KR102526391B1 - System for providing interpretation services based on distribution model - Google Patents

Abstract

A system for providing an interpretation service based on a distribution model is disclosed.
According to an embodiment of the present invention, a system for providing an interpretation service based on a distribution model on a communication network, recognizing original speech included in content collected and transmitted by a user terminal device, and converting the recognized original speech into the original language A first information conversion unit for converting into text; The original language text is allocated and distributed to each of a plurality of first proofreader terminal devices according to a distribution unit set based on the corresponding distribution model, and when the distributed original language text is proofread and collected in the plurality of first proofreader terminal devices, a first distribution correction unit that recombines; an AI translation unit for generating translated text by machine-translating the corrected and combined original language text into a language selected by the user terminal device or another user terminal device according to an artificial intelligence algorithm; The translated text is allocated and distributed to each of a plurality of second proofreader terminal devices according to a distribution unit set based on the corresponding distribution model, and when the distributed translated text is proofread and collected in the plurality of second proofreader terminal devices, a second distribution correction unit that recombines; and a second information conversion unit that converts the corrected and combined translation text into an interpretation voice and provides the result to the other user terminal device.

Description

A system for providing interpretation services based on a distribution model {SYSTEM FOR PROVIDING INTERPRETATION SERVICES BASED ON DISTRIBUTION MODEL}

The present invention relates to a system for providing an interpretation service based on a distribution model. More specifically, it relates to a technology capable of providing an interpretation service corrected quickly and accurately according to a distribution model when interpreting voice information in real time.

As the flow of globalization and the development of information and communication technology, more information is being actively exchanged around the world. In particular, in international business, it is important to secure the accuracy of interpretation in consideration of the importance of content as well as real-time interpretation for smooth communication between speakers using various languages. Therefore, in general, a simultaneous interpreter with a lot of experience in the relevant field directly participates in the conference and interprets on the spot.

However, it is difficult in terms of time and cost to have a simultaneous interpreter with accumulated experience in the field attend every meeting, and even in the case of a simultaneous interpreter with a lot of experience in the field, However, it has limitations in that omissions or errors due to factors such as reduced concentration according to field conditions and overloaded interpreting capacity cannot be completely excluded.

Therefore, recently, a service providing interpretation/translation on-line based on information and communication technology is attracting attention. However, most of these online interpretation/translation services tend to be of low quality due to automatic interpretation by AI (Artificial Intelligence) algorithms. As in, there are limitations in that it is difficult to overcome structural problems such as cost increase and limitations such as omissions and errors that occur during interpretation.

In particular, in the case of an international conference in which a large number of speakers participate and is conducted for a long time, other problems may be additionally accompanied in addition to the problems described above.

First, when interpreting multiple speakers through AI automatic interpretation online, it is quite difficult to process the vast amount of interpretation in real time by distinguishing speakers from each other and securing accuracy. In addition, even when a number of simultaneous interpreters are matched online and real-time interpretation is performed, it is difficult to expect constant quality of interpretation depending on the number of simultaneous interpreters being input, and it is difficult to adjust the overall amount of interpretation assigned to each simultaneous interpreter. Have.

Therefore, there is a demand for the development of a new type of interpretation service that combines automatic interpretation by artificial intelligence (AI) algorithms with interpretation by experts of a certain level.

Registered Patent Publication No. 10-2199050 (registration date: 2020.12.30.)

Therefore, the present invention is to solve the above problems of the prior art, and converts the information recognized in the content for which interpretation is requested (from voice to text, from text to voice) by artificial intelligence (AI) algorithm. It processes the automatic interpretation process, and at this time, the purpose is to allocate and distribute the text (original language text, translated text) to a number of corresponding proofreaders (original language proofreader, translation proofreader) according to the distribution unit set based on the distribution model.

In addition, another object of the present invention is to set and distribute a set of one or more sentence lumps divided on the basis of a period, which is a period, as a distribution unit in the text (original language text and the translated text).

In addition, the present invention is a method of distributing the text (original text, translated text) based on either the number of characters or words according to the distribution unit set based on the distribution model, the distributed text (original text, translation text) is redistributed based on the proofreading time, or redistributing the distributed text (original language text, translated text) based on the group selected by the proofreader (original language proofreader, translation proofreader), or It has another purpose to provide in a combined way.

In addition, the present invention documents (eg, WORD, PDF, PPT, etc.) in a form requested by a user (speaker, listener) based on the translated text corrected by the translation proofreader, and in the document, the original language Another object is to provide information including the speaker of the voice, the proofreader of the original language text, and the proofreader of the translated text.

In addition, the present invention matches a plurality of proofreaders (original language proofreaders, translation proofreaders) according to the interpretation conditions requested by users (speakers and listeners). Based on, there is another purpose of assigning priority according to the suitability of the interpretation condition.

The above object of the present invention is a system for providing an interpretation service based on a distribution model on a communication network, which recognizes original speech included in content collected and transmitted by a user terminal device, and converts the recognized original speech into original language text. A first information conversion unit for conversion; The original language text is allocated and distributed to each of a plurality of first proofreader terminal devices according to a distribution unit set based on the corresponding distribution model, and when the distributed original language text is proofread and collected in the plurality of first proofreader terminal devices, a first distribution correction unit that recombines; an AI translation unit for generating translated text by machine-translating the corrected and combined original language text into a language selected by the user terminal device or another user terminal device according to an artificial intelligence algorithm; The translated text is allocated and distributed to each of a plurality of second proofreader terminal devices according to a distribution unit set based on the corresponding distribution model, and when the distributed translated text is proofread and collected in the plurality of second proofreader terminal devices, a second distribution correction unit that recombines; and a second information conversion unit that converts the corrected and combined translated text into an interpretation voice and provides the converted text to the other user terminal device.

In addition, other configurations, systems, methods for implementing the present invention and computer readable recording media for recording computer programs for executing the methods are also achieved.

According to the present invention, the information recognized in the content for which interpretation is requested is converted (voice to text, text to voice) to process an automatic interpretation process by an artificial intelligence (AI) algorithm, and at this time, based on a distribution model According to the set distribution unit, the text (original language text, translated text) is allocated and distributed to a plurality of corresponding proofreaders (original language proofreaders, translation proofreaders), so that a prompt and accurate proofreading interpretation service can be provided.

In addition, according to the present invention, distribution efficiency is improved by setting and distributing a set of one or a plurality of sentence chunks divided on the basis of a period, which is a period, in the text (original language text and the translated text) as a distribution unit. It has the effect that speed and accuracy can be expected during calibration work.

In addition, according to the present invention, according to the distribution unit set based on the distribution model, the text (original text, translated text) is distributed based on either the number of characters or the number of words, the distributed text (original text , translation text) is redistributed based on the proofreading time input, or redistributed based on the group selected by the proofreader (source language proofreader, translation proofreader), or the distributed text (original language text, translated text) By distributing them in a combined manner, there is an effect of improving distribution efficiency including distribution (allocation) order and redistribution, and easy to secure calibration speed and accuracy.

In addition, according to the present invention, based on the combined translated text, a document is documented in a form requested by a user (speaker, listener), and in the document, the speaker of the original language voice, the proofreader of the original language text, and the translated text There is an effect that it can be provided by including the information of the proofreader.

In addition, according to the present invention, based on the translated text corrected by the translation proofreader, the document is documented in a form requested by the user (speaker, listener) (eg, WORD, PDF, PPT, etc.), and the document includes: Since the information of the speaker of the original language voice, the proofreader of the original language text, and the proofreader of the translated text is provided, it is possible to download the contents of the conference (stenographer function) in a format desired by the user, and the interpretation contents There is an effect of guaranteeing the accuracy of and clarifying responsibility.

In addition, according to the present invention, a plurality of proofreaders (original language proofreaders, translation proofreaders) are matched according to the interpretation conditions applied by users (speakers, listeners). By assigning priority according to the suitability of the interpretation conditions based on statistical information, there is an effect of being matched with more professional proofreaders with accumulated experience in the relevant field.

1 is a diagram showing the configuration of an entire system for providing an interpretation service based on a distribution model according to an embodiment of the present invention.
2 is a diagram showing in detail the internal configuration of a distributed interpretation system 300 according to an embodiment of the present invention.
3 is a mapping diagram for allocating and distributing text to a plurality of proofreader terminal devices according to a distribution unit set based on a distribution model according to an embodiment of the present invention.
4 is a calibration interface screen of any one terminal device among a plurality of calibrator terminal devices according to an embodiment of the present invention.
5 is a calibration interface screen for each proofreader terminal device showing a process of redistributing and redistributing distributed text according to an embodiment of the present invention.
6 is a flowchart of each component for providing an interpretation service based on a distribution model according to an embodiment of the present invention.
7 is a screen showing an example of a process of providing an interpretation service based on a distribution model according to an embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The detailed description of the present invention which follows refers to the accompanying drawings which illustrate, by way of illustration, specific embodiments in which the present invention may be practiced. It should be understood that the various embodiments of the present invention are different from each other but are not necessarily mutually exclusive. Combinations of each block of the accompanying configuration diagram (block diagram) and each step of the flowchart in the present invention can be performed by computer program instructions. Each such block or step may represent a module, segment or portion of code that contains one or more executable instructions for executing the specified logical function(s). It should also be noted that in some alternative embodiments it is possible for the functions recited in blocks or steps to occur out of order. For example, two blocks or steps that are shown adjacent or in sequence may in fact be performed substantially concurrently, or the blocks or steps may be performed in reverse order depending on their function.

It should also be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the detailed description described below is not intended to be taken in a limiting sense, and the scope of the present invention should be construed to cover all ranges equivalent to those claimed by the claims. In addition, similar reference numerals in the drawings indicate the same or similar functions in various aspects, and the form may be exaggerated for convenience.

Meanwhile, the terms used in the present invention are terms used to appropriately express preferred embodiments of the present invention, which may vary according to a user's intention or customs in the field to which the present invention belongs. Therefore, definitions of these terms should be understood based on the content throughout this specification. When it is said that a certain part "includes" a certain element in the entire specification, this means that it may further include other elements rather than excluding other elements unless otherwise stated. In addition, terms such as “unit” and “module” described in the specification refer to a unit that processes at least one function or operation, and may be implemented as hardware or software or a combination of hardware and software.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily practice the present invention.

[Embodiment of the present invention]

In the present invention, “interpretation service” refers to a service that translates the speaker's original language voice between a speaker and a listener and provides the listener with an interpretation voice in another language, and converts information from voice to text and from text to voice. It may be a concept encompassing not only the function of translating text (original language) to text (other language), but also the function of translation service. Although the speaker's original voice in the present invention may be provided as an interpreter voice to the listener alone, the present invention is not limited thereto, and other information (eg, video, image, text) in addition to the speaker's original voice is not limited thereto. is provided as a single content that includes, it is possible to provide various services such as an online video conference interpretation service.

Composition of the entire system

1 is a diagram showing the configuration of an entire system for providing an interpretation service based on a distribution model according to an embodiment of the present invention.

Referring to FIG. 1, the entire system according to an embodiment of the present invention includes a communication network 100, an operating server 200, a distributed interpretation system 300, a first user terminal device 400, and a first corrector terminal device. 500, a second calibrator terminal device 600, and a second user terminal device 700.

First, the communication network 100 according to an embodiment of the present invention is a network capable of performing a series of data transmission and reception operations for data transmission and information exchange, regardless of its communication mode, such as wired and / or wireless can be configured.

More specifically, a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), an Integrated Services Digital Network (ISDN), and a wireless LAN (wireless LAN). It may be any one of a LAN) or a mobile telecommunication network, or various types of wired/wireless communication networks connected and configured by a combination thereof. Preferably, the communication network 100 configured according to the present invention may be an Internet network capable of using World Wide Web (WWW) services.

Next, the operating server 200 according to an embodiment of the present invention may be an operating server capable of performing operating tasks related to interpretation services such as access (login) management, content management, and payment processing.

More specifically, the operation server 200 may perform the function of a web/app operation server of an interpretation service operating company. When a user (speaker, listener) subscribes to the corresponding web/app and logs in to apply for an interpretation service, We can provide you with related services. At this time, the operation server 200 includes the distributed interpretation system 300, the first user terminal device 400, the first proofreader terminal device 500, the second proofreader terminal device 600, and/or the second proofreader terminal device 600, which will be described later. A dedicated web and/or app program that connects the user terminal device 700 through the communication network 100 may be operated or a function supporting the same may be performed. In addition, the operation server 200 allows the distributed interpretation system 300 and other devices 400 to 700 to run an application for implementing a network according to the present invention, on a web/app, API (Application Programming Interface). ) can also perform functions that support the necessary interfaces to build applications such as

Next, the distributed interpretation system 300 according to an embodiment of the present invention converts the information included in the content for which interpretation is requested (from voice to text, from text to voice) and converts it to an artificial intelligence (AI) algorithm. At this time, the text (original language text, translated text) is distributed to a number of corresponding proofreaders (original language proofreaders, translation proofreaders) according to the distribution unit set based on the distribution model to quickly and accurately It may be a system capable of providing interpretation services.

More specifically, the distributed interpretation system 300 recognizes voice information (original language voice) included in the content collected and transmitted by the first user terminal device 400, and converts the recognized original language voice into text (original language text). , and allocates and distributes the original language text to each of the plurality of first proofreader terminal devices 500 according to the distribution unit set based on the corresponding distribution model, and the distributed original language text is distributed to the plurality of first proofreader terminal devices Once corrected and collected in 500, they can be combined again.

The text (translated text) is obtained by machine-translating the corrected and combined original language text into a language selected by the first user terminal device 400 or another user terminal [second user terminal device 700] according to an artificial intelligence algorithm. can create In addition, the translated text is allocated and distributed to each of the plurality of second proofreader terminal devices 600 according to the distribution unit set based on the corresponding distribution model, and the distributed translated text is transmitted to the plurality of second proofreader terminal devices 600. Once corrected and collected, it can be recombined.

Interpretation service can be provided by recognizing the corrected and combined translated text, converting it into an interpretation voice, and providing it to the terminal devices 400 and 700 including the other user terminal [the second user terminal device 700]. , The distributed translation system 300 according to an embodiment of the present invention may be more specifically understood by the following detailed description with reference to FIG. 2 .

Next, the first user terminal device 400 according to an embodiment of the present invention collects content including voice information (original speech) of a user (speaker) using the terminal device, ) may be a digital device capable of transmitting to

More specifically, the first user terminal device 400 collects content including voice information (original voice) of a user (speaker) using the first user terminal device 400, and operates the server 200 and/or Alternatively, it can be provided (transmitted) to the distributed interpretation system (300). Conversely, a function of receiving (receiving) content including interpreted voice information (interpretation voice) of another user terminal device (eg, the second user terminal device 700) provided from the interpretation system 300 is provided. can also be done

In such a first user terminal device 400, a dedicated web or app program capable of transmitting content including the original voice or receiving content including the interpretation voice may be further installed. . In particular, a microphone function may be included to collect the voice of a speaker, and a speaker function may be further included to listen to the interpreted voice of another speaker. These functions (means) are the first user terminal device. (400) It is not necessarily provided inside, and may be implemented by a separate device electrically connected by a wired/wireless communication module (not shown) capable of transmitting and receiving information using wired or short-range wireless communication technology. there is.

Next, the first proofreader terminal device 500 according to an embodiment of the present invention is composed of a plurality of texts (original language text) distributed according to a distribution unit set based on a distribution model in the distribution type interpretation system 300 ) and provide (transmit) to the distributed interpretation system 300 again.

More specifically, the plurality of first proofreader terminal devices 500 convert the original language text converted in the distributed interpretation system 300 to each of the plurality of first proofreader terminal devices 500 according to a distribution unit set based on a distribution model. Once allocated and distributed, a process of proofreading the distributed sentences of the original language text by an original language proofreader using the language of the original language text may be performed. In addition, the original language text corrected by the plurality of first proofreader terminal devices 500 may be provided (transmitted) to the distributed interpretation system 300 and combined.

Next, the second proofreader terminal device 600 according to an embodiment of the present invention is composed of a plurality of texts (translated text) distributed according to a distribution unit set based on a distribution model in the distribution type interpretation system 300. ) and provide (transmit) to the distributed interpretation system 300 again.

More specifically, in the plurality of second proofreader terminal devices 600, the text (translated text) automatically translated by the AI algorithm in the distributed interpretation system 300 is divided into a plurality of second proofreaders according to a distribution unit set based on a distribution model. Once allocated and distributed to each proofreader terminal device 600, a process of proofreading the distributed sentences of the translated text can be performed by the translation proofreader using the language of the translated text. In addition, the translated text corrected by the plurality of second proofreader terminal devices 600 may be provided (transmitted) to the distributed interpretation system 300 and combined.

At this time, the plurality of first corrector terminal devices 500 and the plurality of second corrector terminal devices 600 are the first corrector terminal devices 500 and the plurality of second corrector terminal devices ( 600) may perform a redistribution process based on one or more groups of the distribution units selected by the original language proofreader and the translation proofreader, which can be clearly understood by the following detailed description with reference to FIGS. 4 and 5 there is.

Finally, the second user terminal device 700 according to an embodiment of the present invention receives the voice information (interpretation voice) provided by the distributed interpretation system 300 so that the user (listener) can listen to it. It may be a digital device capable of performing the functions provided.

More specifically, the second user terminal device 700 provides an interpretation service by providing a user (listener) using the second user terminal device 700 with content including the interpretation voice interpreted by the distributed interpretation system 300. can provide. Conversely, the second user terminal device 700 collects content including voice information (original speech) of a user (speaker) using the terminal device, and operates the operation server 200 and/or the distributed interpretation system 300. ) may perform a function of providing (transmitting).

In such a second user terminal device 700, a dedicated web or app program may be further installed to receive content including interpretation voice or to transmit content including the original voice. In particular, a speaker function may be included to listen to the interpretation voice of another speaker, and a microphone function may be included to collect the voice of the speaker. These functions are provided inside the second user terminal device 700. It is not necessarily provided, and may be implemented by a separate device electrically connected by a wired/wireless communication module (not shown) capable of transmitting and receiving information using a wired or short-distance wireless communication technology.

Meanwhile, the first user terminal device 400, the first corrector terminal device 500, the second corrector terminal device 600, and/or the second user terminal device 700 according to the embodiment of the present invention described above is a digital device that has a function of transmitting and receiving the above-described interpretation service by accessing it through the communication network 100, and all known mobile communication terminal devices, information communication devices, multimedia terminal devices, and wired terminal devices include such digital devices. , various terminal devices such as fixed type terminal devices and IP (Internet Protocol) terminal devices may be applied. Specifically, terminal devices include a smart phone, a portable multimedia player (PMP), a personal digital assistant (PDA), a mobile internet device (MID), a telematics terminal, a desktop, and a tablet. It is a digital device equipped with memory means and equipped with a microprocessor, such as a computer (Tablet PC), notebook (Note book), net book, etc. It may be adopted as a terminal device according to the invention.

In addition, the first user terminal device 400, the first proofreader terminal device 500, the second proofreader terminal device 600, and/or the second user terminal device 700 include an interpretation service according to the present invention and information related thereto. It may be provided with a display means capable of displaying. Here, the display means is LCD (Liquid Crystal Display), TFT-LCD (Thin Film Transistor LCD), OLED (Organic Light Emitting Diodes), Light Emitting Diode (LED), AMOLED (Active Matrix Organic LED), Flexible Display and a 3D display. At this time, since the display means may include a touch screen, some or all of the functions of the input means may be performed.

In particular, in the first user terminal device 400 and the second user terminal device 700 in the present invention, a microphone device capable of collecting voice information (original voice) of a speaker and a speaker device capable of delivering an interpretation voice to a listener (including a headset and earphones) may be further included or interlocked therewith to improve the convenience of users (speakers and listeners) when providing an interpretation service.

Meanwhile, in FIG. 1 showing an embodiment of the present invention, the distributed interpretation system 300 and the operating server 200 are shown as being configured separately, but according to the needs of those skilled in the art for implementing the present invention, distribution It will be apparent that the type interpretation system 300 can be configured by integrating the operation server 200.

Configuration of Distributed Interpretation System (300)

In the following detailed description, the internal configuration of the distributed interpretation system 300 that performs important functions for the implementation of the present invention and the functions of each component will be reviewed.

2 is a diagram showing in detail the internal configuration of a distributed interpretation system 300 according to an embodiment of the present invention.

Referring to FIG. 2, the distributed interpretation system 300 according to an embodiment of the present invention includes a first information conversion unit 310, a first distribution correction unit 320, an AI translation unit 330, and a second distribution unit 310. It may include a correction unit 340, a second information conversion unit 350, a database 360, a communication unit 370 and a control unit 380.

According to such an embodiment of the present invention, the first information conversion unit 310, the first distribution correction unit 320, the AI translation unit 330, the second distribution correction unit 340, the second information conversion unit ( 350), the database 360, the communication unit 370, and the control unit 380, at least some of which are the operation server 200, the first user terminal device 400, the first corrector terminal device 500, and the second corrector It may be program modules that transmit and receive data with the terminal device 600 and/or the second user terminal device 700 . These program modules may be included in the distributed interpretation system 300 in the form of an operating system, application program module, and other program modules, and may be physically stored on various well-known storage devices. Hardware components ( For example, it may be implemented as a general-purpose processor, a dedicated processor) and/or software components (eg, firmware, application, program module) and combinations thereof. Also, these program modules may be stored in a remote storage device capable of communicating with the distributed interpretation system 300. Meanwhile, these program modules include routines, subroutines, programs, objects, components, data structures, etc. that perform specific tasks or execute specific abstract data types according to the present invention, but the present invention is not limited thereto. .

First, the first information conversion unit 310 according to an embodiment of the present invention recognizes the original speech included in the content collected and transmitted by the first user terminal device 400, and converts the recognized original speech into the original language. You can perform the function of converting to text.

More specifically, when the first user terminal device 400 collects (obtains) and provides (transmits) content including the original voice of a user (speaker) using the terminal device, the original voice is extracted and recognized from the content. After that, a process of converting into original language text may be performed, and the known speech-to-text technology (STT: Speech-to-Text) may be applied without limitation. Looking at the concept of this SST technology as an example, an input voice is compared with an acoustic model (Acoustic Model, AM) and a language model (Language Model, LM), and finally the voice signal is converted into character symbols based on the value. can

For example, a representative algorithm is the Hidden Markov Model (HMM), which models a basic unit (phoneme) for speech recognition. There are a pattern matching method for finding a voice pattern closest to an input voice, and an identification function method for determining a function by preparing a function for distinguishing one word from another in advance and applying the function to the input voice. In addition to the above methods, various known methods for voice recognition and voice-to-text conversion may be used, and the present invention is not limited to the above methods.

Next, the first distribution proofreading unit 320 according to an embodiment of the present invention allocates the original language text to each of a plurality of first proofreader terminal devices 500 according to a distribution unit set based on a corresponding distribution model, When the distributed original language text is corrected and collected in the plurality of first proofreader terminal devices 500, a function of recombining them may be performed.

More specifically, the first distribution correction unit 320 improves distribution efficiency by setting and distributing a set of one or a plurality of sentence chunks divided based on periods, which are periods in the original language text, as the distribution unit. can make it

In addition, when the distributed original language text is corrected in the plurality of first proofreader terminal devices 500, the first distribution and correction unit 320 collects and combines them to complete (generate) one corrected original language text. can

Next, the AI translation unit 330 according to an embodiment of the present invention converts the corrected and combined original language text into a language selected by the first or second user terminal device 400 or 700 according to an artificial intelligence algorithm. It can perform the function of generating translated text by translating.

More specifically, the AI translation unit 330 processes the translation of natural language used by humans through a computer equipped with an artificial intelligence algorithm, and converts text in a specific language into text in another language (automatically). It can perform translation function. For example, Rule-Based Machine Translation (RBMT) using word dictionaries and translation rules, Statistical Machine Translation (SMT) using large amounts of corpus data on original texts and their translations, and original texts. Known translation technologies such as Neural Machine Translation (NMT), which learns the relationship between the text and the corresponding translation using an artificial neural network, may be adopted and used without limitation.

Next, the second distribution correction unit 340 according to an embodiment of the present invention allocates the translated text to each of a plurality of second proofreader terminal devices 600 according to a distribution unit set based on a corresponding distribution model. When the distributed translated text is proofread and collected in the plurality of second proofreader terminal devices 6000, a function of recombining them can be performed.

More specifically, the second distribution correction unit 340 improves distribution efficiency by setting and distributing a set of one or more sentence chunks divided on the basis of a period as a period in the translated text as the distribution unit. can make it

In addition, when the distributed translation text is corrected in the plurality of second proofreader terminal devices 600, the second distribution and correction unit 340 collects and combines them to complete (generate) one corrected translation text. can

On the other hand, the text (original text and translated text) described above is transmitted to a proofreader through a plurality of first proofreader terminal devices 500 and second proofreader terminal devices 600 according to a distribution unit set based on a corresponding distribution model. The distribution method for each (original language proofreader, translation proofreader) can be divided into three major categories as follows.

First, a method of allocating and distributing to each of the plurality of first and second proofreader terminal devices 500 and 600 according to the distribution unit based on any one of the number of characters or words of the original text and the translated text; Second, based on the proofreading time required for proofreading and collecting the distributed original language text and the distributed translated text, allocated to each of the plurality of first and second proofreader terminal devices 500 and 600 according to the distribution unit Third, one or more distribution units selected by the plurality of first proofreader terminal devices 500 or the plurality of second proofreader terminal devices 600 from among the distributed original language text and the distributed translated text. It may be any one of the redistribution based on the group of , or a method consisting of a combination thereof, and more specific details can be understood by the following detailed description with reference to FIG. 3 .

Next, the second information conversion unit 350 according to an embodiment of the present invention recognizes the corrected and combined translated text provided by the second distribution correction unit 340 and converts (synthesizes) it into an interpretation voice. , It is possible to perform functions provided to terminal devices 400 and 700 including other user terminal devices 700 .

More specifically, the second information conversion unit 350 may perform a process of recognizing the corrected and combined translated text provided by the second distribution correction unit 340 and then converting it into an interpretation voice. Text-to-speech synthesis (TTS: Text-To-Speech) technology can be applied and processed without limitation. Looking at the concept of this TTS technology as an example, it is a text-to-speech technology that converts text (characters) into voice. Voice is divided into certain units, and when text is input, the corresponding pieces can be found in the database and combined to synthesize. .

For example, typical algorithms include concatenative TTS, which cuts and stores voice into very short units such as phonemes, and synthesizes voice by combining phonemes constituting a sentence to be synthesized; There is parametric TTS, which is extracted using (parameters) and then modeled using statistical techniques, and when text is input, it is synthesized from the model into voices of parameters.

In addition to the above methods, various known methods for speech synthesis and text-to-speech conversion may be used, and the present invention is not limited to the above methods.

Next, the database 360 according to an embodiment of the present invention includes information necessary or related to the interpretation service, for example, user information, natural language information, sound model information, language model information, speech synthesis information, words/sentences It may be a storage device capable of storing information, distribution model information, and the like.

On the other hand, although the database 360 is shown as being included in the distributed interpretation system 300 in FIG. 2, according to the needs of those skilled in the art to implement the present invention, the database 360 ) and may be configured separately or included in the operation server 200. The database 360 is a concept including a computer-readable recording medium, and refers to a database in a broad sense including not only a database in a narrow sense but also a data record based on a file system, and even a simple set of logs. If data can be extracted by searching, it can be included in the database 360 according to the present invention.

Next, the communication unit 370 according to an embodiment of the present invention includes the internal components of the distributed interpretation system 300 including the operation server 200, the distributed interpretation system 300, and the first user terminal device 400. , It can perform a function to transmit and receive data (information) with external devices such as the first corrector terminal device 500, the second corrector terminal device 600, and/or the second user terminal device 700. .

Finally, the control unit 380 according to an embodiment of the present invention includes the first information conversion unit 310, the first distribution correction unit 320, the AI translation unit 330, and the second distribution correction unit 340. , It can perform a function of controlling the flow of data between the second information conversion unit 350, the database 360 and the communication unit 370. That is, the control unit 380 according to the present invention controls the flow of data from the outside or between each component of the distributed interpretation system 300, so that the first information conversion unit 310, the first distribution correction unit ( 320), the AI translation unit 330, the second distribution correction unit 340, the second information conversion unit 350, the database 360, and the communication unit 370 may be controlled to perform unique functions.

Meanwhile, although not shown in FIG. 3, according to the interpretation conditions requested by the first user terminal device 400 or the second user terminal device 700, a plurality of the first and second proofreader terminal devices 500 and 600 It may further include a matching unit for matching. The matching unit may be preferentially assigned according to the suitability of the interpretation condition based on the calibration statistical information of the plurality of first and second proofreader terminal devices 500 and 600 .

In addition, although not shown in FIG. 3, a content composition unit may further include a content synthesis unit that combines the corrected and combined translated text and the interpretation voice with the content collected from the first user terminal device 400 and provided. . The content synthesis unit may provide information including the speaker of the original speech, the proofreader of the original language text, and the proofreader of the translated text.

In the distributed interpretation system 300 according to an embodiment of the present invention, when a user inputs a specific language voice (original language voice), the corresponding original language voice is recognized based on deep learning and big data, and the text conversion process is performed. and the converted text information (original language text) is translated into text in another language (translated text) in the cloud through natural language processing that computers can understand, and synthesized with voice (interpretation voice) to provide interpretation services to users. can provide That is, the distributed interpretation system 300 of the present invention can be implemented as a platform technology that encompasses deep learning, big data, and cloud.

Example of distribution model

In the following detailed description, a distribution model, which performs an important function to provide prompt and accurate interpretation service in the distribution type interpretation system 300 according to the present invention, will be described through an example. For convenience of description, the first proofreader terminal device 500 is equivalently expressed as an original language proofreader and the second proofreader terminal device 600 as a translation proofreader.

The text (original language text and translated text) is transmitted through a plurality of first proofreader terminal devices 500 and second proofreader terminal devices 600 according to a distribution unit set based on the corresponding distribution model. The distribution method can be divided into three main types as follows.

First, a method of allocating and distributing to each of the plurality of first and second proofreader terminal devices 500 and 600 according to the distribution unit based on any one of the number of characters or words of the original text and the translated text; Second, based on the proofreading time required for proofreading and collecting the distributed original language text and the distributed translated text, allocated to each of the plurality of first and second proofreader terminal devices 500 and 600 according to the distribution unit Third, one or more groups selected by the plurality of first proofreader terminal devices 500 or the plurality of second proofreader terminal devices 600 from among the distributed original language text and the distributed translated text (grouping ), it may be any one of the redistribution schemes according to the distribution unit or a scheme composed of a combination thereof.

More specifically, first, looking at the method of distributing based on either the number of characters or the number of words according to the present invention, the number of characters or words of text (original text, translated text) is based on n (n is a natural number). , according to the distribution unit composed of sentence chunks included in all of the n sentences, it can be allocated and distributed to the first and second proofreader terminal devices (original language proofreader, interpretation proofreader). Conversely, according to the distribution unit composed of sentence chunks including at least a part of the number of n characters or words, it can be allocated and distributed to the first and second proofreader terminal devices (original language proofreader, translation proofreader).

For example, if the number of words is 10, after classifying up to 10 words in the text (original text, translated text), only a chunk (distribution unit) of sentences that are all included within 10 words can be distributed. In the text (original text: Korean), "Hello. I have a topic to present at this conference. This topic is about the development direction of big data." hello. I had the opportunity to make a presentation on the topic at this conference. This topic is 'until'. Therefore, among the chunks (distribution units) of sentences based on periods, which are periods, 'this subject' is not a chunk (distribution unit) of sentences that include all of the above 10 words, so 'Hello. I have a topic presentation at this meeting.' Only a chunk (distribution unit) of two sentences can be distributed to the first proofreader among multiple proofreaders (original language proofreaders). In the end, 'this topic is about the development direction of big data' can be distributed to other proofreaders other than the first proofreader among multiple proofreaders (original language proofreaders).

In the case of a method of distributing based on n number of characters or number of words (n is a natural number) of such text (original text, translated text), the n standard values are set in the distributed interpretation system 300 or the user (speaker) , listener), but the present invention is not limited thereto and may be variably changed according to the calibration time described later. That is, if the proofreading speed of the text (original language text, translated text) increases, the n reference values may increase, and if the proofreading speed slows down, the n reference values may decrease.

Next, looking at the case of distribution based on the proofreading time according to the present invention, the texts (original language text, translated text) distributed to proofreaders (original language proofreaders, interpretation proofreaders) are sent to the distributed interpretation system 300 after proofreading is completed. It may be a method of distributing based on the time required to input (collect) (calibration time).

More specifically, the proofreading time may mean the time required from the time the proofreader receives the distributed text from the distributed interpretation system 300 to the time when the proofreading is completed and inputted into the distributed interpretation system 300. there is.

For example, if the correction is completed within the minimum set time, in the next distribution order, twice the amount of the correction-completed sentences in the distribution unit may be set as a distribution unit and distributed. In addition, when the correction is completed because the correction time exceeds the maximum set time, in the next distribution order, only 1/2 times of the correction-completed distribution unit sentence chunk may be set and distributed as a distribution unit. In addition, when the correction time exceeds the minimum set time and the correction is completed within the maximum set time, the amount of the corrected distribution unit lump sentences may be maintained and distributed in the next distribution order.

As an example, the minimum setting time may mean a time (for example, within 5 seconds) in which real-time interpretation service can be smoothly performed online, and the maximum setting time is a difficult limit in providing real-time interpretation service online. It may mean time (eg, greater than 10 seconds). On the other hand, in this method of distributing based on proofreading time, the order of allocating texts (original language text, translated text) distributed to each proofreader (original language proofreader, translation proofreader) may be determined. This allocation sequence can be understood by the following detailed description with reference to FIG. 3 .

3 is a mapping diagram for allocating and distributing text to a plurality of proofreader terminal devices according to a distribution unit set based on a distribution model according to an embodiment of the present invention.

Texts (original text, translated text) can be assigned to a plurality of proofreaders in order according to the criteria of the set distribution model. takes precedence, and if the distribution unit held among the plurality of first and second corrector terminal devices 500 and 600 exceeds the set range, the corresponding terminal device may omit the assignment order in the next distribution order. In addition, among the plurality of first and second calibrator terminal devices 500 and 600, redundant distribution may be performed to a terminal device in which the calibration of the previous distribution order is completed in the shortest time.

Referring to FIG. 3 , as an example, a case in which translated text is allocated based on n number of words described above and distributed to a plurality of translation proofreaders (A, B, C, D proofreader terminals) will be described.

First, the translation text (1, 2, 3, 4) can be distributed to each of the plurality of translation proofreaders (terminals of A, B, C, and D proofreaders). 'A proofreader terminal is translated text 1, proofreader terminal B is translated text 2, proofreader terminal C is translated text 3, and proofreader terminal D is translated text 4'. This time point can be set as the starting point of the unit time. Subsequently, among the translated texts (1, 2, 3, 4) distributed in the first distribution process, proofreading is completed in the order of 3, 2, and 4 (shaded circle in FIG. ) can be entered. On the other hand, the translation text 1 distributed to the proofreader A terminal has not been proofread.

Subsequently, in the second distribution process (based on unit time 1, 5 seconds), allocation proceeds according to the order in which proofreading is completed, and translation texts 8, 6, 5, and 7 may be allocated in order for each of a plurality of translation proofreaders. That is, the translation text 8 may be assigned to the A proofreader terminal in which proofreading has not yet been completed in the latest order.

Subsequently, among the translated texts (8, 6, 5, and 7) distributed in the secondary distribution process, proofreading is completed in the order of 6, 5, and 7 (shaded circle in FIG. ) can be entered.

Subsequently, in the third distribution process (based on unit time 2, 10 seconds), allocation proceeds according to the order in which proofreading is completed, and each translation proofreader may be allocated in order of translation text 9+12, 10, and 11. That is, since the proofreader A terminal has not yet completed proofreading of the translation texts 1 and 8, further allocation is suspended (stopped), and the translation text 8 to be distributed to the proofreader A terminal is delivered (moved) to the proofreader B terminal. may be redundantly distributed. At this time, the retention limit (Limit) of the translated text may be set to, for example, the case of retaining two.

Next, looking at the case of redistribution based on one or more groups of the distribution units selected by the proofreader (original language proofreader, interpretation proofreader) according to the present invention, the text (original text , translated text), distributed texts (source language text, translated text) may be grouped and redistributed according to the selection of a proofreader (original language proofreader, translation proofreader).

4 is a calibration interface screen of any one terminal device among a plurality of calibrator terminal devices according to an embodiment of the present invention.

Referring to FIG. 4, a method of redistributing based on one or a plurality of groups of distribution units selected by the plurality of first and second correctors terminal devices 500 and 600 is the plurality of first and second correctors It can be provided on the calibration interface 10 configured differently for each terminal device (500, 600). This proofreading interface 10 may include and consist of a plurality of distributed text blocks 20 (21, 22, 23), proofreading blocks 30 and selection blocks 40.

First, the plurality of distribution text blocks 20: 21, 22, 23 may be sequentially combined among the distributed original language text or the distributed translation text, and the plurality of distribution text blocks 20: 21, 22, 23 is the distribution text block 22 in which the corresponding distribution units assigned and distributed to each of the first and second corrector terminal devices 500 and 600 are arranged in the center, and the distribution text in which the corresponding distribution units sequentially combined are arranged Blocks 21 and 23 can be arranged in order around them.

Next, the correction block 30 may correct and input any one or more of the plurality of distribution text blocks 20 (21, 22, 23). At this time, the distribution text block 22 arranged in the center may be in an active state.

Subsequently, the selection block 40 may group the distribution unit by selecting a correction target (activation) from among the plurality of distribution text blocks 20 (21, 22, 23).

More specifically, as an example, the second distribution text block (translation text 4) distributed on the proofreading interface screen 10 displayed on the terminal device 600 of the second proofreader will be described based on the translated text distributed to the translation proofreader. : 22) is shown arranged in the center. At this time, the first distribution text block (translation text 3: 21) distributed to other translation proofreaders is arranged at the top of the second distribution text block (translation text 4: 22), and the second distribution text block (translation text 4 : 22), a third distribution text block (translation text 5: 23) distributed to another translation proofreader may be arranged at the lower end.

The first and third distribution text blocks (translation text 3, 5: 21, 23) distributed to other translation proofreaders at the top and bottom are inactivated (shaded) and provided, and the second distribution text block distributed to the corresponding translation proofreader. (Translation Text 4: 22) can only be activated and served.

Subsequently, the correction block 30 located at the bottom may be an area for inputting the result after correcting the activated second distribution text block (translation text 4: 22) with the corresponding translation corrector. In this correction block 30, the activated distribution text block is basically reflected (copy input) so that the correction process can proceed. At this time, when the calibration process is completed, the completion button 42 of the calibration completion selection block 40 can be selected and transmitted to the distributed interpretation system 300.

Subsequently, selection buttons 41a and 41b are arranged in the right selection block 40 of the first and third distribution text blocks (translation text 3, 5: 21, 23) distributed to other translation proofreaders. When the buttons 41a and 41b are selected (clicked), the shading of the corresponding first and third distribution text blocks (translation text 3, 5: 21 and 23) may be removed and changed to an active state. For example, when the selection button 41a located on the right side of the first distribution text block (translation text 3: 21) is selected (clicked), the block is activated, and 'translation text 3: 21 + translation text 4' is activated. : 22', redistribution is set, and the distributed translation texts 1 and 4 activated in the proofreading block 30 located at the bottom are basically reflected (copy input) so that the proofreading process can proceed.

In the same principle, when the selection button 41b located on the right side of the third distribution text block (translation text 5: 23) is selected (clicked), the corresponding block is activated, and 'translation text 4: 22 + translation text 5: 23' can be grouped and set. On the other hand, when all of the selection buttons 41a and 41b are selected, the first, second, and third distribution text blocks 21, 22, and 23 are all activated and 'translation text 3: 21 + translation text 4: 22 + translation text 5' : 23' can be grouped and redistributed.

The process of grouping and redistributing distributed texts according to the selection of the translation corrector can be more clearly understood by the following detailed description with reference to FIG. 5 .

5 is a calibration interface screen for each proofreader terminal device showing a process of redistributing and redistributing distributed text according to an embodiment of the present invention.

Referring to FIG. 5, as an example, an interface screen displayed for each terminal device 600 of a translation proofreader is shown. For convenience of description, the proofreading interface screens (A, B, and C) of the terminal device 600 of a total of three translation proofreaders are as follows.

First, on the screen A of the proofreader A, the distributed translation text (translation text 4) is arranged in the second distributed text block and is displayed in an activated state, arranged in the center. Then, on the proofreader B terminal screen (B), the distributed translation text (translation text 5) is arranged in the second distributed text block and arranged in the center in an activated state, and on the proofreader C terminal screen (C), the distributed translation text ( The translated text 6) can be placed in the second distributed text block and centered in an activated state. At this time, at the upper and lower ends of the translated text (second distribution text block) arranged in the center, the translated text distributed to each terminal device 600 of the translation proofreader that is located at the front and rear ends in order of the translated text and is sequentially combined can be arranged

Therefore, as a whole, the distributed translation text (translation text 3, 4, 5) on the terminal screen of proofreader A (translation text 3, 4, 5), the distributed translation text (translation text 4, 5, 6) on the terminal screen of proofreader B, and proofreader C On the terminal screen C, the distributed translation texts (translation texts 5, 6, and 7) are sequentially arranged for each distributed text block, and the proofreading interface screens (A, B, C) of each translation proofreader terminal device 600 are arranged in order. ) can be displayed on.

At this time, each translation proofreader is basically located in the center of the proofreading interface screen (A, B, C) as described with reference to FIG. can However, each translation proofreader examines the overall context of the distributed translation text and, in order to proofread together with the translated text distributed to other translation proofreaders positioned at the front and back in order, deactivated translation text (distribution text block) You can select the selection box located on the right, group, redistribute, and proceed with the correction process. Hereinafter, a few cases are presented and explained to aid understanding.

First, when proofreader A (A) completes proofreading of 'translation text 4' and proofreader B (corrector B) groups and proofreads 'translation text 4 + translation text 5', proofreader A on the distributed interpretation system 300 'Translation text 4' of (A) is discarded (if 'translation text 4 is under proofreading, stop processing'), and proofreaders B's 'translation text 4 + translation text 5' grouped priority can be applied This is because grouped proofreading shows the probability of providing a more accurate proofreading (translation) quality depending on the overall context. Subsequently, proofreader C may complete the proofreading of 'translated text 6'.

As a result, the proofreading result of 'translated text 4 + translated text 5' by proofreader B and the proofreading result of 'translated text 6' by proofreader C are collected by the distributed interpretation system 300, and 'translated text 4' + translation text 5 + translation text 6'.

Next, in the above case, when proofreader C (C) groups 'translated text 5 + translated text 6' and proofreads, proofreader B (B) compares 'translated text 4 + translated text 5' Translation text 5' overlaps. In this case, proofreader C (C) grouped later in time may take precedence over proofreader B (B). Therefore, the proofreading result of 'translation text 4 + translation text 5' by proofreader B is discarded, and proofreader C (C) can correct by grouping all 'translation text 4 + translation text 5 + translation text 6'. . If proofreader B (B) is still proofreading 'translated text 4 + translated text 5', the processing may be stopped, and another distributed translated text may be assigned instead.

Next, when both B proofreader (B) and C proofreader (C) group 'translation text 5 + translation text 6' and proofread, the B proofreader (B ) may take precedence. This is because the proofreading (translation) of the translated text located at the back end (arranged) around the translated text distributed to the proofreader shows the probability of providing a more accurate proofreading (translation) quality according to the overall context.

Next, if proofreader B (B) grouped 'translated text 4 + translated text 5' and proofreader C (C) grouped 'translated text 5 + translated text 6 + translated text 7' to proofread, the front end The C corrector (C) grouped with the back end may take precedence over the B corrector (B) grouped with . Therefore, the proofreading result of 'translation text 4 + translation text 5' by proofreader B (B) is discarded (if it is still being proofreaded, it is stopped), and proofreader C (C) returns 'translation text 5 + translation text 6 + translation Text 7' can be corrected by grouping all of them.

Therefore, in summary of the above-mentioned principle, the calibration result of the terminal device that selects the grouped distributed text block rather than the single distributed text block selected from among the plurality of first and second corrector terminal devices 500 and 600 can be prioritized. .

In addition, when all of the grouped distributed text blocks selected from among the plurality of first and second corrector terminal devices 500 and 600 overlap, the calibration result of the terminal device selected later in time may take precedence.

In addition, when some of the grouped distribution text blocks selected from among the plurality of first and second corrector terminal devices 500 and 600 overlap, all grouped terminal devices that select the later distribution text blocks that are sequentially combined The proofreading result of the distributed text block may be prioritized.

In order to help understand the more specific configuration of the distributed interpretation system 300 according to the present invention described above, in the following detailed description, a user connects to the system and applies for an interpretation service, and accordingly, based on the distribution model A process (method) of providing an interpretation service will be described with an example.

Process for providing interpretation services based on distribution model

In the following detailed description, the process of providing an interpretation service based on the distribution model according to the present invention includes a distributed interpretation system 300, a first user terminal device 400, a first corrector terminal device 500, An example will be described focusing on the data processing process between the second calibrator terminal device 600 and the second user terminal device 700, but this is the most representative process for convenience of explanation, and the present invention is limited thereto. It is not.

6 is a flowchart of each component for providing an interpretation service based on a distribution model according to an embodiment of the present invention.

Referring to FIG. 6 , first, a process (S10) of collecting (obtaining) content including audio information (original speech) of a speaker (user) using the first user terminal device 400 is performed, and the collected audio information is performed. A process of requesting interpretation for content including (original speech) may be performed (S20). In this case, the subject requesting interpretation may be either the first or second user terminal device 400 or 700 . Meanwhile, although not shown, a normal network access process in which the user terminal devices 400 and 700 log in to the distributed interpretation system 300 may be preceded.

Subsequently, the distributed interpretation system 300 performs a speech-to-text (STT) process of extracting and recognizing only the original speech included in the content provided from the first user terminal device 400 and converting it into text information. It can (S30). For example, if the speaker uses Korean and the original speech is Korean, the Korean speech can be recognized through the STT process (S30) and then converted into Korean (Hangul) text. In this STT process (S20), a known speech-to-text conversion technique can be used without limitation, but as an example, a method of modeling a basic unit (phoneme) for speech recognition by applying a hidden Markov model (HMM), They can be combined to form words and sentences.

Subsequently, in order to improve the accuracy of the original language text converted from the original language speech, a process of allocating the original language text to each of the plurality of first proofreader terminal devices 500 may be performed based on the distribution model set according to the distribution unit ( S40). The distribution unit may be a set of one or a plurality of sentence chunks classified on the basis of a period, which is a period in the original language text. A detailed description of such a distribution model (S40) is shown in FIGS. 3 to 5 above. Referring to, it is omitted as it is duplicated with the above description.

Subsequently, a process of distributing the allocated original language text to each of the plurality of first proofreader terminal devices 500 according to the distribution model (S40) may be performed (S50).

Subsequently, the original language proofreader, who is the user of the first proofreader terminal device 500, may perform a process of correcting the text errors and insufficient parts according to the automatic text conversion (STT process) for the distributed original language text (S60). ).

Subsequently, the corrected original language text may be provided (transmitted) from the plurality of first proofreader terminal devices 500 to the distributed interpretation system 300, and a combination process may be performed (S70). In this case, each corrected original language text may be a distribution unit in which sentence lumps, which are distribution units, are gathered, and one corrected original language text completed by sequentially combining these distribution units can be completed.

Subsequently, a machine translation process of translating the corrected and combined original language text into a language selected by the first and second user terminal devices 400 and 700 according to an artificial intelligence (AI) algorithm may be performed (S80). For example, when the selected language is English, the original language text composed of Korean (Hangul) may be translated into translated text composed of English through a machine translation process (S80). In this machine translation process (S80), a technique for automatically translating a text of a specific known language into a text of another language can be used without limitation, but, for example, a neural network machine translation that learns the relationship between the original text and the corresponding translation with an artificial neural network. (NMT: Neural Machine Translation) can be applied to automatically translate Korean text (original text) into English text (translated text).

Subsequently, in order to improve the accuracy of the translated text that has been automatically translated, a process of allocating the translated text to each of the plurality of second proofreader terminal devices 600 may be performed based on a distribution model set according to a distribution unit (S90). ). The distribution unit may be a set of one or a plurality of sentence chunks classified on the basis of periods, which are periods in the original language text. A detailed description of such a distribution model (S90) is shown in FIGS. 3 to 5 above. Referring to, it is omitted as it is duplicated with the above description.

Subsequently, a process of distributing the allocated translated text according to the distribution model (S90) to each of the plurality of second proofreader terminal devices 600 may be performed (S100).

Then, the distributed translation text can be corrected by a translation proofreader, who is a user of the second proofreader terminal device 600, to correct errors and insufficient parts according to the automatic translation (S110).

Subsequently, the corrected translation text may be provided (transmitted) from the plurality of second proofreader terminal devices 600 to the distributed interpretation system 300, and a combination process may be performed (S120). In this case, each corrected text may be a distribution unit in which sentence chunks, which are distribution units, are gathered, and one corrected original text can be completed by sequentially combining these distribution units.

Subsequently, the distributed interpretation system 300 may perform a Text-To-Speech (TTS) process of recognizing the corrected and combined translation text information and converting it into voice information (interpretation voice). (S130). For example, in the case of a translated text composed of English, after recognizing the English text through the TTS process (S130), the corresponding English voice (interpretation voice) can be converted (synthesized). In this TTS process (S130), a known text-to-speech conversion technology can be used without limitation, but, for example, the voice is cut in advance into very short units such as phonemes and stored, and the phonemes constituting the sentence to be synthesized are combined to generate the voice. As a method of synthesizing concatenative speech synthesis (concatenative TTS), speech may be composed by synthesizing speech corresponding to a sentence.

Subsequently, a process of providing the interpretation voice synthesized through the TTS process (S130) in the distributed interpretation system 300 to the second user terminal device 700 may be performed (S140). The interpretation voice provided in this way can be heard by the user (listener) using the second user terminal device 700 through the speaker (including headset and earphone) function of the second user terminal device 700 as a sound wave. Provided so that (S150).

On the other hand, among the interpretation service processes based on the distribution model according to an embodiment of the present invention described above, a process of documenting and providing the translated text in a form requested by the user terminal device (400, 700) is additionally performed. It can be (S160). In this case, the document may include and provide information on the speaker and the proofreader (original language proofreader and/or interpretation proofreader) of the text, so that the quality of the interpretation service can be improved and responsibility can be clarified. In addition, in the process of providing the interpretation voice to the second user terminal device 700 (S140), the additionally corrected and combined translated text and the interpretation voice are combined with the content collected by the first proofreader terminal device 500. may be provided. Therefore, one content including other information (eg, video, image, text) in addition to the speaker's original voice is provided, and various services such as an online video conference interpretation service can be provided.

An example of providing interpretation services based on a distribution model

In the following detailed description, in an international video conference in which a number of speakers speaking different languages participate, for example, video content (including audio information) recorded by a camera of a speaker speaking Korean is transmitted to other participants. The process of providing real-time interpretation in English based on the distribution model according to the present invention is described as an example, but this is described as a representative example for convenience of explanation, and the present invention is not limited thereto.

7 is a screen showing an example of a process of providing an interpretation service based on a distribution model according to an embodiment of the present invention.

Referring to FIG. 7, when a speaker using Korean speaks, a camera included in the first user terminal device 400 takes a picture and at the same time records the voice of the announcement using a microphone function, so that video information including the voice is recorded. can be collected

Then, the video information of the video content can be transmitted in real time to the second user terminal device 700 through the operation server 200, and at the same time, in the distributed interpretation system 300, the video information included in the video content A speech-to-text (STT) process of extracting and recognizing only Korean voice information and converting it into Korean text information may be performed.

Subsequently, the Korean text converted from the Korean speech through the STT process is allocated and distributed to each of the plurality of first proofreader terminal devices 500 according to the distribution unit set based on the distribution model, and the distributed Korean text is distributed to the plurality of If the first proofreader terminal device 500 is corrected and transmitted, a first distribution proofreading process of collecting and recombining them may be performed.

Subsequently, a machine translation process of translating into a language selected by the first and second user terminal devices 400 and 700 may be performed according to an artificial intelligence (AI) algorithm. This machine translation process automatically translates Korean text into text in another language. For example, if the selected language is English, the Korean text can be translated into English text, but the present invention is not limited thereto, and multilingual (e.g. For example, it can be machine translated into more than 150 languages around the world).

Subsequently, in order to improve the accuracy of the automatically translated English text, the translated text is allocated and distributed to each of a plurality of second proofreader terminal devices 600 according to a distribution unit set based on the distribution model, and the distributed English text When is transmitted after being calibrated by the plurality of second calibrator terminal devices 600, a second distribution calibration process of collecting and combining them again may be performed.

Subsequently, the distributed interpretation system 300 may perform a Text-To-Speech (TTS) process of recognizing the corrected and combined English text information and converting it into English voice information (interpretation voice). there is.

Subsequently, a process of providing English voice information synthesized through the TTS process in the distributed interpretation system 300 to each of a plurality of second user terminal devices 700 may be performed. At this time, the first user terminal device 400 When the Korean voice is removed from the video content transmitted from and the English voice information is inserted, the second user terminal device 700 displays the video content through a speaker (including headset and earphone) function through the display screen. 1 The user terminal device 400 can conveniently listen to the interpretation voice along with visual information of the user (speaker).

Embodiments according to the present invention described above may be implemented in the form of program instructions that can be executed through various computer components and recorded on a computer-readable recording medium. The computer readable recording medium may include program instructions, data files, data structures, etc. alone or in combination. Program instructions recorded on the computer-readable recording medium may be specially designed and configured for the present invention, or may be known and usable to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, and magneto-optical media such as floptical disks. media), and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter or the like as well as machine language codes such as those produced by a compiler. The hardware device may be configured to act as one or more software modules to perform processing according to the present invention and vice versa.

In addition, in the detailed description of the present invention, specific details such as specific components and limited embodiments and drawings have been described, but this is only provided to help a more general understanding of the present invention, and the present invention is limited to the above embodiments It is not, and those skilled in the art can make various modifications and variations from these descriptions. Therefore, the spirit of the present invention is not limited to the above-described embodiments, and not only the claims to be described later, but also all modifications equivalent or equivalent to the claims belong to the scope of the spirit of the present invention. will do it

100: communication network
200: operation server
300: distributed interpretation system
310: first information conversion unit
320: first distribution correction unit
330: AI translation department
340: second distribution correction unit
350: second information conversion unit
360: database
370: communication department
380: control unit
400: first user terminal device
500: first corrector terminal device
600: second corrector terminal device
700: second user terminal device

Claims (10)

As a system for providing interpretation services based on a distribution model on a communication network,
a first information converter for recognizing original speech included in content collected and transmitted by the user terminal device and converting the recognized original speech into original language text;
The original language text is allocated and distributed to each first proofreader terminal device using a plurality of languages of the original language text according to a distribution unit set based on the corresponding distribution model, and the distributed original language text is distributed to the plurality of first proofreader terminals. If collected after being calibrated by the device, a first distribution calibration unit that combines them again;
an AI translation unit for generating translated text by machine-translating the corrected and combined original language text into a language selected by the user terminal device or another user terminal device according to an artificial intelligence algorithm;
The translated text is allocated and distributed to each second proofreader terminal device using a plurality of languages of the translated text according to a distribution unit set based on the corresponding distribution model, and the distributed translated text is distributed to the plurality of second proofreader terminals. If collected after being calibrated by the device, a second distribution correction unit that combines them again; and
A second information conversion unit for converting the corrected and combined translated text into an interpretation voice and providing it to the other user terminal device;
The distribution model is
A method of allocating and distributing the distributed original language text and the distributed translated text to each of the plurality of first and second proofreader terminal devices according to the distribution unit based on the proofreading time required to proofread and collect the distributed text,
The method of distributing based on the calibration time,
An allocation sequence is included for each of the plurality of first and second calibrator terminal devices,
The system characterized in that the assignment order takes precedence as the time for completion of proofreading among the distributed original language text and the distributed translated text is shorter.
According to claim 1,
The distribution unit is
A set of one or more sentence chunks divided based on periods, which are periods in the original language text and the translated text,
The distribution model is
a method of allocating and distributing each of the plurality of first and second proofreader terminal devices according to the distribution unit based on any one of the number of characters or words of the original language text and the translated text;
Redistribution according to the distribution unit based on one or a plurality of groups (groupings) selected by the plurality of first proofreader terminal devices or the plurality of second proofreader terminal devices among the distributed original language text and the distributed translated text A system characterized in that it further comprises a method consisting of any one of the methods or a combination thereof.
According to claim 2,
The method of distributing based on either the number of letters or the number of words,
Based on the number of letters or the number of words n (where n is a natural number),
Assigning and distributing to the first and second proofreader terminal devices according to the distribution unit composed of sentence chunks all included within the n sentences,
The system characterized in that allocating and distributing to the first and second proofreader terminal devices according to the distribution unit composed of sentence chunks including at least some of the n sentences.
According to claim 1,
The method of distributing based on the calibration time,
When the correction is completed within the minimum set time, in the next distribution order, twice the amount of the corrected distribution unit sentence chunk is set as a distribution unit and distributed,
When the correction is completed because the correction time exceeds the maximum set time, in the next distribution order, only 1/2 of the corrected distribution unit sentence chunk is set as a distribution unit and distributed,
When the correction time exceeds the minimum set time and the correction is completed within the maximum set time, the system characterized in that in the next distribution order, the amount of the correction-completed distribution unit sentence block is maintained and distributed.
According to claim 1,
The method of distributing based on the calibration time,
If the distribution unit held among the plurality of first and second corrector terminal devices exceeds the set range, the corresponding terminal device omit the allocation order in the next distribution order, and among the plurality of first and second corrector terminal devices A system characterized by redundantly distributing to a terminal device in which the correction of the previous dispensing sequence is completed with the shortest time.
According to claim 2,
The method of redistributing according to the distribution unit based on one or more groups selected by the plurality of first and second corrector terminal devices,
It is provided on a calibration interface configured differently for each of the plurality of first and second calibrator terminal devices,
The calibration interface is
A plurality of distribution text blocks in which the distribution unit sequentially combined among the distributed original language text or the distributed translated text is disposed, respectively;
A correction block for correcting and inputting any one or more of the plurality of distributed text blocks, and
It is configured to include a selection block for selecting and grouping whether or not to be corrected (activated) among the plurality of distribution text blocks,
Among the plurality of distribution text blocks in which the distribution unit is respectively disposed, the distribution unit in which the distribution unit allocated and distributed to each of the first and second corrector terminal devices is arranged is arranged in the center, and is sequentially combined with the distribution unit The system characterized in that the distribution text block on which is arranged is arranged in order around it.
According to claim 6,
The distributed text block arranged in the center is reflected in the correction block in an active state,
The distribution text block arranged in the periphery is in an inactive state, and when a corresponding button of the selection block is selected, it is changed to an active state, grouped with the distribution text block arranged in the center, and reflected in the correction block. .
According to claim 6,
On the calibration interface configured differently for each of the plurality of first and second calibrator terminal devices,
A calibration result of a terminal device that selects a grouped distributed text block is given priority over a single distributed text block selected from among the plurality of first and second corrector terminal devices;
When all of the grouped distributed text blocks selected from among the plurality of first and second corrector terminal devices overlap, the calibration result of the terminal device selected later in time takes precedence;
When some of the grouped distribution text blocks selected from among the plurality of first and second proofreader terminal devices overlap, the proofreading result of the distributed text block of the terminal device that selects the later distribution text block that is sequentially combined is given priority. system characterized by.
According to claim 1,
The second information conversion unit,
Based on the edited and combined translated text, document in a form requested by the user terminal device or other user terminal devices,
The system characterized in that the document includes and provides information on the speaker of the original speech, the proofreader of the original language text, and the proofreader of the translated text.
According to claim 1,
a matching unit that matches a plurality of first and second corrector terminal devices according to an interpretation condition requested by the user terminal device or another user terminal device;
The system according to claim 1 , wherein the matching unit firstly allocates according to the suitability of the interpretation condition based on the calibration statistical information of the plurality of first and second proofreader terminal devices.

Images