vocabularies (for example, with three word vocabularies) ) To provide a third estimate to the natural vocabulary A and the subsequent vocabulary E. Provide the above procedure to the following ten natural sentences (a) to (k), and arrange the sentences in descending order of the sum of the third estimated values, as shown below, where the vocabulary X and "/" represent one Arbitrary words and punctuation. 1. Natural sentence (a) two (A, B, C, D, E) 2. Natural sentence (b) = (A, B, C, D, /, B, C, D, E) 3. Natural sentence (C) = (B, C, D) 4. Natural sentences (d) = (A, B, C, X, D, E) 5. Natural sentences (e) = (A, B, C, X, D ) 6. Natural sentences (f) = (A, B, C, /, C, D, E) 7. Natural sentences (g) = (A, X, B, C, X, D) 8. Natural sentences ( h) = (B, C, /, C, D) 9. Natural sentence (j) = (B, X, C, /, C, X, D) 10. Natural sentence (k) 2 (B, X, X, C, /, C, X, X, D) As mentioned above, the third estimated value is provided to each natural sentence containing the translated phrase, and it is evaluated based on the sum of the third estimated value. The order of the vocabulary has a high degree of similarity, and the number of inconsistent vocabularies does not exist as much as possible between the vocabulary consistent with the original sentence, that is, it is possible to evaluate natural sentences that include as little as possible inconsistent vocabulary.

Page 68 200530856 V. The natural sentence of the original sentence in the case of invention description (62), as a natural sentence with a high degree of consistency. You can understand the similarity of the order in which the vocabulary is considered between the natural sentence and the original sentence, and the number of inconsistent vocabularies in natural sentences that exist between the vocabulary consistent with the original sentence. For example, the following procedure: First, from the original In the translation phrase in the sentence (this time the previous vocabulary is adjacent to the translation phrase), the previous vocabulary group is selected to have the shortest distance from the translated vocabulary, so it is assumed that the translation phrase exists in the natural sentence The position is the reference position of each natural sentence containing the translation phrase selected from the corresponding translation database (when the translation phrase exists in each of the plural places in the natural sentence, it is assumed to exist in The position in any translation phrase is the reference position). From the reference position to the specific number of areas above it, search for the previously selected word. The program calculates the number of vocabulary from kerning (more specifically, the number of first vocabulary from kerning and the number of second vocabulary from kerning), and measures the reference position and The kerning between the previous vocabulary, and the number of the first vocabulary is determined by the kerning based on each measured distance.

Page 69 200530856 V. Description of the invention (63) The number of words with a specific word above each natural sentence from the position where the selected previous word was found in the previous search to the previous word found in the previous search Area (for example, with three vocabularies), search for the previous vocabulary of the previous check, and measure the kerning between the position of the previous vocabulary found in the previous search and the previous vocabulary found this time, and the previous previous search Find the kerning between the previous vocabulary and the reference position, and add the measured kerning to get the kerning between the previous vocabulary and the reference position found in this search, and according to the distance obtained each time The number of first words is determined by the kerning. Check the previous vocabulary in descending order of the kerning of the translation phrases in the original sentence, and repeat the above procedure until there are no longer any optional preceding sacs in the original sentence. For example, when an original sentence contains five vocabularies A to E (original sentence = (A, B, C, D, E)), and the fourth vocabulary D in the original sentence is specified as a translation phrase, first The vocabulary C selected from the original sentence is used as the previous vocabulary, and among the natural sentences containing the translation phrases selected in the corresponding translation database, the number of specific vocabularies from the translation phrase to the top The area searches for the previous vocabulary, and the number of the first vocabulary is determined by the kerning in the natural sentence of the found vocabulary C. Here is an example of the result of calculating the number of the first vocabulary, as shown in the following table. Table 1. An example of the result of calculating the number of vocabulary when the kerning of the word vocabulary C is retrieved.

Page 70 200530856 V. Description of the invention (64) An example of the kerning of the natural vocabulary kerning = 0 1 2 3 CD 1 0 0 0 CXD 0 1 0 0 CXXD 0 0 1 〇 The preceding word comes first, before The middle face sentence of Hui Congzi 'is the B face of the middle face before the word selection of the word check. The original c $ domain of Huiyuan District, which has been searched for, has a unique edge of the word. Look for "Zi 7 sentences" in front of you \) y ε Ran Hui from the word _ 3 of the B three hui, the example of the word such as C before the current number issued by the hui, two words in the Β word are listed in this table. The number of words is as follows. The final result of the word vocabulary is determined by the number of words in the word space. · Table 2 is used to calculate the word space during the period of searching word word B. Calculate the number of word words. Examples of results.

Page 71 200530856 V. Description of the invention (65) An example of the kerning of a natural sentence kerning kerning = 0 1 2 3 4 BCD 2 0 〇0 〇BXCD 1 1 0 0 0 BXXCD 1 0 1 〇0 BCXD 0 2 0 0 0 BXCXD 0 1 1 0 0 BXXCXD 0 1 0 1 0 BCXXD 0 0 2 0 0 BXCXXD 0 0 1 1 0 BXXCXXD 0 0 1 0 1 Further, check the vocabulary A from the original sentence as the previous vocabulary. Previously, In the natural sentence of the previous vocabulary B found in the search, search for the previous vocabulary A from the previous vocabulary B to the area with the number of specific vocabularies (for example, three vocabulary) above. The kerning in the sentence determines the number of first characters. An example of the result of the kerning here is to calculate the number of the first vocabulary, as shown in Table 3 below. Table 3. Calculating the kerning of the period of the vocabulary A, and calculating the number of vocabulary.

Page 72 200530856 V. Description of the invention (66) Examples of results. An example of a natural sentence vocabulary of 5: kerning = 0 1 2 3 4 5 6 ABCD 3 0 0 0 0 0 0 AXBCD 2 1 0 0 0 0 0 AXXBCD 2 0 1 0 0 0 〇 ABXCD 1 2 0 0 0 0 〇AXBXCD 1 1 1 0 0 〇0 AXXBXCD 1 1 0 1 0 0 0 ABXXCD 1 0 2 0 0 0 0 AXBXXCD 1 0 1 1 0 〇0 AXXBXXCD 1 0 1 0 1 0 0 ABCXD 0 3 0 0 0 0 0 AXBCXD 0 2 1 0 〇0 0 AXXBCXD 0 2 0 1 0 0 0 ABXCXD 0 1 2 0 0 0 〇AXBXCXD 0 1 1 1 0 0 0

200530856 V. Description of the invention (67) 1 AXXBXCXD! 0 1 1 1 1 1 0 1 1 1 0 _ 1 0 ABXXCXD 0 1 0 2 0 0 0 AXBXXCXD 0 1 0 1 1 0 0 AXXBXXCXD 0 1 0 1 0 1 0 ABCXXD 0 0 3 0 0 0 0 AXBCXXD 0 0 2 1 0. 0 0 AXXBCXXD 0 0 2 0 1 0 0 ABXCXXD 0 0 1 2 0 0 0 AXBXCXXD 0 0 1 1 • 1 0 0 AXXBXCXXD 0 0 1 1 0 1 0 ABXXCXXD 0 0 1 0 2 0 0 AXBXXCXXD 0 0 1 0 1 1 0 AXXBXXCXXD 0 0 1 0 1 0 1

IRlilll Page 74 200530856 V. Description of the invention (68) Search in a method different from the number of the first vocabulary mentioned above, in each natural sentence containing the translation phrase selected in the corresponding translation database Measure the kerning between the kerics, and add the kerning to the number of the second kerning. More specifically, in the case where the existence of the vocabulary before the adjacent translation phrase is ignored between the preceding vocabulary groups in the original sentence (every natural sentence containing the vocabulary C is regarded as an inconsistency of an arbitrary vocabulary X) , The translation phrase selected by the previous vocabulary group in the original sentence has the shortest kerning distance, and the position of the translation phrase assumed to exist in the natural sentence is the translation database corresponding to the search The reference position of each natural sentence of the selected translation phrase is used to search for the area with a specific number (for example, three words) from the reference position to the top of the selected previous vocabulary. In previous natural words found in previous searches, measure the kerning between the reference position and the previous word (more specifically, the number of inconsistent words between the reference position and the previous word), and assume The kerning obtained by adding "1" to the measured kerning is the kerning of the reference position. According to the kerning of each reference position, the kerning is used to determine the number of the second kerning. Has not selected from the previous vocabulary, and selected from the previous vocabulary group and has the shortest kerning distance with the translation phrase, and found the previous vocabulary from the previous search in the area with the number of specific vocabulary , Search for the previously selected previous vocabulary in each of the natural sentences in the position where the previous vocabulary was found in the previous search, and in the natural sentence where the previous vocabulary was found in this search, measure the previous one found in the previous search Vocabulary and this

Page 75 200530856 V. Description of the invention (69) Find the kerning between the previous vocabulary, add the kerning between the previous vocabulary found in the previous search and the reference position to the measured kerning, and Obtain the kerning between the previous vocabulary and the reference position before the search, and determine the number of the second vocabulary by the kerning according to the obtained kerning. The previous vocabulary is checked in descending order of the kerning of the translated phrases in the original sentence, and the above process is repeated until there is no longer any selectable previous vocabulary in the original sentence.

Clause beginning 丨 \ ly original E 1, D when, C, B, for example, A, C, C, ξ, fourth sentence in the beginning of the Central Plains, M, the sentence _ the beginning of the W, a five-inclusive package, and the union字 之 语 片片 翻翻 为 为 以 字 的 前 之 片片 译 # Nearest neighbor slightly huqunhuihui the face of the character, the C clause in the previous memory, the original first suffix, and the four character forms The order of each column is as follows. The amount of Ego and the result of JB are merged into the word. The test of the original parameter is based on the position of the child in the distance check word. Convergence of the number of eyes is two. The distance measurement is made up of four A. One of the natural sentences on the grid table is an example of the kerning between B and D (actual kerning +1) (kerning between A and B) A Kerning with D (BD + AB) ABCD 2 〇2 AXBCD 2 1 3 AXXBCD 2 2 4 AB (XCD 3 0 3

Page 76 200530856 V. Description of the invention (70) Based on the above measurement results, the number of second vocabulary is determined by the kerning to get the result of calculating the number of second vocabulary from the kerning, as shown in Table 5. Table 5 is an example of the result of calculating the number of second words from the kerning. An example of a natural sentence Kerning Kerning = 0 1 2 3 4 5 6 ABCD • 0 0 2 0 0 0 0 AXBCD 0 0 1 1 0 0 0 AXXBCD 0 0 1 〇1 0 0 ABXCD 0 0 0 2 0 0 0 As mentioned above, when in each natural sentence containing a translation phrase, the number of first vocabulary by kerning and the number of second vocabulary by kerning are completed in relation to the preceding vocabulary group in the original sentence, In each natural sentence, the number of digits is selected, the number of digits in the distance is calculated by the number of digits, and the word is selected by the distance. One of the words, and one of the ratios of the head to head in the calculation of the eye distance meter in the small calculation

The form of the C distance is changed to ο 4, the estimate is ο the last one evaluated is L, B is used, and A 5 / (V is the conclusion of the fruit. The distance from the word is as an example.

D by and

Page 77 200530856 V. Description of the invention (71) The number of word vocabulary in the number of kerning of the first word is 3, and the natural sentence has the number of vocabulary of 0 when the kerning is 0, and the number of vocabulary of 0 when the kerning is 1, When the kerning is 2, the number of vocabularies is 2, as shown in Table 5. Therefore, the number of the first vocabulary with the largest number of vocabulary when the kerning is 0 is selected as the final evaluation of the vocabulary group preceding the natural sentence (A, B, C, D). The above embodiment is applied to this example. The number of vocabularies is different from the number of kerning first vocabularies and the number of kerning second vocabularies. When the kerning is 0, the number of vocabularies 0 is the same as each other. In the case of kerning 1, the number of kerning at kerning is sequentially compared, and the number of first kerning selected by kerning and the number of second kerning selected by kerning are compared in the same kerning. The number of large vocabularies is used as the final evaluation. From the kerning of the sentence to the natural sentence (C, A, kerning, kerning, kerning, the number is 2, but the last B, D in the sentence), depending on the number of settlement of the number of the second vocabulary The order of the vocabulary is similar to the vocabulary part of the vocabulary. When the number of the first vocabulary is 0, the number of the vocabulary of 1 is 1, and the number of vocabulary is 0. The kerning time of the word is evaluated by the kerning second word. In the case of (C, X, X, D) in the above example, the aim is to correctly evaluate and originality, and to change the order from the original sentence. For example, the natural sentence? The word 2 is in ο in ο, which is the confluence of numbers in the middle distance @ the number of words in the distance of the word in the distance of the word in the distance of 11 in the distance. The same is The word in front of the group is written as οο, because the number of items is described here because

C The Bth child / Γ / Γ dissimilarity, A Al ^ J / aT However, the distance L from the word c will be evaluated by the child, and the word A Λ c

D is the view

Page 78 200530856 V. Description of the invention (72) (X, A, B (remove the word vocabulary C), D) and evaluate the result of calculating the number of the second vocabulary from the kerning. Therefore, the original sentence of the above-mentioned natural sentence in the case of higher integrity is evaluated. When executing the above procedure for searching each natural sentence of the selected translation phrase in the corresponding translation database to determine the final evaluation, use the one that exists in the original sentence of each of the above natural sentences. The following vocabulary groups below the translation phrase are executed similarly (the number of first vocabularies is calculated by kerning and the number of second vocabularies is calculated by kerning, and the final evaluation of the subsequent vocabulary groups is determined) . Then, calculate each natural sentence to get the last evaluation of the previous vocabulary group and the last evaluation of the subsequent vocabulary group (from the same kerning, assuming the number of vocabularies calculated by kerning in each final evaluation), to Get an extensive assessment of one of every natural sentence. Based on the extensive evaluation of each natural sentence, the translated sentences of each natural sentence are displayed in the decreasing order of the kerning between the words. More specifically, for example, when the kerning is 0, the arrangement of each natural sentence is repeated in the decreasing order of the number of vocabularies, and when the kerning is 1 in the decreasing order of the number of vocabularies, the The subsequent arrangement of natural sentences with the same number of vocabularies, while increasing the kerning equivalent to the number of compared vocabularies at the time of kerning 1, according to the descending order of the kerning between the vocabularies indicated by extensive evaluation. Arrange each sentence and display relative

Page 79 200530856 V. Description of the invention (73) The translated sentence for each sentence that is missing. Based on this point 5 based on the evaluation of the extensive white I, the following nine missing / ,,,, sentences (a) to (j) shown are processed as mentioned above, in order to arrange Bai Xi k /, ,, sentence 1 the following Results in the sequence • 1. White missing sentence (a) = (Α, Β, C, D, 'Ε) 2. White missing sentence (b) (Α, Β, C, D, | /, Β, C, DE) 3. Pk sentence (c) (Α, Β, C, X, D) 4. Missing sentence (d) (Β, C, D) 5. White Pk sentence (e) = (Α, X, B, C, X, D) 6. Wonderful; sentence (f) = (Β, C, X, D) 7. White Pk sentence (g) = (Β, C, /, C, D) 8. Bai Que M, , Sentence (h) (Β, X, C, /, C, X, D) 9. White Pk sentence (j) (Β, X, X, C, /, C, X, 1,,] D) above The number of the second vocabulary counted by the vocabulary in the procedure § is not needed. When it is not necessary to evaluate the Bairan sentence from the original sentence When arranging the partial vocabulary in the child, you can omit the number of the second vocabulary calculated by the kerning and use the number of the first vocabulary calculated by the kerning as the final evaluation. 0 As usual, based on the first and second estimated values The third / fourth estimated value / the fourth estimated value, the degree of agreement, and the selection or arrangement of natural sentences can usually be evaluated, selected, or arranged by a correlation analysis or a plan analysis using natural sentences. When retrieving a translated sentence for reference as a phrase with a narrower range than the entire original sentence, based on the first

Page 80 200530856 V. Description of the invention (74) The estimated value and the second estimated value, or the third estimated value are more suitable for this example, and are used to evaluate the pattern of the degree of consistency of natural sentences. For example, this pattern exists in the original comparison. Sentence phrases, etc., use only the number of consistent vocabulary between the original sentence and the natural sentence to assess the degree of consistency. For example, a search mode can be provided, such as a phrase search mode, and there is also a general search mode, and based on the first estimated value and the second estimated value or the third estimated value, the evaluation of the degree of consistency and the current phrase are performed. Choose natural sentence selection or rearrange when searching mode. Although an example in which the degree of agreement is first calculated to select a natural sentence (identify the corresponding translation), when a suitable natural sentence cannot be restricted based on the degree of agreement only, a correlation analysis (c ο rre 1 ati ο η ana 1 ysis) select a natural sentence (recognize the corresponding translation), and when a suitable natural sentence cannot be restricted even based on the above-mentioned correlation analysis, a natural sentence is selected by a scheme analysis, this application It is not limited to this. It can calculate the consistency of the natural sentences of the translation phrases selected by the retrieval device, and perform correlation analysis and day-to-day analysis to evaluate individual natural sentences and select a natural sentence according to the priority order, such as the following table The results shown in No. 6 Middle School are based on the evaluation results. Table 6. Example of choosing a natural sentence priority order.

Page 81 200530856 V. Description of the invention (75)

Correlation analysis of the degree of consistency (the existence of highly relevant vocabulary) Planning analysis (the existence of selective vocabulary) Whether the Bureau is 1 No 2 5 Low Yes 3. No 4 6 Furthermore, the degree of consistency of individual natural sentences is calculated to perform each Correlation analysis and day-to-day analysis, and can choose natural sentences with the most extensive evaluation scores. Extensive evaluation score = a X (—degree of consistency) + b X (correlation evaluation score) + c X (daily evaluation score)

The correlation evaluation score based on the result of the correlation analysis is provided to one of the individual natural sentences, and the evaluation score based on the result of the plan analysis is provided to one of the individual natural sentences, and a, b, and c It is a predefined coefficient.

Page 82 200530856 V. Description of the invention (76) The program in the translation recognition process described earlier is expressed by a formula (Figures 2 A and 2 B), which asks whether the natural sentence corresponds to the translation retrieved in step 102. The processed original sentences selected by the database are exactly the same (step 104), and if the answer is affirmative, the translated sentences of natural sentences that are exactly the same as the processed original sentences are read and displayed. However, this application is not limited to this. The above steps 104 and 106 can be ignored, and regardless of whether any natural sentence is exactly the same as the original sentence retrieved and selected in step 102, the procedure in step 108 can be performed unconditionally and Subsequent steps (procedures, such as calculating the degree of consistency). In this mode, when a plurality of natural sentences are selected and retrieved in step 102, if the natural sentences that are completely consistent with the original sentence are included in the plurality of selected natural sentences, the degree of agreement of the natural sentences is the greatest. Therefore, the same program results can be obtained when performing steps 104 and 106. According to the number of natural sentences and the translated sentences registered in the corresponding translation database, the correctness of a translation in the present invention is selected, and as the number of natural sentences and translated sentences increases, the correctness of selecting a translation also increases. . Therefore, it is possible to read the translated original sentence and the translated sentence in the translation recognition system of the present invention, and automatically register and read the natural sentence and the translated sentence at the same time when the corresponding translation database is executed before or after it is executed by itself. According to the present invention, only natural sentences and translated sentences in a corresponding translation database need to be automatically registered, and no editing or processing is required, such as dividing a natural word by a word unit. Therefore, the corresponding translation database is easily and automatically registered at the same time (natural sentences begin with

Page 83 200530856 V. Description of the invention (77) and translated sentences). Although the example of the translation mode that emphasizes the translation of phrases is described above, when a translation sentence is displayed, the mode is not limited to this. As a matter of course, the mode does not emphasize that the translation of the translated phrase is also included in the appropriate score of the present invention. The correctness of selecting a translation in the present invention is also based on the natural sentences registered in the corresponding translation database and the redundant words (r e d u n d a n c y) of the capacity of the translated sentences. When emphasizing the natural sentences registered in the corresponding translation database and the redundant words in the capacity of the translated sentence, choose a correct translation and register in the corresponding translation database (the capacity of the corresponding translation database) The number of natural sentences and the number of translated sentences decreases proportionally. Therefore, when each natural sentence and translated sentence registered in the corresponding translation database is searched for natural sentences and translated sentences with high similarity of content, and natural sentences and translated sentences with high similarity are found, you can Provides the ability to remove a set of natural sentences and translate sentences from the corresponding translation database. Although the foregoing example of choosing to translate using a single corresponding translation database has been described, the application is not limited to this. When the processed original sentence is limited to a specific field, such as translating the main menu of a specific field and using this program, the corresponding translation database can be divided by the field of natural sentences. Therefore, it is possible to avoid an increase in the capacity of individual corresponding translation databases, and to obtain, for example, saving time for searching natural sentences and improving selection

Page 84 200530856 V. The utility of invention description (78) The correctness of alternative translation Although the personal computer 10 is an example of a model for executing commands and other programs according to a specific personal computer 10, a computer system 38 can be used. The terminal-network 30 in the spoken computer system 38, such as the Internet (Local Area Network, storage medium 36, etc.), is connected to the server so that when a user asks through it, the server 34 recognizes and Refer to Figure 4. Refer to Figure 4 to show the query on the ground (the original text of the sentence, the original text of the example to the original text in the language to the location (uniform can be installed in this system, and at the same time access the translation database of " ), And operate as the above-mentioned user through the above to display the translation of the translation phrase is the present invention is not limited to this: again, said not to invent to assemble (conf igure), and the machine 3 2 and a Servers 3 and 4 are both connected to the C Internet) or a local area network (LAN), and include a hardware device server 3 4 to store the corresponding translation information. Terminal 32 is online (Online). Close Appropriate translation, as shown in computer system 38 in Figure 4. In order to enable a user to examine the map = (of (1)), and specifically specify an original # such as' through the user through the terminal 3 2 to transfer the original servo For the operation of 3 or 4, or the user connection includes the original server 34 web page (consistent resource resource URL: URL). In this example,% 1, so that when the server 34 executes the translation recognition program, access 1 ng) is stored in the storage medium 36 and the corresponding original text is identified ((2) in Figure 4) ^ and it will be another translation (sentence ) Pass to the terminal

Page 85 200530856 V. Description of the Invention (79) To answer this query (refer to (3) in Figure 4). Because this installation allows multiple users to share the translation recognition function of the server 34 (and the corresponding translation database), especially when the corresponding translation database needs a large capacity, it can effectively reduce costs. Although the example of the mode in which the present invention is applied to an electronic dictionary system is described above, the application is not limited to this. As a matter of course, the present invention is also applied to a machine translation (automatic translation) system, an automatic interpretation system (a system that understands an acoustically input original sentence using auditory interpretation, recognizes the translated sentence of the original sentence that is understood, and transmits Auditory-recognized translated sentences) and a language learning system, etc. Figure 5 is a schematic diagram showing the installation of a machine translation system 40 provided by the present invention. Referring to Fig. 5, an original sentence input section 4 2 inputs a translated original sentence (original data) to select both a translation phrase 4 4 and a translation sentence assembly 4 6. When a translation phrase 4 4 is selected, a specific phrase in the original sentence that has been input is repeatedly selected as a translation phrase, and at the same time, another phrase is selected in the translated original sentence, and asked A translation recognition section 48 translates by retrieving one of the selected translation phrases. The translation recognition part 48 is a part equivalent to the translation recognition system of the present invention, and executes the translation recognition program, and simultaneously accesses the corresponding translation database stored in the storage medium 50 to identify the translation of the requested translation phrase. And pass the recognized translation to the translation sentence set part 4 6.

Page 86 200530856 V. Description of the invention (80) Transmit the translation phrase based on the part used to select the translation phrase 4 4 and the translation sentence set part 4 6 connect the translation transmitted from the translation recognition part 4 8 (if necessary, change the arrangement The sequence) is to gather the translated sentences in the original language, the translations transmitted by the translation recognition section 48 and the original sentences of the translated phrases transmitted by the original sentence input section 42. The translated sentences collected in the translated sentence collection part 46 are transmitted to the translated sentence output part 52 for executing the translated sentence output program, such as being displayed on a display, recording to a storage medium, and an audible output. In the above-mentioned installation, since the present invention is applied to the recognition of translation phrases, a translation having a high probability of obtaining a naturally translated sentence as one of the sentences in the original sentence is obtained from the original sentence, and the sentence is translated from The translated sentence transmitted in the output section 52 becomes a natural translated sentence as one of the original sentences. Furthermore, the procedure can be simplified because it does not require complicated procedures, such as identifying a part of the report or analyzing the substructure. As usual, referring to Figure 5, through this procedure, the resulting translated sentences can be transmitted online. It has been explained, for example, recognition and collective translation to answer online queries of translated sentences, as shown in Figure 4. Although one of a pair of related sentences (a pair of sentences in different languages) is registered in the corresponding translation database (storage device), and the other parts are called a natural sentence and a translated sentence, the first language to the second In the language, the corresponding translation database of the translation (identification of translated sentences) can be used as the database of the translation from the second language to the first language (identification of translated sentences).

P.87 200530856 V. Description of the Invention (81) Although the examples in English, Japanese, and Chinese are described as an example of an original language or an object language, they are not limited to these languages. As a matter of course, it can be applied to, for example, French, German, Italian, Spanish, and Portuguese as the original language or the target language. The present invention has been described above with reference to the preferred embodiments, but it is not intended to limit the scope of patent rights claimed by the present invention. The scope of patent protection shall depend on the scope of the attached patent application and its equivalent fields. Anyone skilled in this field can make changes or modifications without departing from the spirit or scope of this patent, which are all equivalent changes or designs made in the spirit disclosed by the present invention, and should be included in the scope of patent application described below. Inside. 200530856 Brief description of the drawings [Simplified illustration of the drawings] Fig. 1 is a schematic diagram showing the configuration of the personal computer in the preferred embodiment. Figures 2A and 2B are flowcharts showing the contents of the translation recognition program. Figure 3 is a table showing examples of natural sentences and corresponding translations, which includes a vocabulary "operation" (operation) registered in a translation database. Figure 4 is a description of a mode of applying a computer system of the present invention, Grid diagram of a terminal connected to a server. Figure 5 shows a grid diagram of the configuration of a machine translation system to which the present invention is applied. [Description of main component symbols] Personal computer 1 0 Central processing unit 1 0 A Read memory 1 0 B Random access memory 1 0 C Input / output port 1 0 D Bus 1 0 E Display 1 2

Page 89 200530856 Brief description of the keyboard 14 Mouse 1 6 Hard disk device 1 8 CD read-only memory device 20 Scanner 2 2 CD read-only memory 24 Network 30 Terminal 3 2 Server 34 Storage media 3 6 Computer system 3 8 Machine translation system 4 0 Original sentence input part 4 2 Translation phrase 4 4 Translation sentence set part 4 6 Translation recognition part 4 8 Storage media 5 0 Translation sentence output part 5 2 Step 1 0 0, 1 0 2, 1 0 4, 1 0 6, 1 0 8, 1 1 0, 1 1 2, 1 1 4 Steps 1 1 6, 1 1 8, 1 1 9, 1 2 0, 1 2 2, 1 2 4, 1 2 6, 1 2 8, 132, 134 130

Claims (1)

200530856 6. Scope of patent application 1. A system for identifying a corresponding translation, the system includes: a storage device for storing a plurality of natural sentences including a plurality of words in the original language, the natural sentence and A translation sentence of a target language is associated; a retrieval device for retrieving a natural sentence from a plurality of natural sentences stored in the original language of the storage device, where the natural sentence contains a translation of the original sentence of the original language A phrase; and a first translation recognition device for evaluating the degree of consistency of the natural sentence selected by the retrieval device, and identifying a translation of the phrase in at least the translated sentence of the natural sentence as the translation For the translation of the phrase, the natural sentence is selected based on the assessment of the degree of consistency. 2. The system for identifying a corresponding translation as described in item 1 of the scope of the patent application, wherein the retrieval device retrieves from the plurality of natural sentences stored in the original language of the storage device exactly the same as the original sentence A consistent natural sentence; and when a natural sentence that completely matches the original sentence is selected by a search by the retrieval device, the first translation recognition device recognizes the translated sentence of the natural sentence that completely matches the original sentence as The translated sentence of the original sentence. 3. If the system for identifying a corresponding translation is described in item 1 of the scope of the patent application, wherein the translation recognition device calculates the search selected by the retrieval device 200530856 VI. The scope of the patent application Natural sentence 'the value of the vocabulary consistent with the original sentence' and evaluate the degree of consistency with the original sentence, so that when the number of vocabulary consistent with the calculation increases, the degree of consistency with the original sentence varies Its improvement. 4. If the system for identifying a corresponding translation is described in item 3 of the scope of patent application, wherein the first translation recognition device divides the value of the vocabulary consistent with the calculation by the value of the vocabulary including the translated phrase in the calculation, To calculate the degree of consistency. 5. If the system for identifying a corresponding translation is described in item 3 of the scope of the patent application, the translation recognition device also calculates the value of the vocabulary inconsistent with the original sentence selected by the retrieval of the retrieval device, Evaluate the degree of agreement with the original sentence, so that when the number of words that are consistent with the calculation decreases, the degree of inconsistency with the original sentence increases. 6. If the system for identifying a corresponding translation is described in item 3 of the scope of patent application, when calculating the value of the consistent vocabulary, the first translation recognition device does not include the pre-defined frequently from the calculated vocabulary. Words that appear. 7. If a system for identifying a corresponding translation is used in item 3 or item 5 of the scope of patent application, when calculating the number of the consistent vocabulary or the number of the inconsistent vocabulary, the vocabulary is regarded as consistent Vocabulary Page 92 200530856 VI. Scope of patent application Other devices are used to calculate vocabulary with different suffixes due to different singular / plural or tense. 8. For the system for identifying a corresponding translation in item 3 of the scope of patent application, when calculating the number of the consistent vocabulary, the first translation recognition device does not need to repeatedly calculate the consistency that appears two or more times. Vocabulary. 9. If the system for identifying a corresponding translation is described in item 3 or 5 of the scope of patent application, wherein the first translation recognition device evaluates the relationship between the natural sentence selected by the retrieval device and the original sentence The similarity of the vocabulary arrangement order, and the degree of agreement with the original sentence is evaluated, so that when the similarity of the vocabulary arrangement order increases, the degree of agreement with the original sentence increases accordingly. 10. If the system for identifying a corresponding translation is described in item 3 or 5 of the scope of the patent application, wherein the first translation recognition device is in a vocabulary consistent with the natural sentence selected by the retrieval device, Calculate the number of inconsistent vocabularies, and evaluate the degree of agreement with the original sentence, so that when the number of inconsistent vocabularies in the calculated coincidence decreases, the degree of agreement with the original sentence increases. 1 1. If the system for identifying a corresponding translation in item 1 of the scope of patent application, further includes a search device for searching for a natural sentence selected by the retrieval of the retrieval device for the translation phrase, and Often appears in the primitive language Page 93 200530856 VI. Words that appear frequently in the same sentence as stated in the scope of the patent application; where the first translation recognition device refers to specific frequently-occurring phrases found by the search device and existing in the original sentence, and the reference contains A translation sentence of a natural sentence of each of the translated phrases, the translated sentence in the search for a natural sentence selected by a search device that includes each of the translated phrases, and a frequently used translation phrase The translation of certain frequently occurring phrases is used to identify and find frequently occurring translations as a translation of the translated phrase in the original sentence. 1 2. If the system for identifying a corresponding translation according to item 1 of the scope of the patent application, further includes an identification device for identifying the optional phrase in the original sentence, and replaces the search that does not include the retrieval device The selected phrase in the selected natural sentence, wherein the first translation recognition device recognizes the translation of the translation phrase in at least the translated sentence of the natural sentence, and the translation sentence of the natural sentence includes each recognized by the recognition device The optional phrase and the translated phrase in the natural sentence selected by the retrieval of the retrieval device are used as a translation of at least the translated phrase in the original sentence. 13. If the system for identifying a corresponding translation is described in item 12 of the scope of patent application, the identification device searches for a plurality of natural sentences of the storage device stored in a natural sentence containing the subject vocabulary, and searches for stored in The natural sentence in the storage device of the natural sentence having the same sentence structure is used as the natural sentence selected by the search, and the self-selected sentence selected by the search is identified Page 94 200530856 6. Scope of Patent Application The phrase in place of the subject phrase in the sentence is used as the optional phrase. 1 4. A system for identifying a corresponding translation, the system includes: a storage device for storing a plurality of natural sentences contained in a plurality of words in an original language, and the translation of the natural sentence and a target language Sentence association, a retrieval device for retrieving a natural sentence from a plurality of natural sentences stored in the original language of the storage device, the natural sentence including a translation phrase of the original sentence of the original language; a search device, A natural sentence selected by the retrieval of the retrieval device for the translation phrase and a frequently occurring translation appearing in the same sentence of the original language; and a second translation recognition device for searching for a message containing The translation sentence of the natural sentence of each of the translation phrases, and the search for a specific frequently-occurring phrase for a frequently-occurring translation phrase, the translation phrase refers to the ones found in the search device and existing in the original sentence. The particular frequently occurring phrase, and referring to each of the natural sentences picked up by the retrieval device contained in the retrieval Natural sentence translated sentence of the translated phrases and for identifying the found frequently appearing translation, as the original sentence Translation of phrases. 1 5. If the system for identifying a corresponding translation is provided in item 1 or item 14 of the scope of the patent application, wherein when the translation phrase includes a plurality of words, the retrieval device retrieves from the stored in the storage device the Among the plural natural sentences in the original language, the search contains at least one plural word including the translation of the phrase Page 95 200530856 6. Scope of Patent Application Natural sentence of Huihui. 16. A method for identifying a corresponding translation, the method includes: a first step in which each of the natural language of the original language including a plurality of words stored in a storage device is retrieved A sentence, the natural sentence being associated with a translation sentence of an object language; and a second step in which the consistency between the selected natural sentence retrieved in the first step and the original sentence is evaluated Degree, and at least the translation of the phrase in the translation sentence of the natural sentence selected based on the evaluation of the degree of consistency is identified as the translation of at least the phrase in the original sentence. 17. A method for identifying a corresponding translation, the method includes: a first step in which each natural sentence containing an original language of a plurality of words stored in a storage device is retrieved , The natural sentence is associated with a translation sentence of an object language; a second step, in the second step, the first step retrieves the selected natural sentence system for searching the translation phrase and frequently appears in a Frequently occurring vocabulary in the same sentence of the original language; and a third step, in which the translated sentence containing the natural sentence of each of the translated phrase and a specific frequently occurring phrase is used to search Recognize the translation of frequently occurring translation phrases and find the frequently occurring translations as the translation of the translation phrase in the original sentence. The translation of the translation phrase refers to the original sentence and the specific frequently found in the second step appeared Page 96 200530856 VI. Patent application phrase, and in the first step of searching the selected natural sentence, refer to the translation of the natural sentence containing each of the translated phrases. 1 8. —A program for connecting a computer to a storage device that stores, from a plurality of natural sentences of the original language stored in the storage device, a plurality of vocabularies related to translations in a target language A plurality of natural sentences in one of the original languages, the program having a function of a retrieval device to retrieve natural sentences of translated phrases contained in the original sentences of the original language; and a first translation recognition device for evaluating the retrieval The device searches for the degree of agreement between the selected natural sentence and the original sentence, and is used to identify the identification translation of the translation phrase in the translation sentence of at least the natural sentence as the translation phrase in the original sentence. Translation, and the natural sentence is selected based on the degree of consistency of the assessment. 1 9. A program for connecting a computer to a storage device, from the storage device, from a plurality of natural sentences of the original language stored in the storage device, storing a plurality of vocabulary that are consistent with a translation in an object language A plurality of natural sentences in an original language to have a function of a retrieval device, the function of the retrieval device is to retrieve the natural sentences of the translated phrases contained in the original sentences of the original language; a search device for searching the Retrieval of translation phrases by the retrieval device. Natural sentences picked up frequently appear in the same sentences as the original language; and 200530856 6. Patent application scope-a second translation recognition device for searching The translated sentence of the natural sentence of each of the translated phrases, as well as the search for a specific frequently-occurring phrase for a frequently-occurring translation phrase, and used to identify the frequently-occurring translation as the translation piece in the original sentence Language translation, the particular frequently-occurring phrase refers to the use of the search device to discover and exist in the original The phrase sentence in particular often appear, as well as reference translation of each sentence contains natural language translation of the film. Chapter 98