[go: up one dir, main page]

WO2018080228A1 - Serveur pour traduction et procédé de traduction - Google Patents

Serveur pour traduction et procédé de traduction Download PDF

Info

Publication number
WO2018080228A1
WO2018080228A1 PCT/KR2017/011991 KR2017011991W WO2018080228A1 WO 2018080228 A1 WO2018080228 A1 WO 2018080228A1 KR 2017011991 W KR2017011991 W KR 2017011991W WO 2018080228 A1 WO2018080228 A1 WO 2018080228A1
Authority
WO
WIPO (PCT)
Prior art keywords
translation
language
accuracy
path
optimal
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/KR2017/011991
Other languages
English (en)
Korean (ko)
Inventor
박성국
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Neopyxis Co Ltd
Original Assignee
Neopyxis Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Neopyxis Co Ltd filed Critical Neopyxis Co Ltd
Publication of WO2018080228A1 publication Critical patent/WO2018080228A1/fr
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/40—Processing or translation of natural language

Definitions

  • the voice recognition-based interface adopts language, which is a basic means of communication between people, and thus enhances user experience and accessibility of the interface.
  • the amount of acquired training data becomes an important factor that makes a difference in the performance of the translation engine.
  • the difference in the amount of the learning data may cause a performance gap of the translation engine between languages. If the performance gap of the translation engine widens, the development of translation technology for languages in regions with poor IT (Information Technology) environments may be delayed. Therefore, development of translation technology considering the performance of translation between languages is required.
  • Embodiments attempt to increase the translation accuracy of translation techniques.
  • Embodiments seek to bridge the gap in translation services between languages.
  • Embodiments attempt to improve the quality of translation services through feedback from users related to translation.
  • Embodiments provide an optimal translation path in consideration of translation accuracy between languages.
  • Server for translation includes a database for recording translation accuracy of the translation engines for each language; And a processor for generating an optimal translation path between an input language and a target language, the processor determining at least one relay language and at least one optimal translation engine from the database based on the input language and the target language; The optimal translation path specified by the at least one relay language and the at least one optimal translation engine may be generated.
  • the processor may include: an optimal translation engine for translating text of the input language into text of the at least one relay language; At least one optimal translation engine for translation between texts of the at least one relay language; And an optimal translation engine for translating text of a relay language into text of the target language from the translation engines based on the translation accuracy for each language.
  • the processor determines, based on the input language and the target language, one optimal translation engine from the translation engines, and based on translation accuracy corresponding to the determined optimal translation engine, at least One relay language may be determined, and the optimal translation path may be generated based on the determined at least one relay language.
  • the server may further include a communication module configured to receive the text of the input language and the target language, wherein the processor is configured to convert the text of the input language into text of the target language and the target language.
  • An optimal translation path may be generated for translating the text of the text into the text of the input language.
  • the processor comprises: translation engines and relay languages for translating text of the input language; Translation engines and relay languages for translation between text in said at least one relay language; And at least one translation engine selected by the user from at least one of translation engines for translating the text of the relay language into the text of the target language and the at least one relay language.
  • a translation method may include: obtaining a first translation accuracy between an input language and a target language; Obtaining second translation accuracies between the input language and a plurality of languages; Obtaining third translation accuracies between the plurality of languages and the target language; And generating an optimal translation path for translation between the input language and the target language based on the first translation accuracy, the second translation accuracy and the third translation accuracy.
  • the generating of the optimal translation path may include selecting a relay language from the plurality of languages based on the second translation accuracy and the third translation accuracy; Comparing the first translation accuracy with a fourth translation accuracy of a translation path that is a sequence of the input language, the relay language, and the target language; And determining one of a translation path corresponding to the first translation accuracy and a translation path corresponding to the fourth translation accuracy based on the comparison result as the optimal translation path.
  • the selecting of the relay language may include selecting a translation path having the highest translation accuracy among translation paths corresponding to the plurality of languages; And determining a language belonging to the selected translation path as the relay language.
  • the fourth translation accuracy comprises: translation accuracy between the input language and the relay language; And a product of translation accuracy between the relay language and the target language.
  • the second translation accuracy and the third translation accuracy are in a critical accuracy range.
  • generating the optimal translation path may include determining a translation path corresponding to the first translation accuracy as the optimal translation path when the first translation accuracy exceeds a threshold accuracy. Can be.
  • the generating of the optimal translation path may include determining a translation path corresponding to the first translation accuracy as the optimal translation path when the second translation accuracy and the third translation accuracy are less than a threshold accuracy. It may include the step.
  • the translation accuracy between a first language and a second language comprises: an amount of retention of translation data between the first language and the second language; A word order matching rate between the first language and the second language; Whether a capacity of a character of the first language matches a capacity of a character of the second language; And based on at least one of feedback corresponding to a translation between the first language and the second language, wherein the first language and the second language include the input language, the target language, and the plurality of languages. can do.
  • the generating of the optimal translation path may include translating time corresponding to the first translation accuracy; Translation time periods corresponding to the second translation accuracies; And reflecting at least one of the translation time periods corresponding to the third translation accuracy to at least one of the first translation accuracy, the second translation accuracy, and the third translation accuracy, to generate the optimal translation path. It may include the step.
  • Translation method comprises the steps of receiving texts of the input language; Recognizing complexity corresponding to the texts; And generating translation paths corresponding to the complexity, wherein generating the optimal translation path may include generating the optimal translation path based on the generated translation paths. .
  • Translation method comprises the steps of receiving a text of the input language; Generating translation paths based on the first translation accuracy, the second translation accuracys and the third translation accuracys; And outputting texts in which the text is translated using the translation paths, and generating the optimal translation path comprises determining a translation path selected from a user among the translation paths as the optimal translation path. It may include.
  • a translation method for translating text of an input language into a target language may include selecting at least one relay language within a relay language group including a plurality of languages; Translating text of the input language into the at least one relay language to generate relay text; And translating the relay text into the target language, wherein selecting the at least one relay language comprises: a first translation accuracy between the input language and the target language; Second translation accuracies between the input language and the plurality of languages; And selecting the at least one relay language based on third translation accuracies between the plurality of languages and the target language.
  • the server obtains a first translation accuracy between an input language and a target language, obtains second translation accuracies between the input language and a plurality of languages, and between the plurality of languages and the target language. Obtain a third translation accuracy of and generate an optimal translation path for translation between the input language and the target language based on the first translation accuracy, the second translation accuracy and the third translation accuracy It may include a processor.
  • Embodiments can increase the translation accuracy of a translation technique.
  • Embodiments can bridge the gap in translation services between languages.
  • Embodiments may improve the quality of the translation service through feedback of the user related to the translation.
  • Embodiments may provide an optimal translation path in consideration of translation accuracy between languages.
  • FIG. 1 is a diagram illustrating a server for translation, according to an exemplary embodiment.
  • FIG 2 illustrates an optimal translation path according to an embodiment.
  • FIG. 3 is a diagram for describing an optimal translation path, according to an exemplary embodiment.
  • FIG. 4 is a diagram for describing an optimal translation path, according to an exemplary embodiment.
  • FIG. 5 is a flowchart illustrating a translation method, according to an exemplary embodiment.
  • FIG. 6 is a diagram for describing translation accuracy between languages, according to an exemplary embodiment.
  • FIG. 7 is a diagram for describing a process of generating an optimal translation path, according to an exemplary embodiment.
  • FIG. 8 is a diagram for describing a translation method, according to an exemplary embodiment.
  • FIG. 9 is a diagram for describing a translation method, according to an exemplary embodiment.
  • FIG. 10 is a flowchart illustrating a translation method according to an embodiment.
  • FIG. 11 is an exemplary diagram of a configuration of a server for translation according to an embodiment.
  • first or second may be used to describe various components, but such terms should be interpreted only for the purpose of distinguishing one component from another component.
  • first component may be referred to as a second component
  • second component may also be referred to as a first component.
  • Embodiments may be implemented in various forms of products, such as personal computers, laptop computers, tablet computers, smart phones, televisions, smart home appliances, intelligent cars, kiosks, wearable devices, and the like.
  • the embodiments may be applied to recognize a user in a smart phone, a mobile device, a smart home system, and the like.
  • Embodiments may be applied to a payment service through user recognition.
  • the embodiments may be applied to an intelligent vehicle system that automatically starts the vehicle by recognizing the user.
  • FIG. 1 is a diagram illustrating a server for translation, according to an exemplary embodiment.
  • the server 101 may be linked with the translation engines 102 to generate an optimal translation path for translation.
  • the server 101 is a server that processes operations for generating a translation path, and may be implemented as a software module, a hardware module, or a combination thereof.
  • the server 101 may execute an application or a program for performing a translation method or a method for generating a translation path, and may load an application program recorded in an internal memory.
  • Application programs recorded in the internal memory may be executed through wireless communication with the server.
  • the translation engines 102 may be subjects for processing an operation of translating text of a specific language into text of another language, and include translation engines provided for each country and translation engines developed to provide a translation service.
  • the server 101 may include a database, a processor, and a communication module.
  • the database here records the translation accuracy for each language of the translation engines 102
  • the processor can execute a program, control the server 101
  • the communication module can input text, target language, text of the input language and target. Receive text in a language.
  • the communication module receives the text of the input language and the target language from the user terminal 103 and transmits it to the translation engines 102, and receives the text of the target language from the translation engines 102 to receive the user terminal ( 103).
  • the user terminal 103 may be connected to the server 101 by wired or wireless communication to execute an application or program for translation.
  • the translation assistance apparatus 104 may receive a voice of an input language and transmit it to the user terminal 103, and receive and output a voice of an interpreted language from the user terminal 103.
  • the translation assistance device 104 may be connected to the user terminal 103 through short-range communication.
  • the server 101 may determine at least one relay language and at least one optimal translation engine from the database based on the input language and the target language.
  • the server 101 may generate an optimal translation path specified by at least one relay language and at least one optimal translation engine.
  • the server 101 may provide an automatic translation portal service that generates an optimal translation path by linking between different translation engines and translates the text of the received input language into the text of the target language through the optimal translation path.
  • the server 101 may be a gateway server that provides a translation service in association with the translation engines 102.
  • the server 101 may generate an optimal translation path corresponding to the input language and the target language so as to provide an optimal translation quality based on the translation accuracy and translation speeds of the translation engines.
  • the server 101 may generate an optimal translation path within any one of the translation engines 102.
  • the server 101 determines any one optimal translation engine from the translation engines 102 based on the input language and the target language and determines at least one relay language based on the translation accuracy corresponding to the determined optimal translation engine. Can be.
  • the server 101 may generate an optimal translation path based on the determined at least one relay language.
  • the server 101 may select a translation engine A among the translation engines 102 and generate an optimal translation path within the translation engine A, in order to translate the text of Korean as the input language into the text of Chinese as the target language. have. For example, server 101 may determine translation paths "Korean-> Chinese” and "Korean-> English-> within translation engine A based on translation accuracy and translation indices between languages of translation engine A. One of "Chinese”, "Korean-> English-> Chinese” can be determined as the optimal translation path.
  • the server 101 may generate an optimal translation path to which different translation engines among the translation engines 102 are linked.
  • the server 101 may determine from the translation engines 102 an optimal translation engine for translating the text of the input language into the text of the at least one relay language based on the translation accuracy by language.
  • the server 101 may determine at least one optimal translation engine from the translation engines 102 for translation between texts of at least one relay language based on language-specific translation accuracy.
  • the server 101 may determine from the translation engines 102 an optimal translation engine for translating the text of the at least one relay language into the text of the target language based on the translation accuracy by language.
  • the server 101 may generate an optimal translation path through the translation engine B and the translation engine C among the translation engines 102 in order to translate the text of Korean as the input language into the text of Vietnamese as the target language. For example, the server 101 combines a translation path of "Korean-> English" through the translation engine B and a translation path of "English-> Vietnamese" through the translation engine C, so that "Korean-> English-> Vietnamese To generate an optimal translation path. At this time, the server 101 may determine the translation engine B and the translation engine C as optimal translation engines.
  • FIG 2 illustrates an optimal translation path according to an embodiment.
  • the server may generate an optimal translation path for translating text of Korean as an input language into text of English as a target language.
  • the server may generate an optimal translation path for translating the English text into the Korean text in the reverse order of the generated optimal translation path.
  • the server may determine the translation engine A and the translation engine B as optimal translation engines based on the translation accuracy and translation speeds of the translation engines recorded in the database.
  • the server may set the translation engine A to correspond to a path for translating Korean into Japanese, and the translation engine B may be set to correspond to a path for translating Japanese into English.
  • the server may adopt the pre-generated optimal translation path as it is and provide the translation in the reverse order.
  • the server may set the English text to be translated into Japanese text by the translation engine B and the Japanese text to be translated into Korean text by the translation engine A.
  • FIG. 3 is a diagram for describing an optimal translation path, according to an exemplary embodiment.
  • the server may generate an optimal translation path for translating text of Korean as an input language into text of Chinese as a target language.
  • the server may generate an optimal translation path for translating the text of Chinese, the target language, into the text of Korean, the input language, independently of the previously generated optimal translation path.
  • the server may generate an optimal translation path employing a translation engine A to translate Korean text into Chinese text.
  • the server generates an optimal translation path in which the Chinese text is translated into Japanese text by translation engine B, and the Japanese text is translated into Korean text by translation engine C, in order to translate Chinese text into Korean text. can do.
  • optimal translation paths between forward and reverse languages between the input language and the target language may be generated independently of each other.
  • the server may generate bidirectional optimal translation paths as shown in FIG. 2 in consideration of time or speed for generating an optimal translation path, or generate optimal translation paths independent of FIG. 3 and B.
  • FIG. 4 is a diagram for describing an optimal translation path, according to an exemplary embodiment.
  • the server may generate an optimal translation path based on a user's selection.
  • the server may output translation engines and relay languages for translating an input language of Korean into a relay language.
  • the server may output translation accuracy and translation speeds corresponding to translation engines and relay languages to induce a user's selection.
  • the translation accuracy can be updated based on the evaluation results by the users.
  • the evaluation result may be defined as discretely distinguished or continuous values, such as "upper, upper, middle, middle, lower and lower".
  • the server may generate an optimal translation path based on the translation engine selected from the user and the relay language.
  • the server may output translation engines and relay languages for translation between the texts of the relay languages, and generate an optimal translation path based on the translation engine and the relay language selected from the user.
  • the server may output translation engines for translating the text of the relay language into the target language, and generate an optimal translation path based on the translation engine selected from the user.
  • the server may select a translation engine A for translating Korean text and Japanese as a relay language based on a user's selection result, and a translation engine B and a relay language for translating Japanese text. You can select Chinese separately, and choose a translation engine C for translating Chinese text.
  • the server may generate an optimal translation path based on the user's selection result in "Korean-> Japanese (Translation Engine A)-> Chinese (Translation Engine B)-> English (Translation Engine C)".
  • the number of relay languages may be a preset value or may be adaptively set according to a user's selection. If the relay language is not selected by the user, the server may generate an optimal translation path to translate the text into the text of the target language according to the selection result of the translation engine.
  • FIG. 5 is a flowchart illustrating a translation method, according to an exemplary embodiment.
  • the server for translation may obtain a first translation accuracy between an input language and a target language (501).
  • the server is a server that processes operations for translation and may be implemented as a software module, a hardware module, or a combination thereof.
  • the server may execute an application or a program for performing the translation method, and load the application program recorded in the internal memory.
  • Application programs recorded in the internal memory may be executed through wireless communication with the server.
  • the server may access the server based on a wireless communication network
  • the wireless communication network for accessing the server may be a communication standard of a mobile communication device such as Wifi, 2G, 3G, 4G, 5G, and LTE, and next generation communication standards.
  • Smart phones, tablet PCs, notebook computers, wearable devices, portable devices, and the like capable of executing various applications through communication with a server.
  • the input language is a language of text to be translated and may be, for example, a country-specific language such as Korean, Chinese, English, or Spanish.
  • text includes characters, words and sentences expressed in the input language.
  • the text of the input language may be expressed by Korean characters, words, and sentences such as " ⁇ ", "school", and "hello".
  • the target language means that other language when the text of the input language is to be translated into the text of another language. For example, if the target language is German, the text "hello" of the input language may be translated into the text "Hallo" of the target language.
  • the server may receive the selected information from the user to identify the input language and the target language.
  • the server may receive a voice of an input language and recognize the input language based on the received voice to identify the input language.
  • the server may convert the voice of the recognized input language into the text of the input language.
  • the server may detect a voice of the input language and recognize the input language based on the detected voice pattern.
  • the server may receive the text input from the user and recognize the input language based on the pattern of the received text.
  • the server may also receive text in the input language from an external server or device.
  • FIG. 6 is a diagram for describing translation accuracy between languages, according to an exemplary embodiment.
  • the server may obtain a first translation accuracy between the input language and the target language from a pre-built database.
  • the database may be implemented as a memory included in the server or an external device such as a server that can be connected to the server by wire, wireless, or network.
  • Translation accuracy A between a plurality of languages can be recorded in a database as shown in FIG. 6.
  • the server may identify the input language and the target language and obtain a first translation accuracy between the input language and the target language from the database. For example, if the input language is Korean and the target language is English, the server may request A ( 1, 2) can be obtained.
  • the translation accuracy is a parameter regarding how accurate the translation between the two languages is.
  • the translation accuracy may be set to a different value depending on the direction of the translation (eg, first language-> second language or second language-> first language), or may be defined as a value independent of the direction of translation.
  • the unit may be defined in various ways according to the design intention, for example, it may be expressed in%.
  • the translation accuracy between the first language and the second language may include an amount of retention of translation data between the first language and the second language; A word order agreement rate between the first language and the second language; Whether the capacity of the characters of the first language matches the capacity of the characters of the second language; And feedback corresponding to the translation between the first language and the second language, based on at least one element, and the translation accuracy may be updated as these elements are updated.
  • the capacity of a character means a capacity required to express a character of a specific language.
  • Korean Hangul can be represented by 2 bits
  • English alphabet can be represented by 1 bit, so the capacity of the Hangul and the alphabet does not match.
  • Translation accuracy may be defined by weighting elements that affect translation accuracy.
  • the translation accuracy may be defined by differently setting the weight of the retained amount of translation data and the weight of the word order matching ratio.
  • the method of defining the translation accuracy may be applied by applying various techniques according to the design intention.
  • the translation accuracy may be an indicator related to the performance of the translation engine providing translation between specific languages.
  • the translation engine may be an independent server or device separate from the server according to an embodiment.
  • the server may transmit the text of the input language to the translation engine, and receive the text of the target language as the translation result from the translation engine.
  • the translation accuracy recorded in the database may be translation accuracy of the plurality of translation engines.
  • the server may process an operation of selecting translation engines for translating the text of the input language into the text of the target language based on the translation accuracy recorded in the database. Translation accuracy can be obtained from the execution subject of the translation engine and recorded in the database.
  • the database may record translation time corresponding to translation accuracy.
  • the translation time required may be recorded in a relation mapped to the translation accuracy.
  • the server may select translation engines for translating text in the input language into text in the target language, taking into account translation accuracy and translation time requirements.
  • the server may obtain second translation accuracy between the input language and the plurality of languages (502).
  • the plurality of languages may include languages other than the target language among languages recorded in the database. For example, if the input language is Korean and the target language is German, the server may select languages other than German from among a plurality of languages recorded in the database, and obtain second translation accuracy between Korean and the selected languages.
  • the server may obtain third translation accuracies between the plurality of languages and the target language (503). For example, if the input language is Korean and the target language is German, the server may select languages other than German from among a plurality of languages recorded in the database, and obtain third translation accuracies between the selected languages and German.
  • the server may generate an optimal translation path for translation between the input language and the target language based on the first translation accuracy, the second translation accuracy, and the third translation accuracy (504).
  • the translation path may be defined as a sequence of languages for translating text of an input language into text of a target language
  • an optimal translation path means an optimized translation path generated by a server. For example, if the input language is a Korean target language is German, the translation path may be expressed as a sequence of languages from the input language to the target language such as "Korean-> English-> French-> Spanish-> German". .
  • the server may select a relay language from the plurality of languages based on the second translation accuracy and the third translation accuracy.
  • the relay language means a language between the input language and the target language among the sequence of languages constituting the translation path.
  • the relay language may include at least one language.
  • the server may select a translation path having the highest translation accuracy among translation paths corresponding to a plurality of languages, and determine a language belonging to the selected translation path as a relay language.
  • the translation paths corresponding to the plurality of languages may include the following examples.
  • the server may select the translation path 1 and determine the languages "Japanese" and "English” belonging to the selected translation path 1 as the relay languages.
  • the accuracy of the translation path may include: translation accuracy between the input language and the at least one relay language; Translation accuracy between at least one relay language; And translation accuracy between the at least one relay language and the target language.
  • the accuracy of a translation path may be defined as the product of translation accuracy between languages in the sequence of languages included in the translation path.
  • the server may obtain translation accuracy between the plurality of languages recorded as shown in FIG. 6 and calculate translation accuracy of the plurality of translation paths.
  • the plurality of translation paths are a sequence of "input language-> at least one relay language-> target language”.
  • the server selects a translation path having the highest translation accuracy among the calculated translation accuracy.
  • the server may determine a language belonging to the sequence as the relay language.
  • FIG. 7 is a diagram for describing a process of generating an optimal translation path, according to an exemplary embodiment.
  • the server may be configured to perform either of a translation path that is a sequence of "input language-> at least one relay language-> target language” and a translation path of "input language-> target language” based on at least one of translation accuracy and translation time. Can be determined as the optimal translation path.
  • the translation accuracy of the translation path which is a sequence of "input language-> target language”
  • the translation accuracy of the translation path which is a sequence of "input language-> target language”. 4 It will be referred to as translation accuracy.
  • the fourth translation accuracy may be defined as the product of translation accuracy between languages belonging to the sequence of the translation path.
  • the server may compare the first translation accuracy and the fourth translation accuracy, and determine one of a translation path corresponding to the first translation accuracy and a translation path corresponding to the fourth translation accuracy based on the comparison result. .
  • the server optimally translates a translation path 702, which is a sequence of "Korean-> Japanese-> English-> German," among translation paths for translation between Korean input language and German target language. Can be determined by the path.
  • the server may select a translation path 702 having the highest translation accuracy among the translation paths, and determine "Japanese” and "English” belonging to the selected translation path 702 as a relay language.
  • the translation accuracy of the selected translation path is referred to as the fourth translation accuracy
  • the translation accuracy of the translation path 701 of "Korean-> German” is referred to as the first translation accuracy.
  • the server may compare the fourth translation accuracy and the first translation accuracy, and determine the translation path 702 as the optimal translation path because the fourth translation accuracy is greater than the first translation accuracy.
  • Translation accuracy of "Korean-> German” is 95%
  • translation accuracy of "Korean-> Japanese” is 90%
  • translation accuracy of "Japanese-> English” is translation accuracy of "English-> German” Since the accuracy is 88%
  • the first translation accuracy is 50% and the fourth translation accuracy is 75.24% (which can be calculated as 0.95 * 0.9 * 0.88).
  • the translation accuracy is calculated as a product of the translation accuracy between the languages, but not limited to such a calculation technique, a variety of techniques such as applying a weight differentially or a penalty as the path is lengthened may be applied. If the translation performance is better when directly translating the input language from the input language to the target language than the translation accuracy or the time required for translation, the sequence including the relay language, the server determines the translation path which is a sequence of the input language and the target language. The optimal translation path can be determined.
  • the server may generate an optimal translation path 702 and translate the Korean text into the German text using the generated optimal translation path 702.
  • the translation operation of converting text into text of another language may be directly processed by the server or may be processed by an external server or device for each language belonging to the optimal translation path 702.
  • the server may access an integrated cloud server, and translation operations according to the sequence of optimal translation paths 702 may be processed on the cloud server.
  • the server may generate an optimal translation path 702 suitable for an input language that is Korean and a target language that is German based on the translation accuracy and translation time recorded in the database.
  • the server queries the translation engines recorded from the database based on the generated optimal translation path 702, and uses the query result to translate the translation engine "Japanese-> English" suitable for the translation of "Korean-> Japanese".
  • a translation engine suitable for and a translation engine suitable for translation of "English-> German" can be determined.
  • translation between languages belonging to the optimal translation path 702 is processed through a translation engine separate from the server, and the server may generate an optimal translation path 702 and relay the translated texts. have.
  • the server performs the translation process through the program recorded in the internal memory or the external translation engine based on the translation accuracy and the translation time recorded in the database. It can be determined. If the performance of the translation process through the program recorded in the internal memory is superior to the translation accuracy or translation time of the translation process through the external server or the translation engine, the server may process the translation operation locally.
  • the server may determine a sequence of the input language and the target language as the optimal translation path. For example, if the translation accuracy between the input language Korean and the target language Japanese is 95% and the threshold accuracy is 90%, the server determines the sequence of "Korean-> Japanese" as the optimal translation path and additional translation paths. Or do not perform a search of the relay language.
  • the server may select languages in which translation accuracy with an input language and translation accuracy with a target language are within a critical accuracy range among a plurality of languages, and select at least one relay language from among the selected languages.
  • the server may select languages in which the translation accuracy is in the range of 60% to 100% (or more than 60%) among the translation accuracy between the input language Korean and the plurality of languages.
  • the server generates translation paths that include the selected languages, and the generated translation paths can compare translation accuracy with each other.
  • the server may determine that the input accuracy is less than the threshold accuracy if the translation accuracy between the languages recorded in the database and the input language is less than the threshold accuracy or if the translation accuracy between the languages recorded in the database and the target language is less than the threshold accuracy.
  • Target language can be determined as the optimal translation path. For example, if the input language is Vietnamese, the target language is Indonesian, the translation accuracy between Vietnamese and plural languages is less than 60% and the translation accuracy between plural languages and Indonesian is less than 60%, Indonesia "sequence can determine the translation path.
  • the server may derive an optimal translation path with the following algorithm.
  • a (i, j) is the translation accuracy between language i and language j
  • N is the number of languages
  • i-> j is a command to create an optimal translation path for translation from language i to language j
  • 1node is a command for calculating the translation accuracy R1 of the translation path which is a sequence of "language i-> language j”
  • 2node is the translation accuracy R2 of the translation path which is a sequence of "language i-> language k”.
  • 3node is a command for calculating the translation accuracy R3 of the translation path, which is a sequence of "language i-> language k-> language l-> language j".
  • the server may determine a translation path corresponding to the largest translation accuracy among the translation accuracy of R1 to RM as an optimal translation path.
  • the above algorithm is merely an example, and the server may derive an optimal translation path by employing algorithms such as dynamic programming and tree based optimal path search.
  • the server may generate an optimal translation path based on not only the translation accuracy between languages belonging to the translation path but also translation time corresponding to the translation accuracy.
  • the weights applied to the translation accuracy and the translation time may be variously modified according to the design intention.
  • FIG. 8 is a diagram for describing a translation method, according to an exemplary embodiment.
  • the server may generate translation paths for each text.
  • the server may receive texts 801 of the input language.
  • the texts 801 may be divided into units such as sentences, phrases, and clauses.
  • the server can recognize the complexity 802 corresponding to the texts 801.
  • the server may recognize the complexity corresponding to the text based on the words included in the text, the number of words, the length of the text, the structure of the text, and the like.
  • the complexity is the complexity of the processing required for translation, and can be expressed in units that represent discrete or continuous values.
  • the server may generate translation paths 803 corresponding to the complexity 802.
  • the server may set the translation paths 803 differently depending on the complexity 802, where the translation accuracy and translation time recorded in the database may be taken into account.
  • the server may generate an optimal translation path based on the generated translation paths 803.
  • the server may process translations of the texts 801 in parallel using the translation paths 803, and generate an optimal translation path for processing the translations in parallel based on the translation accuracy and translation time required. Can be.
  • FIG. 9 is a diagram for describing a translation method, according to an exemplary embodiment.
  • a server may translate text of an input language into texts of a target language through a plurality of translation paths, and receive feedback of a user regarding the translated texts.
  • the server may receive the text of the input language and generate translation paths based on the translation accuracy and translation time. For example, the server may generate higher translation paths with higher translation accuracy.
  • the server may translate and output the text of the input language into the texts 901 to 903 of the target language using the generated translation paths.
  • the user may evaluate or select corresponding translation paths based on the output texts 901 to 903.
  • the server may be fed back a selection result or evaluation result of any one of the texts 901 to 903 or a feedback or selection result of any one of translation paths corresponding to the texts 901 to 903.
  • the server may determine a translation path selected from the user as an optimal translation path among translation paths corresponding to the texts 901 to 903 output from the user.
  • the server may update the translation accuracy or translation time recorded in the database based on the feedback result by the user.
  • FIG. 10 is a flowchart illustrating a translation method according to an embodiment.
  • the server may select at least one relay language within a relay language group including a plurality of languages (1001).
  • the server may further include: first translation accuracy and translation time between the input language and the target language; Second translation accuracies and translation time between the input language and the plurality of languages; And at least one relay language based on at least one of third translation accuracies and translation time between the plurality of languages and the target language.
  • the server may generate a relay text by translating text of an input language into at least one relay language (1002).
  • the server can generate the relay text locally.
  • the server may receive relay text from a translation engine providing translation processing of the input language and at least one relay language.
  • the server may translate the relay text into a target language (1003). As described above, the server may generate or receive text of the target language through a local method or a method employing a translation engine.
  • FIG. 11 is a diagram illustrating a configuration of a server according to an embodiment.
  • the server 1101 includes a processor 1102 and a memory 1103.
  • the processor 1102 may include at least one of the devices described above with reference to FIGS. 1 through 10, or may perform at least one method described above with reference to FIGS. 1 through 10.
  • the memory 1103 may store a program in which a translation method is implemented.
  • the memory 1103 may be a volatile memory or a nonvolatile memory.
  • the processor 1102 may execute a program and control the server 1101. Code of a program executed by the processor 1102 may be stored in the memory 1103.
  • the server 1101 may be connected to an external device (eg, a personal computer or a network) through an input / output device (not shown), and may exchange data.
  • the server may generate an optimal translation path or select a relay language in consideration of translation accuracy and translation time recorded in a database to provide translation between an input language and a target language.
  • an embodiment can increase the quality of translation services with large gaps between languages, and can create a translation path optimized for languages with poor quality of translation services due to poor IT environments.
  • an embodiment generates a translation path through the languages when the translation accuracy between specific languages is higher than other languages, and thereby translates the translation into higher quality than the direct translation path between the input language and the target language. Can be provided. Since raising translation quality between all languages is limited in terms of the amount or processing speed of translation data, one embodiment can break down language barriers between languages by creating an optimal translation path that is appropriate for input and target languages. .
  • one embodiment may apply a real-time translation service to a variety of fields, for example, one embodiment may be applied to a variety of technologies such as replacing the interpreter or inserting subtitles of foreign broadcasts in real time.
  • the embodiments described above may be implemented as hardware components, software components, and / or combinations of hardware components and software components.
  • the devices, methods, and components described in the embodiments may include, for example, processors, controllers, arithmetic logic units (ALUs), digital signal processors, microcomputers, field programmable gates (FPGAs). It may be implemented using one or more general purpose or special purpose computers, such as an array, a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions.
  • the processing device may execute an operating system (OS) and one or more software applications running on the operating system.
  • the processing device may also access, store, manipulate, process, and generate data in response to the execution of the software.
  • OS operating system
  • the processing device may also access, store, manipulate, process, and generate data in response to the execution of the software.
  • processing device includes a plurality of processing elements and / or a plurality of types of processing elements. It can be seen that it may include.
  • the processing device may include a plurality of processors or one processor and one controller.
  • other processing configurations are possible, such as parallel processors.
  • the software may include a computer program, code, instructions, or a combination of one or more of the above, and configure the processing device to operate as desired, or process it independently or collectively. You can command the device.
  • Software and / or data may be any type of machine, component, physical device, virtual equipment, computer storage medium or device in order to be interpreted by or to provide instructions or data to the processing device. Or may be permanently or temporarily embodied in a signal wave to be transmitted.
  • the software may be distributed over networked computer systems so that they may be stored or executed in a distributed manner.
  • Software and data may be stored on one or more computer readable recording media.
  • the method according to the embodiment may be embodied in the form of program instructions that can be executed by various computer means and recorded in a computer readable medium.
  • the computer readable medium may include program instructions, data files, data structures, etc. alone or in combination.
  • the program instructions recorded on the media may be those specially designed and constructed for the purposes of the embodiments, or they may be of the kind well-known and available to those having skill in the computer software arts.
  • Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks.
  • Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like.
  • the hardware device described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Machine Translation (AREA)

Abstract

La présente invention concerne un serveur et un procédé de traduction. Un serveur, selon un mode de réalisation, peut acquérir au moins une langue pont et au moins un moteur de traduction optimale à partir d'une base de données sur la base d'une langue d'entrée et d'une langue cible, et peut générer un chemin de traduction optimale spécifié par au moins une langue pont et au moins un moteur de traduction optimale. Un serveur, selon un mode de réalisation, peut générer un chemin de traduction optimale pour une traduction entre une langue d'entrée et une langue cible sur la base de taux de précision de traduction entre la langue d'entrée et la langue cible, de taux de précision de traduction entre la langue d'entrée et de multiples langues, et de taux de précision de traduction entre les multiples langues et la langue cible.
PCT/KR2017/011991 2016-10-27 2017-10-27 Serveur pour traduction et procédé de traduction Ceased WO2018080228A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
KR1020160140895A KR101719107B1 (ko) 2016-10-27 2016-10-27 번역을 위한 서버 및 번역 방법
KR10-2016-0140895 2016-10-27

Publications (1)

Publication Number Publication Date
WO2018080228A1 true WO2018080228A1 (fr) 2018-05-03

Family

ID=58583925

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/KR2017/011991 Ceased WO2018080228A1 (fr) 2016-10-27 2017-10-27 Serveur pour traduction et procédé de traduction

Country Status (2)

Country Link
KR (1) KR101719107B1 (fr)
WO (1) WO2018080228A1 (fr)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110008468A (zh) * 2019-03-04 2019-07-12 广州虎牙信息科技有限公司 语言系统管理方法、信息转换方法、服务器以及存储介质
CN111756825A (zh) * 2020-06-12 2020-10-09 引智科技(深圳)有限公司 实时云端语音翻译处理方法及系统
CN112818706A (zh) * 2021-01-19 2021-05-18 传神语联网网络科技股份有限公司 基于逆向结果稳定性的语音翻译实时争端记录系统与方法

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102637340B1 (ko) * 2018-08-31 2024-02-16 삼성전자주식회사 문장 매핑 방법 및 장치
CN110502762B (zh) * 2019-08-27 2023-07-28 北京金山数字娱乐科技有限公司 一种翻译平台及其管理方法
KR102542362B1 (ko) * 2019-09-27 2023-06-12 주식회사 엔디소프트 자동 번역 기능을 이용한 실시간 다국어 다중 번역 채팅 방법
CN114154522A (zh) * 2021-12-09 2022-03-08 上海开放大学 一种基于多种语言学习平台翻译智能处理方法和系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2001188789A (ja) * 1999-12-28 2001-07-10 Nec Corp 翻訳ならびに翻訳料分配方法及びそのシステム
KR20110043645A (ko) * 2008-07-03 2011-04-27 구글 인코포레이티드 기계 번역을 위한 파라미터들의 최적화
US20130226553A1 (en) * 2012-02-23 2013-08-29 Google Inc. Translating text using a bridge language
KR20150103367A (ko) * 2013-01-03 2015-09-10 업투데이트 인코포레이티드 데이터베이스 질의 번역 시스템
JP2016051433A (ja) * 2014-09-02 2016-04-11 日本電気株式会社 情報処理システム、翻訳方法、及びそのためのプログラム

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20100020381A (ko) * 2008-08-12 2010-02-22 오의진 2단계 번역에 의한 다국어 번역방법

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2001188789A (ja) * 1999-12-28 2001-07-10 Nec Corp 翻訳ならびに翻訳料分配方法及びそのシステム
KR20110043645A (ko) * 2008-07-03 2011-04-27 구글 인코포레이티드 기계 번역을 위한 파라미터들의 최적화
US20130226553A1 (en) * 2012-02-23 2013-08-29 Google Inc. Translating text using a bridge language
KR20150103367A (ko) * 2013-01-03 2015-09-10 업투데이트 인코포레이티드 데이터베이스 질의 번역 시스템
JP2016051433A (ja) * 2014-09-02 2016-04-11 日本電気株式会社 情報処理システム、翻訳方法、及びそのためのプログラム

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110008468A (zh) * 2019-03-04 2019-07-12 广州虎牙信息科技有限公司 语言系统管理方法、信息转换方法、服务器以及存储介质
CN110008468B (zh) * 2019-03-04 2023-08-01 广州虎牙信息科技有限公司 语言系统管理方法、信息转换方法、服务器以及存储介质
CN111756825A (zh) * 2020-06-12 2020-10-09 引智科技(深圳)有限公司 实时云端语音翻译处理方法及系统
CN112818706A (zh) * 2021-01-19 2021-05-18 传神语联网网络科技股份有限公司 基于逆向结果稳定性的语音翻译实时争端记录系统与方法
CN112818706B (zh) * 2021-01-19 2024-02-27 传神语联网网络科技股份有限公司 基于逆向结果稳定性的语音翻译实时争端记录系统与方法

Also Published As

Publication number Publication date
KR101719107B1 (ko) 2017-04-06

Similar Documents

Publication Publication Date Title
WO2018080228A1 (fr) Serveur pour traduction et procédé de traduction
WO2021141419A1 (fr) Procédé et appareil pour générer un contenu personnalisé en fonction de l'intention de l'utilisateur
WO2018074681A1 (fr) Dispositif électronique et procédé de commande associé
WO2018074716A1 (fr) Procédé et système pour recommander une interrogation à l'aide d'un contexte de recherche
WO2016093552A2 (fr) Dispositif terminal et son procédé de traitement de données
WO2016013885A1 (fr) Procédé d'extraction d'image et dispositif électronique associé
WO2016208941A1 (fr) Procédé de prétraitement de texte et système de prétraitement permettant de mettre en œuvre ledit procédé
WO2020055156A1 (fr) Système et procédé pour générateur de scène
WO2020159140A1 (fr) Dispositif électronique et son procédé de commande
WO2015050321A1 (fr) Appareil pour générer un corpus d'alignement basé sur un alignement d'auto-apprentissage, procédé associé, appareil pour analyser un morphème d'expression destructrice par utilisation d'un corpus d'alignement et procédé d'analyse de morphème associé
WO2022114437A1 (fr) Système de tableau noir électronique pour réaliser une technologie de commande d'intelligence artificielle par reconnaissance vocale dans un environnement en nuage
WO2023068691A1 (fr) Procédé de traitement de langage naturel via la réalisation d'une analyse sémantique au moyen d'informations syntaxiques, et appareil pour celui-ci
CN108628911A (zh) 针对用户输入的表情预测
US11320914B1 (en) Computer interaction method, device, and program product
WO2017043715A1 (fr) Serveur, terminal d'utilisateur et procédé de commande de serveur et de terminal d'utilisateur
WO2014148664A1 (fr) Système de recherche en plusieurs langues, procédé de recherche en plusieurs langues et système de recherche d'image basé sur la signification d'un mot
WO2016036049A1 (fr) Programme informatique, procédé, système et appareil de fourniture de service de recherche
WO2018117660A1 (fr) Procédé de reconnaissance de parole à sécurité améliorée et dispositif associé
WO2020022645A1 (fr) Procédé et dispositif électronique pour configurer un clavier d'écran tactile
WO2017018736A1 (fr) Procédé destiné à la génération automatique d'index dynamique servant au contenu affiché sur un dispositif électronique
WO2024106925A1 (fr) Système et procédé de communication basés sur une réalité augmentée et dispositif informatique pour la mise en œuvre de celui-ci
WO2016204391A1 (fr) Dispositif, procédé et programme informatique fournissant un service de recherche
WO2014157746A1 (fr) Système de partage d'images permettant de rechercher des images par mot-clé et procédé s'y rapportant
WO2025187942A1 (fr) Résumé vidéo génératif personnalisé
WO2018008861A1 (fr) Système et procédé de commande html, utilisant un serveur utilisateur

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17863595

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 04.09.2019)

122 Ep: pct application non-entry in european phase

Ref document number: 17863595

Country of ref document: EP

Kind code of ref document: A1