JP2019139629A - Machine translation device, translation learned model and determination learned model - Google Patents

Abstract

To reduce translation processing time in machine translation.SOLUTION: A machine translation device 1 comprises: translation sections 10 for respective categories, each of which uses a translation learned model obtained by machine learning using the corpus of category to which each translation section 10 belongs, and translates sentences in the category; a determination unit 11 which determines the category of input sentences using a determination learned model obtained by machine learning based on the corpus used in machine learning of each translation section 10; and a control unit 12 which translates the input sentences using the translation section 10 of the category determined by the determination unit 11 and then outputs translation results. The translation learned model may be a learned model based on a recurrent neural network. The determination learned model may be a learned model based on a recurrent neural network.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a machine translation device that performs machine translation, a translation learned model, and a determination learned model.

  In recent years, neural machine translation using a neural network has been attracting attention in machine translation. Compared with conventional statistical-based machine translation, more natural translation is possible. Therefore, a translation system based on machine learning using a neural network has been developed. For example, in Patent Document 1 below, a machine translation device using a neural network is disclosed.

Japanese Patent Laid-Open No. 06-332935

  However, in machine translation using a neural network or the like, there is a problem that as the number of vocabulary to be translated increases, for example, the size of the neural network increases, resulting in an increase in translation processing time.

  Accordingly, the present invention has been made in view of such problems, and an object thereof is to provide a machine translation device, a translation learned model, and a judgment learned model that can reduce translation processing time in machine translation.

  In order to solve the above-described problem, a machine translation device according to one aspect of the present invention is a translation unit for each category, and each translation unit uses a translation learned model obtained by machine learning using a corpus of a category to which the translation unit belongs. Uses a translation learning unit that translates sentences of the category to be used, and determines the category of the input sentence that has been input, using a learning model that has been obtained by machine learning based on the corpus used in the machine learning of each translation unit And a control unit that translates the input sentence using the translation unit of the category determined by the determination unit and outputs a translation result.

  According to such a machine translation apparatus, a sentence is translated using the translation learned model of the translation unit for each category. In this case, for example, the translation target vocabulary handled by a single translation unit can be limited to the vocabulary used in the category to which the translation belongs, and therefore the number of vocabularies handled by the translation learned model can be reduced. Thereby, the translation processing time of a single translation part can be suppressed, and the translation processing time of the whole machine translation apparatus can be suppressed. The determination learned model is a learned model obtained by machine learning based on the corpus used in the machine learning of the translation learned model of the translation unit. Since a category is determined using such a determination learned model, for example, a category of a sentence that the translation learned model is good at can be determined more accurately. Thereby, since translation is performed by a more accurate category translation unit, translation accuracy can be improved.

  According to the present invention, the translation processing time in machine translation can be suppressed.

It is a functional block diagram of the machine translation apparatus 1 which concerns on embodiment of this invention. It is a figure which shows the example of learning of the translation part. It is a figure which shows the example of learning of the determination part. It is a conceptual diagram which shows an example of the determination process by the determination part. 4 is a flowchart illustrating an example of a determination process performed by a determination unit 11. It is a graph which shows an example of the relationship between the number of vocabularies and translation processing time. It is a hardware block diagram of the machine translation apparatus 1 which concerns on embodiment of this invention.

  Hereinafter, embodiments of the apparatus will be described in detail with reference to the drawings. In the description of the drawings, the same elements are denoted by the same reference numerals, and redundant description is omitted. The embodiments in the following description are specific examples of the present invention, and are not limited to these embodiments unless specifically described to limit the present invention.

  FIG. 1 is a functional block diagram of the machine translation apparatus 1. As shown in FIG. 1, the machine translation apparatus 1 includes a translation unit 10A, a translation unit 10B, a translation unit 10C, a determination unit 11, and a control unit 12. The translation unit 10A, the translation unit 10B, and the translation unit 10C are collectively referred to as the translation unit 10 as appropriate.

  The machine translation device 1 is a computer device that translates a sentence (input sentence) input by a user or the like and outputs a translation result to the user or the like. Translation means that a sentence expressed in a certain language (first language) is replaced with another language (second language). An example is translation from a Japanese sentence to an English sentence, but is not limited to this. A sentence is composed of one or more sentences. A sentence is a series of words written in letters. The translation result is information based on the translation result, and is, for example, a translated sentence (translated sentence), but is not limited thereto.

  Hereinafter, each functional block of the machine translation apparatus 1 shown in FIG. 1 will be described.

  The translation unit 10 is a translation unit 10 for each category, and each translation unit 10 uses a learned model (translation learned model) obtained by machine learning using a corpus of a category to which the translation unit 10 belongs, to translate sentences in the category. translate. As shown in FIG. 1, the machine translation apparatus 1 includes a category A translation unit 10A (translation unit 10A belongs to category A), a category B translation unit 10B (translation unit 10B belongs to category B), and a category C translation. 10C (translation unit 10C belongs to category C). Note that the number of categories is not limited to three, and may be any number as long as it is one or more.

  A category is a classification of text, text type, text genre, text area, text field, or text domain. The category corresponds to a usage scene, for example, a category strong against spoken language often used in hospitals, a category strong against written language often used in manuals in the IT (Information Technology) field, and the like. In the present embodiment, it is assumed that category A is a medical category, category B is a travel category, and category C is a shopping category.

  The corpus is a part or all of data (database) obtained by structuring natural language sentences and collecting them on a large scale. The corpus in the present embodiment assumes a corpus (a bilingual corpus, bilingual data) in which a first language (Japanese) is associated with a second language (English) that is a translation of the first language. The process of obtaining a learned model by machine learning using a corpus and the process of translating a sentence using the learned model are processes according to the prior art.

  A learned model is a combination of a computer program and parameters. The learned model is a combination of the structure of the neural network and a parameter (weighting coefficient) that is the strength of connection between the neurons of the neural network. The learned model is an instruction to the computer, and is a combination of programs so that one result can be obtained (predetermined processing is executed), that is, a computer program that causes the computer to function.

  FIG. 2 is a diagram illustrating a learning example of the translation unit 10. For example, the translation unit 10A belonging to the category A uses a category A dedicated corpus and a general-purpose corpus as learning data used when performing machine learning. The category A dedicated corpus is a corpus specialized for category A, for example, a corpus related to sentences often used in the medical field. The general-purpose corpus is a corpus related to general-purpose words. In order to reduce the number of vocabularies, a general-purpose corpus is composed of sentences or words that do not include special technical terms or terms that are used only in special categories. The learning of the translation unit 10B and the translation unit 10C is the same as the learning of the translation unit 10A. That is, the translation unit 10B belonging to the category B uses a category B-dedicated corpus (for example, a corpus related to sentences often used in the travel field) and a general-purpose corpus as learning data, and the translation unit 10C belonging to the category C receives category data as learning data. A C-dedicated corpus (for example, a corpus related to sentences often used in the shopping field) and a general-purpose corpus are used.

  As described above, each translation unit 10 belonging to each category is trained with a corpus obtained by adding a general-purpose corpus to a dedicated corpus of the category. Note that the dedicated corpus may be exclusive. In this case, the same sentence is not included in another corpus. When machine learning is performed, a general-purpose corpus that is commonly used as described above is used in some or all of the translation units 10 because only a dedicated corpus cannot handle general-purpose phrases. In addition, any one or more of the translation unit 10A, the translation unit 10B, and the translation unit 10C may use only a dedicated corpus without using a general-purpose corpus as learning data.

  The translation learned model may be a learned model based on a neural network. The translation learned model may be a learned model based on a recurrent neural network. The translation learned model is not limited to a neural network, but may be a learned model based on information processing capable of machine learning.

  The translation learned model is a translation unit 10 for each category, and each translation unit 10 translates a sentence of the category to which the translation unit 10 belongs, a determination unit 11 that determines a category of an input sentence that has been input, and a determination The translation learning used when the translation unit 10 of the machine translation device 1 includes the control unit 12 that translates the input sentence using the translation unit 10 of the category determined by the unit 11 and outputs the translation result. A neural network in which a weighting coefficient is learned using a corpus (dedicated corpus) of a category to which the translation unit 10 using the translation learned model belongs, and the determination unit 11 includes the translation learned model. Even if a category is judged using a judgment learned model obtained by machine learning based on a corpus (dedicated corpus) used in learning There.

  The translation-learned model is a translation-learned model for causing a computer to function so as to translate input text of a predetermined category and output a translation result, using a corpus (dedicated corpus) of the category. The category may be a category of an input sentence that is determined using a determination learned model obtained by machine learning based on the corpus (dedicated corpus). .

  The determination unit 11 uses a corpus (dedicated corpus) in which the category of the input sentence (from the control unit 12 described later) is used in the machine learning of each translation unit 10 (the translation unit 10A, the translation unit 10B, and the translation unit 10C). Judgment is performed by using a learning model that has been obtained by machine learning based on the above. The process of obtaining a learned model by machine learning based on a corpus and the process of determining a sentence category using a learned model are processes according to the prior art.

  FIG. 3 is a diagram illustrating a learning example of the determination unit 11. The determination unit 11 includes, as learning data, a category A dedicated corpus used in the machine learning of the translation learned model of the translation unit 10A, a category B dedicated corpus used in the machine learning of the translation learned model of the translation unit 10B, and a translation unit A corpus dedicated to category C used in machine learning of the 10C translation learned model is used. More specifically, the determination unit 11 includes, as learning data, a combination of a first language of the category A dedicated corpus and information (label) indicating the category A, and information indicating the first language and category B of the category B dedicated corpus. A set of (label) and a set of information (label) indicating the first language of category C-dedicated corpus and category C are used. In the learning of the determination unit 11, the first language sentence of the dedicated corpus of a predetermined category is used as an input, and the output (correct data) is learned as the label of the category.

  The learning of the determination unit 11 will be described with a specific example. The determination unit 11 first labels medical corpus (category A), travel corpus (category B), and shopping corpus (category C). As an example, the medical corpus text is 1, the travel corpus text is 2, and the shopping corpus text is 3. Then, the sentence of each corpus is input to a judgment learned model (such as a recurrent neural network) of the judgment unit 11, and an output vector in which the dimension of the label of the input sentence is “1” and the others are “0” is learned as correct data. Let In this case, since there are three types of categories, it becomes a three-dimensional output vector, “(1,0,0)” for a medical corpus, “(0,1,0)” for a travel corpus, and In the case of a shopping corpus, “(0, 0, 1)” is correct data. By learning in this way, a category suitable for translating the input sentence can be determined (identified).

  The determination learned model may be a learned model based on a neural network. The determination learned model may be a learned model based on a recurrent neural network. The determination learned model is not limited to a neural network, and may be a learned model based on information processing capable of machine learning.

  The determination learned model is the translation unit 10 for each category, and each translation unit 10 uses the translation learned model obtained by machine learning using the corpus (dedicated corpus) of the category to which the translation unit 10 belongs. Control for translating, translating unit 10, determining unit 11 for determining the category of the input sentence that has been input, and translating the input sentence by using the translating unit 10 for the category determined by determining unit 11, and outputting the translation result And a weighting coefficient based on a corpus (dedicated corpus) used in the machine learning of each translation unit 10, which is used when the determination unit 11 of the machine translation device 1 including the unit 12 determines a category. May be configured by a learned neural network.

  The decision learned model is a decision learned model for causing a computer to function so as to determine a category of an input sentence, and a neural network in which a weighting coefficient is learned based on a predetermined corpus (dedicated corpus). The input sentence may be translated using a translation learned model obtained by machine learning using the corpus (dedicated corpus) of the category determined based on the determined learned model.

  FIG. 4 is a conceptual diagram illustrating an example of determination processing by the determination unit 11. FIG. 5 is a flowchart illustrating an example of the determination process performed by the determination unit 11. Hereinafter, an example of the determination process performed by the determination unit 11 will be described with reference to FIG. 4 and based on the flowchart of FIG. 5. First, an input sentence is input from the control unit 12 (step S1). Next, the input sentence input in S1 is subjected to morphological analysis (divided into words) to generate a word string (step S2). Next, words are extracted from the word string generated in S2 (in order from the front) (step S3). Next, the word is converted into a word ID and input (sequentially) to the input layer of the neural network (determined learning model) (step S4). Next, it is determined whether or not words remain in the word string (step S5). If it is determined in S5 that it remains (S5: YES), the process returns to S3, and the processes of S3 to S5 are repeated for the remaining words. On the other hand, if it is determined in S5 that it does not remain (S5: NO), that is, if all the word strings of the input sentence are input, “<EOS>” (a symbol indicating the end of the sentence) is input to the neural network. Input to the layer (step S6). Next, an output vector having the number of categories to be classified as the number of dimensions is calculated from the vectors of the intermediate layer of the neural network to obtain an output layer. This output vector is subjected to likelihood calculation by Softmax processing or the like, and the category corresponding to the dimension with the highest likelihood is determined as a determination result (identified as the optimum category) (step S7).

  Below, it supplements about the description regarding the above-mentioned FIG.4 and FIG.5. The neural network described above may be a recurrent neural network. The input layer of the neural network is set in advance to a predetermined number of dimensions. For example, when the number of vocabulary of the input sentence is 50,000 words, it is set to 500 dimensions. The output layer of the above-described neural network is set to a vector having the number of categories as the number of dimensions. The above-described intermediate layer vector is used to make a determination (identification) at a non-linear boundary. The intermediate layer vector is also set to a predetermined number of dimensions. For example, the intermediate layer vector is set to the same number of dimensions as the input layer. By linearly mapping the intermediate layer vector to the output layer vector, the output layer vector can be calculated. Although the highest value of the vector in the output layer may be used, the vector value can be normalized and converted into a probability distribution value by performing Softmax processing. Thereby, as correct answer data used at the time of learning, a vector of “1” only for the corresponding category and “0” for the other categories can be used as correct answer data.

  In S4, the word is converted into a vector having a predetermined number of dimensions and input to the input layer. In S4, the word ID is an ID assigned to each word, and is assigned in order from “1” in descending order of frequency. For example, IDs “1” to “30000” are assigned to each word, and 30,000 or less types of infrequent words are discarded. When this word ID is input to the input layer, the ID is compressed (converted) into, for example, a 300-dimensional vector. As the type of word ID is smaller, the size of this vector can be reduced, and the amount of calculation can be reduced. Note that the compression from 30,000 words to 300 dimensions is lossless compression. The compression uses a conventional technique such as word2vec. In compression, the vectors are basically constructed so that the Euclidean distance between the vectors of each word is as large as possible.

  Returning to the description of each functional block shown in FIG. 1, the control unit 12 inputs an input sentence from another device or the like via a user, a network, or the like. Next, the control unit 12 outputs the input sentence to the determination unit 11 and translates the input sentence using the category translation unit 10 determined by the determination unit 11 according to the output. Next, the control unit 12 outputs (displays) the translation result to another device or the like via a user, a network, or the like.

  Next, the effect of the machine translation apparatus 1 configured as in the present embodiment will be described.

  According to the machine translation apparatus 1 of this embodiment, a sentence is translated using the translation learned model of the translation unit 10 for each category. In this case, for example, the vocabulary to be translated handled by the single translation unit 10 can be limited to the vocabulary used in the category to which the translation belongs, and therefore the number of vocabularies handled by the translation learned model can be reduced. Thereby, the translation processing time of the single translation part 10 can be suppressed, and the translation processing time of the whole machine translation apparatus 1 can be suppressed. The determination learned model is a learned model obtained by machine learning based on the corpus used in the machine learning of the translation learned model of the translation unit 10. Since a category is determined using such a determination learned model, for example, a category of a sentence that the translation learned model is good at can be determined more accurately. As a result, translation is performed by the translation unit 10 of a more accurate category, so that translation accuracy can be improved. That is, the accuracy of translation is not reduced while the categories are dispersed and the number of vocabularies in each category is reduced to speed up the processing time. In addition, since the corpus used in the machine learning of the translation learned model is also used in the machine learning of the judgment learned model, the best performance can be achieved by matching the learning data, and the learning data can be reused. Time and cost for preparing learning data can be reduced.

  In addition, when the translation learned model is a learned model based on a recurrent neural network, optimal machine translation is performed in consideration of not only the word (word level) of the input sentence but also the word order, wording or expression of the input sentence. Since this can be done, translation accuracy can be improved.

  In addition, when the judgment learned model is a learned model based on a recurrent neural network, not only the word (word level) of the input sentence but also the word order, wording or expression of the input sentence is considered, and the optimum category is determined. It can be performed. As a result, translation is performed by the translation unit 10 of a more accurate category, so that translation accuracy can be improved.

  The machine translation apparatus 1 according to the present embodiment distributes vocabulary to be handled to a plurality of translation units 10 and increases the number of vocabularies that can be handled while shortening the translation processing time. Specifically, in the machine translation apparatus 1, for example, a plurality of translation units 10 composed of, for example, a neural network are prepared, and a good category is assigned to each. By preparing the translation unit 10 for each category, the number of vocabularies handled by a single translation unit 10 can be limited to the vocabulary used in the category, and thus the number of vocabularies can be reduced. Therefore, the translation processing time of the translation unit 10 can be reduced. However, if the user selects which category of the translation unit 10 should be used for translating the input sentence, there is a possibility that convenience may be impaired. Therefore, the machine translation apparatus 1 includes a determination unit 11 that determines (identifies) a category in which the input sentence is optimally translated. Then, the input sentence is sent to the translation unit 10 of the determined optimum category, and the translation result is output as a translated sentence (output sentence).

  Further, as a method for learning the determination learned model of the determination unit 11, the dedicated corpus used when learning the translation unit 10 of each category is labeled, the sentence of the corpus for the category is input, and output (correct data) ) Learns to be the label for that category. This is because a sentence close to a sentence that the later translation unit 10 is good at is determined as the category. Further, for example, by using a recurrent neural network similar to the translation unit 10 in the subsequent stage as the determination learned model, it is possible to perform optimum category classification considering not only the words of the input sentence but also the word order and expression.

  With the machine translation apparatus 1 of this embodiment, the number of vocabularies can be distributed to a plurality of translation units 10, and the translation process can be speeded up. In addition, the determination unit 11 allows the user to input text into the machine translation device 1 without being conscious of which category to translate, so that the machine translation device 1 can handle an enormous vocabulary when viewed from the user. Therefore, the machine translation apparatus 1 that can handle a large vocabulary can be operated in real time by a CPU (Central Processing Unit) without using a GPU (Graphics Processing Unit), and the computer cost can be reduced. . FIG. 6 is a graph showing an example of the relationship between the number of vocabularies and the translation processing time. As shown in the graph, in the machine translation according to the prior art, the translation processing time increases in proportion to the square as the number of vocabulary increases. On the other hand, the processing time of the machine translation apparatus 1 increases linearly as the number of vocabularies increases, but is shorter than that of the prior art. Accordingly, a plurality of translation units 10 having a small number of vocabularies are prepared, and the determination unit 11 distributes the translations to the appropriate category translation units 10 for faster processing. Further, by specializing the translation unit 10 in a dedicated category, there is an effect that a phrase peculiar to the category can be translated well, and there is a merit in terms of accuracy improvement.

  In the prior art using a neural network, the larger the vocabulary number, the larger the size of the input vector and the size of the output vector, and there is a problem that it takes time for learning and calculation of translation. This is because an ID is assigned to each of a large amount of vocabulary data, and an input vector and an output vector having the number of IDs as the number of dimensions are used for learning a neural network, so that the vector size increases and the neural network size also increases. This is because the calculation time becomes long. As the number of vocabulary handled increases, the number of vocabulary increases. At present, the number of vocabulary of about 50,000 words can be translated in real time by the CPU. In order to handle a larger vocabulary, it is necessary to use high-performance GPUs in parallel, which increases the computer cost.

  The machine translation apparatus 1 according to this embodiment relates to distributed processing and accuracy improvement of a machine translation system using a neural network. The machine translation apparatus 1 prepares a plurality of translation units 10, specializes each translation unit 10 in a specific category, and reduces the number of vocabularies handled by each translation unit 10, thereby reducing calculation time. In addition, by allocating input sentences to optimal categories by the determination unit 11, a machine translation system that can handle sentences of various categories can be realized from the input side.

  A machine translation device 1A, which is a modification of the machine translation device 1, includes a domain classification unit that classifies an input sentence into a predetermined type of domain, and a machine translation unit for each domain, and is used for learning of the machine translation unit A dedicated corpus is used for learning of the domain classifying unit, the domain classifying unit identifies a domain in which the input sentence is most suitable, and the machine translation unit of the identified domain translates the input sentence, and a translation result is obtained. Output. The domain classification means may be a classifier configured by a recurrent neural network. Each domain corpus that causes the machine translation unit of each domain to learn may be used for learning of the recurrent neural network constituting the domain classification unit. The machine translation apparatus 1A includes a morpheme analysis unit that divides an input sentence into words, and includes means for converting a word into a vector in the recurrent neural network of the domain classification unit, and the vector is input to the recurrent neural network. It may be an input to the layer. The machine translation device 1A sequentially inputs the words of the input sentence to the recurrent neural network, generates a vector of the output layer from the intermediate layer at the time when the input is completed, and the likelihood of each domain from the vector value of the output layer May be calculated.

  In addition, the block diagram used for description of the said embodiment has shown the block of the functional unit. These functional blocks (components) are realized by any combination of hardware and / or software. Further, the means for realizing each functional block is not particularly limited. That is, each functional block may be realized by one device physically and / or logically coupled, and two or more devices physically and / or logically separated may be directly and / or indirectly. (For example, wired and / or wireless) and may be realized by these plural devices.

  For example, the machine translation apparatus 1 according to an embodiment of the present invention may function as a computer that performs processing of the determination method and the machine translation method of the present invention. FIG. 7 is a diagram illustrating an example of a hardware configuration of the machine translation apparatus 1 according to the embodiment of the present invention. The machine translation device 1 described above may be physically configured as a computer device including a processor 1001, a memory 1002, a storage 1003, a communication device 1004, an input device 1005, an output device 1006, a bus 1007, and the like.

  In the following description, the term “apparatus” can be read as a circuit, a device, a unit, or the like. The hardware configuration of the machine translation device 1 may be configured to include one or a plurality of each device illustrated in the figure, or may be configured not to include some devices.

  Each function in the machine translation apparatus 1 is such that a predetermined software (program) is read on hardware such as the processor 1001 and the memory 1002 so that the processor 1001 performs an operation, and communication by the communication device 1004, memory 1002 and storage This is realized by controlling reading and / or writing of data in 1003.

  For example, the processor 1001 controls the entire computer by operating an operating system. The processor 1001 may be configured by a central processing unit (CPU) including an interface with peripheral devices, a control device, an arithmetic device, a register, and the like. For example, the translation unit 10, the determination unit 11, the control unit 12, and the like described above may be realized by the processor 1001.

  Further, the processor 1001 reads a program (program code), software module, and data from the storage 1003 and / or the communication device 1004 to the memory 1002, and executes various processes according to these. As the program, a program that causes a computer to execute at least a part of the operations described in the above embodiments is used. For example, the translation unit 10, the determination unit 11, and the control unit 12 may be realized by a control program stored in the memory 1002 and operated by the processor 1001, and may be realized similarly for other functional blocks. Although the above-described various processes have been described as being executed by one processor 1001, they may be executed simultaneously or sequentially by two or more processors 1001. The processor 1001 may be implemented by one or more chips. Note that the program may be transmitted from a network via a telecommunication line.

  The memory 1002 is a computer-readable recording medium and includes, for example, at least one of ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), RAM (Random Access Memory), and the like. May be. The memory 1002 may be called a register, a cache, a main memory (main storage device), or the like. The memory 1002 can store a program (program code), a software module, and the like that can be executed to implement the determination method and the machine translation method according to an embodiment of the present invention.

  The storage 1003 is a computer-readable recording medium such as an optical disc such as a CD-ROM (Compact Disc ROM), a hard disc drive, a flexible disc, a magneto-optical disc (eg, a compact disc, a digital versatile disc, a Blu-ray). (Registered trademark) disk, smart card, flash memory (for example, card, stick, key drive), floppy (registered trademark) disk, magnetic strip, and the like. The storage 1003 may be referred to as an auxiliary storage device. The storage medium described above may be, for example, a database, server, or other suitable medium including the memory 1002 and / or the storage 1003.

  The communication device 1004 is hardware (transmission / reception device) for performing communication between computers via a wired and / or wireless network, and is also referred to as a network device, a network controller, a network card, a communication module, or the like. For example, the control unit 12 and the like may be realized by the communication device 1004.

  The input device 1005 is an input device (for example, a keyboard, a mouse, a microphone, a switch, a button, a sensor, or the like) that accepts an external input. The output device 1006 is an output device (for example, a display, a speaker, an LED lamp, etc.) that performs output to the outside. The input device 1005 and the output device 1006 may have an integrated configuration (for example, a touch panel).

  Each device such as the processor 1001 and the memory 1002 is connected by a bus 1007 for communicating information. The bus 1007 may be configured with a single bus or may be configured with different buses between apparatuses.

  The machine translation apparatus 1 includes hardware such as a microprocessor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a programmable logic device (PLD), and a field programmable gate array (FPGA). A part or all of each functional block may be realized by the hardware. For example, the processor 1001 may be implemented by at least one of these hardware.

  The notification of information is not limited to the aspect / embodiment described in the present specification, and may be performed by other methods. For example, notification of information includes physical layer signaling (for example, DCI (Downlink Control Information), UCI (Uplink Control Information)), upper layer signaling (for example, RRC (Radio Resource Control) signaling, MAC (Medium Access Control) signaling), It may be implemented by broadcast information (MIB (Master Information Block), SIB (System Information Block))), other signals, or a combination thereof. The RRC signaling may be referred to as an RRC message, and may be, for example, an RRC connection setup message, an RRC connection reconfiguration message, or the like.

  Each aspect / embodiment described herein includes LTE (Long Term Evolution), LTE-A (LTE-Advanced), SUPER 3G, IMT-Advanced, 4G, 5G, FRA (Future Radio Access), W-CDMA. (Registered trademark), GSM (registered trademark), CDMA2000, UMB (Ultra Mobile Broadband), IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802.20, UWB (Ultra-WideBand), The present invention may be applied to a Bluetooth (registered trademark), a system using other appropriate systems, and / or a next generation system extended based on these systems.

  As long as there is no contradiction, the order of the processing procedures, sequences, flowcharts, and the like of each aspect / embodiment described in this specification may be changed. For example, the methods described herein present the elements of the various steps in an exemplary order and are not limited to the specific order presented.

  Input / output information or the like may be stored in a specific location (for example, a memory) or may be managed by a management table. Input / output information and the like can be overwritten, updated, or additionally written. The output information or the like may be deleted. The input information or the like may be transmitted to another device.

  The determination may be performed by a value represented by 1 bit (0 or 1), may be performed by a true / false value (Boolean: true or false), or may be performed by comparing numerical values (for example, a predetermined value) Comparison with the value).

  Each aspect / embodiment described in this specification may be used independently, may be used in combination, or may be switched according to execution. In addition, notification of predetermined information (for example, notification of being “X”) is not limited to explicitly performed, but is performed implicitly (for example, notification of the predetermined information is not performed). Also good.

  Although the present invention has been described in detail above, it will be apparent to those skilled in the art that the present invention is not limited to the embodiments described herein. The present invention can be implemented as modified and changed modes without departing from the spirit and scope of the present invention defined by the description of the scope of claims. Therefore, the description of the present specification is for illustrative purposes and does not have any limiting meaning to the present invention.

  Software, whether it is called software, firmware, middleware, microcode, hardware description language, or other names, instructions, instruction sets, codes, code segments, program codes, programs, subprograms, software modules , Applications, software applications, software packages, routines, subroutines, objects, executable files, execution threads, procedures, functions, etc. should be interpreted broadly.

  Also, software, instructions, etc. may be transmitted / received via a transmission medium. For example, software may use websites, servers, or other devices using wired technology such as coaxial cable, fiber optic cable, twisted pair and digital subscriber line (DSL) and / or wireless technology such as infrared, wireless and microwave. When transmitted from a remote source, these wired and / or wireless technologies are included within the definition of transmission media.

  Information, signals, etc. described herein may be represented using any of a variety of different technologies. For example, data, commands, commands, information, signals, bits, symbols, chips, etc. that may be referred to throughout the above description are voltages, currents, electromagnetic waves, magnetic fields or magnetic particles, light fields or photons, or any of these May be represented by a combination of

  Note that the terms described in this specification and / or terms necessary for understanding this specification may be replaced with terms having the same or similar meaning.

  As used herein, the terms “system” and “network” are used interchangeably.

  In addition, information, parameters, and the like described in this specification may be represented by absolute values, may be represented by relative values from a predetermined value, or may be represented by other corresponding information. . For example, the radio resource may be indicated by an index.

  The names used for the parameters described above are not limiting in any way. Further, mathematical formulas and the like that use these parameters may differ from those explicitly disclosed herein. Since various channels (eg, PUCCH, PDCCH, etc.) and information elements (eg, TPC, etc.) can be identified by any suitable name, the various names assigned to these various channels and information elements are However, it is not limited.

  As used herein, the terms “determining” and “determining” may encompass a wide variety of actions. “Judgment” and “determination” are, for example, judgment, calculation, calculation, processing, derivation, investigating, looking up (eg, table) , Searching in a database or other data structure), ascertaining what has been ascertaining, and so on. In addition, “determination” and “determination” are reception (for example, receiving information), transmission (for example, transmitting information), input (input), output (output), and access. (Accessing) (eg, accessing data in a memory) may be considered as “determined” or “determined”. In addition, “determination” and “determination” means that “resolving”, “selecting”, “choosing”, “establishing”, and “comparing” are regarded as “determining” and “determining”. May be included. In other words, “determination” and “determination” may include considering some operation as “determination” and “determination”.

  The terms “connected”, “coupled”, or any variation thereof, means any direct or indirect connection or coupling between two or more elements and It can include the presence of one or more intermediate elements between two “connected” or “coupled” elements. The coupling or connection between the elements may be physical, logical, or a combination thereof. As used herein, the two elements are radio frequency by using one or more wires, cables and / or printed electrical connections, and as some non-limiting and non-inclusive examples By using electromagnetic energy, such as electromagnetic energy having a wavelength in the region, microwave region, and light (both visible and invisible) region, it can be considered to be “connected” or “coupled” to each other.

  The reference signal may be abbreviated as RS (Reference Signal), and may be referred to as a pilot depending on an applied standard.

  As used herein, the phrase “based on” does not mean “based only on,” unless expressly specified otherwise. In other words, the phrase “based on” means both “based only on” and “based at least on.”

  Any reference to elements using the designations "first", "second", etc. as used herein does not generally limit the amount or order of those elements. These designations can be used herein as a convenient way to distinguish between two or more elements. Thus, a reference to the first and second elements does not mean that only two elements can be employed there, or that in some way the first element must precede the second element.

  The “means” in the configuration of each apparatus described above may be replaced with “unit”, “circuit”, “device”, and the like.

  As long as “including”, “comprising” and variations thereof are used in the specification or claims, these terms are inclusive of the term “comprising”. Intended to be Furthermore, the term “or” as used herein or in the claims is not intended to be an exclusive OR.

  Throughout this disclosure, if articles are added by translation, for example, a, an, and the, in English, these articles may be plural unless the context clearly indicates otherwise. Shall be included.

  DESCRIPTION OF SYMBOLS 1 ... Machine translation apparatus, 10 ... Translation part, 11 ... Determination part, 12 ... Control part, 1001 ... Processor, 1002 ... Memory, 1003 ... Storage, 1004 ... Communication apparatus, 1005 ... Input device, 1006 ... Output device.

Claims (5)

A translation unit for each category, each translation unit translating a sentence of the category using a translation learned model obtained by machine learning using a corpus of the category to which the translation belongs;
A determination unit that determines an input sentence category using a determination learned model obtained by machine learning based on the corpus used in the machine learning of each of the translation units;
A control unit that translates the input text using the translation unit of the category determined by the determination unit, and outputs a translation result;
A machine translation apparatus comprising:   The machine translation apparatus according to claim 1, wherein the translation learned model is a learned model based on a recurrent neural network.   The machine translation apparatus according to claim 1, wherein the determination learned model is a learned model based on a recurrent neural network. A translation unit for each category, each translation unit translating a sentence of the category to which the translation unit belongs, a determination unit that determines a category of the input sentence that has been input, and the translation of the category determined by the determination unit A translation learning model used when the translation unit of the machine translation device includes a controller that translates the input sentence using a unit and outputs a translation result,
It is composed of a neural network in which weighting coefficients are learned using a corpus of a category to which the translation unit using the translation learned model belongs,
The determination unit determines a category using a determination learned model obtained by machine learning based on the corpus used in learning of the translation learned model.
Translation learned model. A translation unit for each category, and each translation unit translates a sentence of the category using a translation learned model obtained by machine learning using a corpus of the category to which the translation unit belongs, and an input sentence input A determination unit that determines a category of the machine translation device, and a control unit that translates the input sentence using the translation unit of the category determined by the determination unit and outputs a translation result. A model that has been learned and used for judgment when judging categories,
It is constituted by a neural network in which weighting factors are learned based on the corpus used in the machine learning of each translation unit,
Judgment learned model.

Images