Information processing system and information processing method

Abstract

Provided is a novel information processing system that exhibits exceptional convenience, usefulness, or reliability. The information processing system can be used for comparing an original sentence with a translated sentence and checking whether translation is accurately performed, and has a function of performing information processing that differ between a case where literal translation is required and a case where inclusion of liberal translation is permitted. The information processing system has a function of, when a translated sentence includes a liberally translated portion in the case where literal translation is required, detecting said liberally translated portion, and urging a user of the information processing system to revise said portion, and further has a function of determining the correspondence relationship between words and phrases of the original sentence and the translated sentence. Still further, the information processing system has a function of, in the case where inclusion of liberal translation is permitted, determining whether the translated sentence is synonymous with the original sentence.

Description

Information processing system and information processing method

[0001] One aspect of the present invention relates to an information processing system and an information processing method.

[0002] Note that one embodiment of the present invention is not limited to the above technical field. The technical field of one embodiment of the invention disclosed in this specification and the like relates to an object, a method, or a manufacturing method. Alternatively, one embodiment of the present invention relates to a process, a machine, manufacture, or a composition of matter. Therefore, more specifically, examples of the technical field of one embodiment of the present invention disclosed in this specification include a semiconductor device, a display device, a light-emitting device, a power storage device, a memory device, a driving method thereof, or a manufacturing method thereof.

[0003] Research and development into machine translation, which uses a computer to translate one natural language into another, is currently underway. Examples of machine translation include rule-based machine translation, which performs translation based on rules, statistical machine translation, which performs translation using a language model and a translation model, and neural machine translation, which performs translation using an artificial neural network (ANN, hereinafter simply referred to as a neural network).

[0004] There are two types of translation: literal translation and free translation. Literal translation means translating the original text so that every word is replaced faithfully. Free translation means translating the original text without being bound by every word, but by taking into account the context and nuances, and focusing on the overall meaning. For this reason, free translation may add words that are not in the original text.

[0005] Neither literal translation nor free translation is necessarily correct, but rather should be selected appropriately depending on the purpose, situation, etc. There are cases where literal translation is preferable, and cases where free translation is preferable. In machine translation, a translation device capable of free translation is also being considered (Patent Document 1).

[0006] In recent years, the development of language models using neural networks has been actively pursued, with large-scale language models (LLMs) attracting particular attention. A large-scale language model is a natural language processing model trained using a large amount of data. A large-scale language model can realize, for example, a dialogue model that responds to user instructions. Non-Patent Document 1 discloses GPT-4 (registered trademark) (Generative Pre-trained Transformer 4) as a large-scale language model, and also discloses ChatGPT as a chat service.

[0007] The use of large-scale language models has significantly increased the capabilities of natural language processing models. However, as language models become larger, it is difficult to incorporate and operate language models in-house due to the equipment and cost involved. Therefore, one way to use language models is to use external services that provide language models.

[0008] JP 2009-205518

[0009] Summary of ChatGPT / GPT-4 Research and Perspective Towards the Future of Large Language Models, Yiheng Liu et al. (Submitted on 4 Apr 2023, [online], Internet <URL: https: / / arxiv.org / abs / 2304.01852>

[0010] The accuracy of machine translation is improving every year, and machine-translated text can be used as is in cases where it is sufficient for the reader to understand the general content or to understand the general content.

[0011] However, machine-translated texts often contain mistranslations, such as "information omissions" (missing information), where elements contained in the original text are not reflected in the translation, and "information overload" (overflow), where elements not contained in the original text are included in the translation.

[0012] There are many documents, such as contracts, manuals, official documents, patent documents, and academic papers, in which mistranslation is unacceptable. When translating such documents, whether the translation is done by hand or machine, the process of checking and correcting the translation after it has been completed is important. Note that the process of manually checking and correcting machine-translated text is sometimes called post-editing, and the international standard ISO 18587 specifies requirements for post-editing.

[0013] Checking and correcting translations requires the translator's specialized knowledge and skills, while also being required to complete the task within limited time and cost. Furthermore, when checking and correcting translations, the content of the checks differs depending on whether the translation is a literal or a free translation. In other words, the content of the checks differs for translations that require a literal translation and translations that are allowed to include a free translation. Therefore, if an information processing system with a checking function that can flexibly handle both literal and free translations were to be realized, it would be possible to efficiently check and correct translations.

[0014] Therefore, an object of one embodiment of the present invention is to provide an information processing system capable of efficiently checking and correcting a translation. Another object of one embodiment of the present invention is to provide an information processing system capable of determining whether a translation includes a free translation. Another object of the present invention is to provide an information processing system capable of determining whether an original text and a translation are synonymous when a translation includes a free translation. Another object of the present invention is to provide an information processing system capable of outputting, when a translation is a literal translation, a correspondence between a word included in the original text and a word in the translation corresponding to the word. Another object of the present invention is to provide an information processing method used in any one or more of the above information processing systems. Another object of the present invention is to provide an information processing device including any one or more of the above information processing systems.

[0015] An object of one embodiment of the present invention is to provide a novel information processing system with excellent convenience, usefulness, or reliability, or to provide a novel information processing method with excellent convenience, usefulness, or reliability, or to provide a novel information processing system, a novel information processing method, or a novel semiconductor device.

[0016] Note that the description of these problems does not preclude the existence of other problems. Note that one embodiment of the present invention does not necessarily solve all of these problems. Note that problems other than these will become apparent from the description of the specification, drawings, claims, etc., and it is possible to extract other problems from the description of the specification, drawings, claims, etc.

[0017] One aspect of the present invention is an information processing system that includes a first information processing device and a second information processing device, wherein the first information processing device has the following functions: a function of accepting input of a first document in a first language and a second document in a second language that is a translation of the first document; a function of obtaining the results of the determination by sending an instruction statement for determining whether the second document contains a free translation of the first document and a prompt including the first document and the second document to the second information processing device; a function of outputting a correspondence between words contained in the second document that has been determined not to contain a free translation and words contained in the first document by sending an instruction statement for outputting a prompt including the first document and the second document to the second information processing device; and a function of obtaining the correspondence by sending the prompt including the first document and the second document to the second information processing device.

[0018] In the above, the first information processing device preferably has a function of receiving a selection of whether to allow the second document to include a free translation, and a function of transmitting a prompt including an instruction for performing a synonym determination between the first document and the second document selected to allow the second document to include a free translation, and the first document and the second document to the second information processing device, thereby obtaining a result of the synonym determination.Furthermore, the first information processing device preferably has a function of receiving a correction of the second document when an error is found in the synonym determination between the second document and the first document.

[0019] Alternatively, in the above, it is preferable that the first information processing device has a function of accepting corrections to a second document that has been determined to contain a free translation, and repeating the process of accepting corrections, sending prompts for determination, and obtaining the results of the determination until it is determined that the second document does not contain a free translation.

[0020] Alternatively, in the above, it is preferable that the first information processing device has a function of accepting corrections to the second document when there is an error in the correspondence between words contained in the second document and words contained in the first document.

[0021] Alternatively, one aspect of the present invention is an information processing system including a first information processing device and a second information processing device, wherein the first information processing device has the following functions: a function of accepting input of a first document in a first language and a second document in a second language that is a translation of the first document; a function of generating a literal translation of the first document in the second language; a function of generating a free translation of the first document in the second language; a function of calculating a similarity between the second document and the literal translation; a function of calculating a similarity between the second document and the free translation; a function of comparing the similarity of the literal translation with the similarity of the free translation, and determining that the second document does not contain a free translation of the first document if the similarity of the literal translation is higher; an instruction statement for outputting a correspondence between words included in the second document that has been determined not to contain a free translation and words included in the first document; and a function of acquiring the correspondence by sending a prompt including the first document and the second document to the second information processing device.

[0022] Alternatively, one aspect of the present invention is an information processing method comprising the steps of: accepting input of a first document in a first language and a second document in a second language that is a translation of the first document; creating a first prompt including an instruction statement for determining whether the second document contains a free translation of the first document and the first document; making a determination in accordance with the first prompt; if it is determined that the second document does not contain a free translation, creating a second prompt including an instruction statement for outputting a correspondence between words contained in the second document and words contained in the first document, the first document, and the second document; and outputting the correspondence in accordance with the second prompt.

[0023] In the above information processing method, it is preferable to have the following steps: accepting a selection of whether to allow the second document to include a free translation; an instruction sentence for performing a synonym determination between the first document and the second document selected as allowing the second document to include a free translation; creating a third prompt including the first document and the second document; and performing a synonym determination in accordance with the third prompt.

[0024] Alternatively, in the above information processing method, it is preferable that the determination is performed by a large-scale language model and the correspondence relationship is output by the large-scale language model. Also, it is preferable that the determination is performed by a large-scale language model, the correspondence relationship is output by the large-scale language model, and the synonym determination is performed by the large-scale language model.

[0025] According to one aspect of the present invention, it is possible to provide an information processing system capable of efficiently checking and correcting a translation. Alternatively, according to one aspect of the present invention, it is possible to provide an information processing system capable of determining whether a translation contains a free translation. Alternatively, according to one aspect of the present invention, it is possible to provide an information processing system capable of determining whether an original text and a translation are synonymous when a translation contains a free translation. Alternatively, according to one aspect of the present invention, it is possible to provide an information processing system capable of outputting, when a translation is a literal translation, a correspondence between a word or phrase included in an original text and a corresponding word or phrase in a translation. Alternatively, it is possible to provide an information processing method used in any one or more of the above information processing systems. Alternatively, it is possible to provide an information processing device including any one or more of the above information processing systems.

[0026] According to one aspect of the present invention, a novel information processing system with excellent convenience, usefulness, or reliability can be provided. Alternatively, according to one aspect of the present invention, a novel information processing method with excellent convenience, usefulness, or reliability can be provided. Also, a novel information processing system can be provided. Alternatively, a novel information processing method can be provided.

[0027] Note that the description of these effects does not preclude the existence of other effects. Note that one embodiment of the present invention does not necessarily have all of these effects. Note that effects other than these will become apparent from the description in the specification, drawings, claims, etc., and it is possible to extract other effects from the description in the specification, drawings, claims, etc.

[0028] Fig. 1 is a diagram illustrating the configuration of an information processing system. Fig. 2 is a diagram illustrating an information processing method. Fig. 3 is a diagram illustrating the information processing method. Fig. 4 is a diagram illustrating the information processing method. Fig. 5 is a diagram illustrating the configuration of an information processing system. Fig. 6 is a diagram illustrating the configuration of an information processing device. Fig. 7A is a diagram illustrating the information processing method. Fig. 7B is a diagram illustrating the configuration of an information processing system.

[0029] An information processing system according to one embodiment of the present invention is an information processing system that can be used to compare an original text in a certain natural language with a translation into another natural language to check whether the translation has been performed correctly. In other words, the information processing system according to one embodiment of the present invention is an information processing system that can assist a translator in checking the translation. Note that the translation is not limited to a translation obtained by machine translation, but also includes a translation obtained by manual translation (also called human translation).

[0030] Furthermore, the information processing system of one embodiment of the present invention has a check function that can flexibly handle both literal translation and free translation. That is, it has a function of performing different information processing depending on whether a literal translation is required or a free translation is permitted. Note that the information processing system of one embodiment of the present invention is preferably an information processing system that has a function that can handle both a case where a literal translation is required and a case where a free translation is permitted, but may also be an information processing system that has a function that can handle either one of the cases.

[0031] By using an information processing system according to one embodiment of the present invention, when a literal translation is required (when free translation is not permitted), if a translated text contains a free translation, the free translation can be detected and the user of the information processing system can be prompted to correct it. Furthermore, by using the information processing system according to one embodiment of the present invention, a translation determined to not contain a free translation (or a translation corrected until it is determined to not contain a free translation) can be compared with the original text, and the correspondence between the words and phrases contained in each document can be presented to the user of the information processing system, making it easier to find mistranslations such as missing information or excessive information, and allowing the mistranslations to be efficiently corrected.

[0032] Alternatively, by using the information processing system of one embodiment of the present invention, it is possible to determine whether a translated sentence is synonymous with an original sentence when free translation is permitted. Furthermore, by presenting the result of the synonymity determination to a user of the information processing system, it is possible to prevent a non-synonymous translated sentence (mistranslation) from being overlooked, and the mistranslation can be efficiently corrected.

[0033] As an example, an information processing system according to one embodiment of the present invention has a function of accepting input of a document in a first language (original text) and a document in a second language (translation text) obtained by translating the original text into a second language when a literal translation is required. The information processing system according to one embodiment of the present invention also has a function of determining whether the input translation text contains a free translation of the original text, and if the determination indicates that the free translation is not included (i.e., the translation text is a literal translation), a function of outputting a correspondence between words included in the original text and words included in the translation text. A user of the information processing system can check whether the translation is correct by confirming the correspondence. Furthermore, if the input translation text is determined to contain a free translation, the user of the information processing system can correct the translation. Furthermore, if the correspondence contains an error or discrepancy, the user of the information processing system can correct the translation.

[0034] Furthermore, an information processing system according to one embodiment of the present invention has a function of accepting input of a document in a first language (original text) and a document in a second language (translated text) obtained by translating the original text into a second language, when free translation is permitted. The information processing system also has a function of determining whether the translated text is synonymous with the original text (synonymous determination). The information processing system also has a function of outputting the results of the synonymous determination. A user of the information processing system can check whether the translation has been performed correctly by checking the results of the synonymous determination. Furthermore, a user of the information processing system can correct the translated text if the input translated text contains a portion that is not synonymous with the original text.

[0035] An information processing system according to one aspect of the present invention has a function of repeatedly performing the above series of processes from input to output in units of one sentence, two or more sentences, one paragraph, two or more paragraphs, etc. The user can check the original text, the translation, and the word correspondence or the synonym determination results as a set and make corrections as necessary, but this checking and selection may be performed for each of the above repetition units, for multiple repetition units, or after the entire processing is completed.

[0036] In addition, terms using ordinal numbers such as first, second, and third used in this specification are used for convenience in identifying components and do not limit the number.

[0037] The embodiments will be described in detail with reference to the drawings. However, the present invention is not limited to the following description, and it will be readily understood by those skilled in the art that various changes in form and details can be made without departing from the spirit and scope of the present invention. Therefore, the present invention should not be interpreted as being limited to the description of the embodiments shown below. In the configuration of the invention described below, the same parts or parts having similar functions will be denoted by the same reference numerals in different drawings, and repeated explanations will be omitted.

[0038] In the drawings accompanying this specification, components are classified by function and shown as independent blocks in block diagrams, but in reality, it is difficult to completely separate components by function, and one component may be involved in multiple functions.

[0039] Embodiment 1 In this embodiment, an information processing system and an information processing method according to one embodiment of the present invention will be described with reference to FIGS.

[0040] FIG. 1 is a diagram illustrating a configuration of an information processing system according to one embodiment of the present invention.

[0041] 1, the information processing system described in this embodiment includes an information processing device 10 and an information processing device 40. The information processing device 10 and the information processing device 40 are connected via a network 30, and can transmit and receive document data and the like.

[0042] Furthermore, the information processing system described in this embodiment may be configured so that a user can input documents, etc. by directly operating the information processing device 10, or may be configured so that a user can input documents, etc. by using an information terminal 20 connected to the information processing device 10 via a network 31, as shown in Figure 1.

[0043] An example of an information processing method used in the information processing system described in this embodiment will be described with reference to FIGS.

[0044] [Information Processing Method] FIGS. 2 to 4 are flowcharts illustrating an example of an information processing method according to one embodiment of the present invention.

[0045] An information processing method according to one aspect of the present invention is "started," and in step S101 of FIG. 2, a first document (original text) expressed in a first language and a second document (translated text) obtained by translating the first document into a second language are input.

[0046] Next, in step S111 of Fig. 2, a selection is made as to whether or not the translation (second document) is allowed to include a free translation. Note that this selection may be made by the user of this information processing system each time, or may be input in advance in step S101 along with the original text (first document) and the translation (second document). In this way, the selection of whether or not to allow the translation (second document) to include a free translation is accepted.

[0047] In step S111 of FIG. 2, if it is selected that the translation (second document) may include a free translation (if the inclusion of a free translation is permitted), the process proceeds from step S111 to connector A, which connects to the flowchart shown in FIG. 4.

[0048] Alternatively, if it is selected in step S111 of FIG. 2 that the translation (second document) should not contain a free translation (if a literal translation is required), the process proceeds to step S121.

[0049] <Flow when a Literal Translation is Required> Next, in step S121 of FIG. 2, it is determined whether the translation (second document) contains a free translation.

[0050] 3, step S121A including the processes of steps S201 to S204 will be described as an example of the process in step S121 in FIG.

[0051] As the first process of step S121A, a prompt is created in step S201 of FIG. 3 to determine whether the translation (second document) contains a free translation of the original text (first document).

[0052] The prompt for determining whether a free translation is included includes an original text (first document), a translation (second document), and an instruction such as, "Please point out any free translations in the original text and translation below." The prompt can be considered an input sentence that causes the language model to perform a desired operation. When a prompt is given to the language model, the language model generates a response content based on the prompt.

[0053] The prompt created in step S201 is sent to the large-scale language model in step S202, and the large-scale language model can determine whether the translation (second document) contains a free translation of the original text (first document) in step S203. Thereafter, in step S204, the determination result output from the large-scale language model is received, and the process can proceed to step S122 in FIG. 2.

[0054] In this way, the information processing system can determine whether the translation (second document) contains a free translation.

[0055] Next, in step S122 of FIG. 2 , the determination result obtained in step S121 is presented to the user of the information processing system. If it is determined that the translation (second document) contains a free translation, the portion determined to be a free translation can be presented to the user. In step S123, the user corrects the portion of the translation (second document) determined to be a free translation. The information processing system accepts the corrected translation (second document). Note that the above correction can be repeated until it is determined that the translation (second document or the second document corrected in step S123) does not contain a free translation.

[0056] If it is determined in step S121 that the translation does not include a free translation, the process proceeds to step S131.

[0057] Next, in step S131 of FIG. 2, a prompt is created to determine the correspondence (correspondence determination) between the original text (first document) and the translation (second document or the second document corrected in step S123).

[0058] The correspondence determination prompt includes an original text (first document), a translation (second document or the second document corrected in step S123), and an instruction such as "The following original text and translation are literal translations. Please create a correspondence table for each word." By sending this prompt to the large-scale language model, the correspondence between the original text (first document) and the translation (second document or the second document corrected in step S123) can be output from the large-scale language model.

[0059] 2, the large-scale language model determines the correspondence between the original text (first document) and the translation (second document or the second document corrected in step S123). The large-scale language model also outputs, as a result of the determination of the correspondence, for example, a correspondence table.

[0060] Furthermore, if there is a difference in the correspondence between the original text (first document) and the translation (second document or the second document corrected in step S123), an instruction may be created to explain the difference. In this case, for example, the instruction included in the prompt may be something like, "The following original text and translation are literal translations. Please create a correspondence table for each word. Then, please report any differences in the correspondence from the correspondence table."

[0061] Examples of the above prompts and their results are shown in Tables 1 to 9. Tables 1 to 3 are examples where there is no difference in the correspondence between the original text and the translation, Tables 4 to 6 are first examples where there is a difference in the correspondence between the original text and the translation, and Tables 7 to 9 are second examples where there is a difference in the correspondence between the original text and the translation. Note that in the first example, there is a omission (a mouse) in the translation, and in the second example, there is an error in the numbers (codes) in the translation (correct: 101, incorrect: 102).

[0062] Tables 1, 4, and 7 are examples of prompts to be input to the large-scale language model, Tables 2, 5, and 8 are examples of correspondence tables output from the large-scale language model, and Tables 3, 6, and 9 are examples of explanations of correspondences output from the large-scale language model.

[0063]

[0064]

[0065]

[0066]

[0067]

[0068]

[0069]

[0070]

[0071]

[0072] Next, in step S133 of FIG. 2, a choice is made as to whether to revise the translation (the second document or the second document revised in step S123).

[0073] If the selection in step S133 indicates that the correspondence relationships output from the large-scale language model in step S132 do not contain any differences, as in the examples of Tables 1 to 3, the information processing method of one aspect of the present invention can be "ended." Alternatively, if the selection in step S133 indicates that the correspondence relationships output from the large-scale language model in step S132 contain differences, as in the examples of Tables 4 to 9, the process proceeds to step S134, where the translated sentence is corrected. Note that the above correction can be repeated until it is determined that the correspondence relationships between the original sentence and the translated sentence do not contain any differences.

[0074] The selection in step S133 can also be made by the user of this information processing system by referring to the correspondence output from the large-scale language model in step S132.

[0075] <Flow when free translation is allowed> A case where it is selected in step S111 that free translation may be included in the translation (second document) (when free translation is allowed) will be described with reference to FIG. 4 .

[0076] In step S141 connected from connector A shown in FIG. 4, a prompt is created to determine whether the original text (first document) and the translation (second document) are synonymous.

[0077] The synonym determination prompt includes an original text (first document), a translation (second document), and an instruction such as, "Please indicate in a table the parts of the original text and translation below that are free translations. Then, please indicate whether the meanings of the original text and the translation are the same by selecting "synonymous" or "not synonymous." By sending this prompt to a large-scale language model, the large-scale language model can determine whether the original text (first document) and the translation (second document) are synonymous and output the determination result.

[0078] For the synonym determination, a machine learning model can be used, which can be a fine-tuned neural network model, BERT (Bidirectional Encoder Representation from Transformers).

[0079] When a machine learning model is used for synonym determination, an entailment recognition task may be used, replacing "entailed" and "not entailed" with "synonymous" and "not synonymous." For example, two sentences may be compared, and learning may be performed by labeling sentences with equivalent information as synonymous and sentences with unequal information as not synonymous. Furthermore, for example, a first document and a second document may be compared, and learning may be performed by labeling the information in the second document with respect to the first document as being insufficient, excessive, equivalent (synonymous), or contradictory.

[0080] Alternatively, the synonymity determination may be performed by comparing sentence vectors and evaluating the cosine similarity of the vectors. Similarity values ​​above a certain value may be determined to be synonymous, and values ​​below the certain value may be determined not to be synonymous.

[0081] Alternatively, LLM can be used for synonym determination. When LLM is used for synonym determination, a method may be used in which, instead of a synonymity value, a binary output is provided: whether the first document and the second document are synonymous or not. Furthermore, in synonym determination using LLM, if the information contained in the second document is excessive or missing compared to the information contained in the first document, that information may be added and output together with the determination. Note that in synonym determination using LLM, the first document and the second document may be compared, and learning may be performed by labeling the information in the second document as insufficient, excessive, equivalent (synonymous), or contradictory to the information in the first document.

[0082] Alternatively, a combination of the cosine similarity of vectors and a synonym determination label may be used for synonym determination. For example, the cosine similarity may be converted into a cosine distance, and an overabundance label may be set to 1, an underabundance label to -1, a synonymous label to 0, and a contradictory label to nan, and the label evaluation value may be calculated by multiplying the cosine distance by the label evaluation value. In this case, the closer to 0 the value is, the more synonymous it is, and the excess or deficiency of information can also be expressed. This is more effective than simply expressing the similarity, as it allows evaluation of the excess or deficiency of information in cases where it is particularly undesirable to add extra information that is not in the original text.

[0083] 4, the synonymity determination result obtained in step S142 is presented to the user of the information processing system, and the user selects whether to revise the translation (second document). The information processing system then accepts the selection.

[0084] If the synonymy determination results output from the large-scale language model in step S142 do not include a determination result of "not synonymous," the information processing method according to one embodiment of the present invention can be "ended." Alternatively, if the synonymy determination results output from the large-scale language model in step S142 include a determination result of "not synonymous," the user can be presented with the parts determined to be not synonymous, and in step S144, the user corrects the parts of the translated text (second document) determined to be not synonymous. The information processing system accepts the corrected translated text (second document). Note that this correction can be repeated until the results of the synonymy determination between the original text and the translated text no longer include a result determined to be not synonymous.

[0085] The selection in step S143 can also be made by the user of this information processing system by referring to the synonymity determination result output from the large-scale language model in step S142.

[0086] 2 and 4, it is possible to proceed to input processing of the next original sentence before selecting "end." The above processing can be repeated until processing of all original sentences is completed.

[0087] In the information processing method of one embodiment of the present invention described using Figures 2 to 4, which device or terminal, information processing device 10, information processing device 40, or information terminal 20, performs the processing of each step will be described using Figure 5.

[0088] 5 is a relationship diagram showing the processes performed by the information processing device 10, the information processing device 40, and the information terminal 20, using the symbols for each step used in Fig. 2 to Fig. 4. In addition, the block arrows labeled with network 30 indicate data transmission and reception between the information processing device 10 and the information processing device 40, and the block arrows labeled with network 31 indicate data transmission and reception between the information processing device 10 and the information terminal 20.

[0089] As shown in FIG. 5, the information processing device 10 has a function to perform processing of step S201, a function to perform processing of step S202, a function to perform processing of step S204, a function to perform processing of step S122, a function to perform processing of step S131, and a function to perform processing of step S141.

[0090] As shown in FIG. 5, the information processing device 40 has a function to perform the process of step S203, a function to perform the process of step S132, and a function to perform the process of step S142.

[0091] Also, as shown in FIG. 5, the information terminal 20 has a function to perform processing of step S101, a function to perform processing of step S111, a function to perform processing of step S123, a function to perform processing of step S133, a function to perform processing of step S134, a function to perform processing of step S143, and a function to perform processing of step S144.

[0092] An example of the configuration of the information processing device 10, the information processing device 40, the information terminal 20, the network 30, and the network 31 will be described below.

[0093] <<Configuration Example of Information Processing Device 10>> A configuration example of the information processing device 10 included in the information processing system of one embodiment of the present invention will be described with reference to FIG. 6 .

[0094] As shown in FIG. 6, the information processing device 10 includes an input unit 110 , a storage unit 120 , a processing unit 130 , an output unit 140 , and a transmission path 150 .

[0095] [Input Unit 110] The input unit 110 can accept data from outside the information processing device 10. For example, the input unit 110 can accept data from the information terminal 20. The input unit 110 can also accept data from the information processing device 40. Specifically, a device such as a wired communication port, a wireless communication port, or an optical communication port can be used.

[0096] The input unit 110 can supply the received data to one or both of the storage unit 120 and the processing unit 130 via the transmission path 150 .

[0097] [Storage Unit 120] The storage unit 120 has a function of storing a program executed by the processing unit 130. The storage unit 120 may also have a function of storing data generated by the processing unit 130 (e.g., calculation results, analysis results, inference results), data accepted by the input unit 110, and the like.

[0098] The storage unit 120 may have a database. Furthermore, the information processing device 10 may have a database separate from the storage unit 120. The information processing device 10 may have a function to retrieve data from a database that exists outside the storage unit 120, outside the information processing device 10, or outside the information processing system. Furthermore, the information processing device 10 may have a function to retrieve data from both its own database and an external database.

[0099] Either or both of a storage and a file server can be used as the memory unit 120. Also, a database that records paths of files stored in a file server can be used as the memory unit 120.

[0100] The storage unit 120 includes at least one of a volatile memory and a non-volatile memory. Examples of the volatile memory include a dynamic random access memory (DRAM) and a static random access memory (SRAM). Examples of the non-volatile memory include a resistive random access memory (ReRAM), a phase change random access memory (PRAM), a ferroelectric random access memory (FeRAM), a magnetoresistive random access memory (MRAM), and a flash memory. The storage unit 120 may also include at least one of NOSRAM (registered trademark) and DOSRAM (registered trademark). The storage unit 120 may also include a recording media drive. Examples of recording media drives include a hard disk drive (HDD) and a solid state drive (SSD).

[0101] NOSRAM is an abbreviation for "Nonvolatile Oxide Semiconductor Random Access Memory." NOSRAM refers to a memory in which memory cells are two-transistor (2T) or three-transistor (3T) gain cells, and the transistors are transistors (also called OS transistors) that use metal oxide in their channel formation regions. OS transistors have an extremely small leakage current, i.e., a current that flows between the source and drain in an off state. NOSRAM can be used as a nonvolatile memory by retaining a charge corresponding to data in the memory cell using its extremely small leakage current characteristic. In particular, NOSRAM can read stored data without destroying it (nondestructive read), making it suitable for arithmetic processing in which only data read operations are repeated a large number of times. NOSRAM can increase its data capacity by stacking layers, and therefore can be used as a large-scale cache memory, main memory, or storage memory to improve the performance of semiconductor devices.

[0102] DOSRAM is an abbreviation for "Dynamic Oxide Semiconductor RAM" and refers to a RAM having 1T (transistor) 1C (capacitor) type memory cells. DOSRAM is a DRAM formed using OS transistors, and is a memory that temporarily stores information sent from an external device. DOSRAM is a memory that takes advantage of the low off-state current of OS transistors.

[0103] In this specification and the like, a metal oxide refers to an oxide of a metal in a broad sense. Metal oxides are classified into oxide insulators, oxide conductors (including transparent oxide conductors), oxide semiconductors (also referred to as oxide semiconductors or simply as OSs), and the like. For example, when a metal oxide is used for a semiconductor layer of a transistor, the metal oxide may be referred to as an oxide semiconductor.

[0104] The metal oxide included in the channel formation region preferably contains indium (In). When the metal oxide included in the channel formation region contains indium, the carrier mobility (electron mobility) of the OS transistor is increased. Furthermore, the metal oxide included in the channel formation region is preferably an oxide semiconductor containing element M. The element M is preferably at least one of aluminum (Al), gallium (Ga), and tin (Sn). Other elements applicable to element M include boron (B), silicon (Si), titanium (Ti), iron (Fe), nickel (Ni), germanium (Ge), yttrium (Y), zirconium (Zr), molybdenum (Mo), lanthanum (La), cerium (Ce), neodymium (Nd), hafnium (Hf), tantalum (Ta), and tungsten (W). However, a combination of two or more of the above elements may be used as element M. The element M is, for example, an element having a high bond energy with oxygen. For example, it is an element having a higher bond energy with oxygen than indium. The metal oxide contained in the channel formation region is preferably a metal oxide containing zinc (Zn), since zinc-containing metal oxides may be easily crystallized.

[0105] The metal oxide contained in the channel formation region is not limited to a metal oxide containing indium, but may be, for example, a metal oxide containing zinc but not indium, such as zinc tin oxide or gallium tin oxide, a metal oxide containing gallium, or a metal oxide containing tin.

[0106] [Processing Unit 130] The processing unit 130 has a function of performing processes such as calculation, analysis, and inference using data supplied from one or both of the input unit 110 and the storage unit 120. The processing unit 130 can supply generated data (e.g., calculation results, analysis results, and inference results) to one or both of the storage unit 120 and the output unit 140.

[0107] The processing unit 130 has a function of acquiring data from the storage unit 120. The processing unit 130 may also have a function of recording or registering data in the storage unit 120.

[0108] The processing unit 130 may include, for example, an arithmetic circuit, a central processing unit (CPU), and a graphics processing unit (GPU).

[0109] The processing unit 130 may have a microprocessor such as a DSP (Digital Signal Processor). The microprocessor may be implemented by a PLD (Programmable Logic Device) such as an FPGA (Field Programmable Gate Array) or an FPAA (Field Programmable Analog Array). The processing unit 130 may also have a quantum processor. The processing unit 130 can perform various data processing and program control by interpreting and executing instructions from various programs using the processor. Programs that can be executed by the processor are stored in at least one of the memory area of ​​the processor and the storage unit 120.

[0110] The processing unit 130 may include a main memory. The main memory may include at least one of a volatile memory such as a random access memory (RAM) and a non-volatile memory such as a read-only memory (ROM). The main memory may also include at least one of the above-mentioned NOSRAM and DOSRAM.

[0111] The RAM may be, for example, a DRAM or an SRAM, and a virtual memory space is allocated and used as a working space for the processing unit 130. The operating system, application programs, program modules, program data, lookup tables, and the like stored in the storage unit 120 are loaded into the RAM for execution. The data, programs, and program modules loaded into the RAM are each directly accessed and operated by the processing unit 130.

[0112] The ROM can store a BIOS (Basic Input / Output System), firmware, etc., which do not require rewriting. Examples of ROM include mask ROM, OTPROM (One Time Programmable Read Only Memory), and EPROM (Erasable Programmable Read Only Memory). Examples of EPROMs include UV-EPROMs (Ultra-Violet Erasable Programmable Read Only Memories), which allow stored data to be erased by exposure to ultraviolet light, EEPROMs (Electrically Erasable Programmable Read Only Memories), and flash memories.

[0113] The processing section 130 can include one or both of an OS transistor and a transistor having silicon in a channel formation region (a Si transistor).

[0114] The processing unit 130 preferably includes an OS transistor. Because an OS transistor has an extremely small off-state current, using the OS transistor as a switch for retaining charge (data) flowing into a capacitor functioning as a memory element can ensure a long data retention period. By using this characteristic in at least one of the register and cache memory of the processing unit, the processing unit can be operated only when necessary, and can be turned off in other cases by saving information from the previous processing in the memory element. In other words, normally-off computing is possible, and the power consumption of the information processing system can be reduced.

[0115] It is preferable that the information processing device 10 uses AI for at least some of its processing.

[0116] In particular, it is preferable that the information processing device 10 uses an artificial neural network, which is realized by a circuit (hardware) or a program (software).

[0117] In this specification, a neural network refers to a general model that mimics the neural circuit network of a living organism, determines the connection strength between neurons through learning, and has problem-solving capabilities. A neural network has an input layer, an intermediate layer (hidden layer), and an output layer.

[0118] In this specification and the like, when discussing neural networks, determining the connection strengths (also called weighting coefficients) between neurons from existing information may be referred to as "learning."

[0119] In this specification and the like, the act of constructing a neural network using connection strengths obtained by learning and deriving a new conclusion from it may be referred to as "inference."

[0120] The information processing device 10 can perform processing using a natural language processing model that uses AI. For example, it can perform machine translation processing using a natural language processing model that uses AI. As a natural language processing model that performs machine translation processing, it is possible to execute processing using a natural language processing model that uses AI, such as Sequence-to-Sequence (seq2seq), Transformer, BERT (Bidirectional Encoder Representations from Transformers), or T5 (Text-to-Text Transformer).

[0121] [Output Unit 140] The output unit 140 can output at least one of the calculation results, analysis results, and inference results in the processing unit 130 to the outside of the information processing device 10. Specifically, a device such as a wired communication port, a wireless communication port, or an optical communication port can be used.

[0122] For example, the output unit 140 can transmit data to the information processing device 40. The output unit 140 can also transmit data to the information terminal 20.

[0123] [Transmission Path 150] The transmission path 150 has a function of transmitting data. Data can be transmitted and received between the input unit 110, the storage unit 120, the processing unit 130, and the output unit 140 via the transmission path 150. Specifically, a bus line on a motherboard, a wired communication cable, or an optical communication cable can be used.

[0124] <<Configuration Example of Information Processing Device 40>> The information processing device 40 can process received data and transmit the processing results. For example, the information processing device 40 can perform processing such as calculations using data received from the information processing device 10. Furthermore, the information processing device 40 can transmit the processing results to the information processing device 10. This can reduce the calculation load on the information processing device 10.

[0125] The information processing device 40 can perform processing using a natural language processing model that uses AI. For example, the information processing device 40 can execute processing using a natural language processing model that uses AI, such as BERT (Bidirectional Encoder Representations from Transformers) or T5 (Text-to-Text Transformer).

[0126] Furthermore, the information processing device 40 can perform processing using a model (such as a sentence generation model or a dialogue model) that utilizes a large-scale language model. The determination of the correspondence between the words and phrases included in the original sentence and the words and phrases included in the translation in step S132 described in FIG. 2 is preferably performed using a model that utilizes a large-scale language model. The determination of whether the translation contains a free translation of the original sentence in step S203 described in FIG. 3 is preferably performed using a model that utilizes a large-scale language model. The determination of whether the original sentence and the translation are synonymous in step S142 described in FIG. 4 is preferably performed using a model that utilizes a large-scale language model. For example, processing can be performed using large-scale language models such as GPT-3, GPT-3.5, GPT-4 (registered trademark), LaMDA (Language Model for Dialogue Applications), PaLM (Pathways Language Model), Llama2, and Llama3. In particular, it is preferable to use GPT-4 (registered trademark).

[0127] For example, the information processing device 40 can translate a document written in language A into a document written in language B. Furthermore, the information processing device 40 can translate a document written in language A into a document written in language B in accordance with a given directive. Furthermore, constraints can be included in the directives. This makes it possible to control the degree of freedom in translation by giving constraints.

[0128] Furthermore, the information processing device 40 can execute processing using a general-purpose language processing model that can perform various natural language processing tasks.

[0129] The information processing device 40 is a large-scale computer such as a server computer or a supercomputer. Preferably, the information processing device 40 has a function as a parallel computer. By using the information processing device 40 as a parallel computer, it is possible to perform large-scale calculations required for AI learning and inference, for example.

[0130] The information processing device 40 is a computer with higher processing power than the information processing device 10. For example, if both the information processing device 10 and the information processing device 40 have functions as parallel computers, the information processing device 40 has higher processing power than the information processing device 10 and can perform large-scale calculations. Also, for example, if both the information processing device 10 and the information processing device 40 can perform processing using a model that utilizes a large-scale language model, the information processing device 40 can execute processing using a large-scale AI model compared to the information processing device 10.

[0131] It should be noted that the service provider does not necessarily have to own the information processing device 40. For example, the service provider can use part of the service provided by another business or the like using the information processing device 40.

[0132] <<Configuration Example of Information Terminal 20>> The information terminal 20 can accept data input by a user of the information processing system of one aspect of the present invention. The information terminal 20 can also provide the user with data output by the information processing system of one aspect of the present invention.

[0133] Furthermore, the information terminal 20 can transmit data received from the user to the information processing device 10. Furthermore, the information terminal 20 can provide data received from the information processing device 10 to the user.

[0134] Furthermore, the information terminal 20 can transmit data generated based on data received from the user to the information processing device 10. Furthermore, the information terminal 20 can provide data generated based on data received from the information processing device 10 to the user.

[0135] For example, dedicated application software or a web browser is installed on the information terminal 20. A user can access the information processing device 10 via either of these. This allows the user to enjoy services using the information processing system according to one aspect of the present invention, for example, using a computer with lower processing power than the information processing device 10.

[0136] The information terminal 20 may also be called a client computer, etc. Each of the information terminals 20 is an information terminal device used by a user of the information processing system according to one aspect of the present invention.

[0137] For example, a desktop computer 20a, a notebook computer 20b, a smartphone 20c, and a tablet computer 20d can be used as the information terminal 20. The tablet computer 20d can also be used as a notebook computer by connecting it to a housing 21 having a keyboard.

[0138] This allows a user of the information processing system to use a computer, such as an information terminal 20, which has lower processing power than the information processing device 10 or the information processing device 40, to control an information processing system according to one embodiment of the present invention, give instructions to the information processing device 40, and enjoy services.

[0139] <<Network 30>> The network 30 connects the information processing device 10 and the information processing device 40. This allows input data and processed data to be transmitted and received between the two. It also allows the load related to information processing to be distributed.

[0140] In this embodiment, the network 30 is mainly described as a computer network that is larger in scale than the network 31. For example, a global network can be used for the network 30. Specifically, the Internet, which is the foundation of the World Wide Web (WWW), can be used.

[0141] <<Network 31>> The network 31 connects a plurality of information terminals 20 and information processing devices 10. This allows data to be transmitted and received between them. It also allows the load related to information processing to be distributed. Furthermore, a service provider can provide a service using an information processing method according to one aspect of the present invention to a user via the network 31, for example.

[0142] For example, a local network can be used for the network 31. Also, an intranet or an extranet can be used for the network 31. Also, a personal area network (PAN), a local area network (LAN), a campus area network (CAN), a metropolitan area network (MAN), a wide area network (WAN), a global area network (GAN), etc. can be used for the network 31.

[0143] When a provider of a service using an information processing method according to one embodiment of the present invention and a user who receives the service belong to the same organization, such as a company, it is preferable that data be transmitted and received between the information terminal 20 and the information processing device 10 using, for example, a network 31 established within the organization. This allows data to be transmitted and received between the information terminal 20 and the information processing device 40 more securely than when data is transmitted over the Internet. Furthermore, confidential information within the organization can be prevented from leaking to the outside.

[0144] When wireless communication is performed, communication standards such as the fourth generation mobile communication system (4G), fifth generation mobile communication system (5G), and sixth generation mobile communication system (6G), or specifications standardized by the IEEE such as Wi-Fi (registered trademark) and Bluetooth (registered trademark), can be used as communication protocols or communication technologies.

[0145] Embodiment 2 In this embodiment, an information processing system and an information processing method according to one embodiment of the present invention will be described. The information processing system has a configuration example that is partially different from the configuration example 1 of the information processing system described in Embodiment 1.

[0146] 7A and 7B , an example of an information processing system different from the information processing system configuration example 1 will be described. Here, different parts will be described in detail, and for parts having the same configuration, the description of embodiment 1 will be referred to.

[0147] In the information processing system described here, the processing content of step S121 in Fig. 2 is different from the flow described in Fig. 3 of configuration example 1 of the information processing system. Specifically, the processing content of step S121 in Fig. 2 is performed using step S121B, which includes processing of steps S301 to S304 shown in Fig. 7A. The processing of steps S301 to S304 will be described using Fig. 7A.

[0148] As the first process of step S121B, in step S301 of FIG. 7A, the first document (original text) is literally translated into the second language to generate a third document (literal translation).

[0149] 7A, the first document (original text) is translated into the second language so as to include a free translation, thereby generating a fourth document (freely translated text). Note that the processing of steps S301 and S302 may be performed in such a way that step S302 is performed first and then step S301 is performed, or steps S301 and S302 may be performed in parallel.

[0150] It is preferable to use a dedicated translation model trained on a dataset consisting of literal translations for generating literal translations (step S301), and a dedicated translation model trained on a dataset consisting of free translations for generating free translations (step S302). This reduces the amount of calculation without requiring the creation of prompts, and because the sentences are generated using different models, different sentences can be generated for literal translations and free translations. The natural language processing model described in embodiment 1 can be used as the translation model.

[0151] Next, in step S303 of FIG. 7A, the similarity between the second document (translation) input in step S101 of FIG. 2 and each of the third document (literal translation) and the fourth document (free translation) is calculated.

[0152] To calculate the similarity, the sentences can be vectorized and the cosine similarity can be used. SentenceBERT or the like can be used to vectorize the sentences. The cosine similarity is expressed as a real number between -1 and 1, and the closer the value is to 1, the more similar the two documents being compared are (high similarity), and the closer the cosine similarity value is to -1, the more dissimilar they are (low similarity).

[0153] Alternatively, an edit distance such as the Levenshtein distance or the Jaro-Winkler distance may be used instead of the similarity. When the edit distance is used instead of the similarity, the closer the value is to 0, the more similar the two documents are to be compared (high similarity), and the larger the edit distance value, the more dissimilar the documents are to each other (low similarity).

[0154] Alternatively, a BLEU (Bilingual Evaluation Understudy) score may be used instead of the similarity. The BLEU score is expressed as a real number between 0 and 1, and the closer the value is to 1, the more similar the two documents being compared are (high similarity), and the closer the BLEU score is to 0, the more dissimilar the documents are (low similarity).

[0155] Next, in step S304 of FIG. 7A, the similarity between the second document (translation) and the third document (literal translation) is compared with the similarity between the second document (translation) and the fourth document (free translation). If the comparison result shows that the similarity of the fourth document (free translation) is higher than the similarity of the third document (literal translation), it is determined that the second document (translation) contains a free translation. Alternatively, if the comparison result shows that the similarity of the third document (literal translation) is higher than the similarity of the fourth document (free translation), it is determined that the second document (translation) does not contain a free translation. After the determination in step S304, the process can proceed to step S122 of FIG. 2.

[0156] The above is an explanation of the differences in the information processing method from Configuration Example 1 of the information processing system described in Embodiment 1. Next, the relationship between the processes performed by the information processing device 10, the information processing device 40, and the information terminal 20 in this embodiment will be described with reference to FIG.

[0157] 7B is a relationship diagram showing the processes performed by the information processing device 10, the information processing device 40, and the information terminal 20 using the symbols for each step used in FIGS. 2, 4, and 7A, and is a different configuration example from that of FIG. 5. In addition, the block arrow labeled with network 30 indicates transmission and reception of data between the information processing device 10 and the information processing device 40, and the block arrow labeled with network 31 indicates transmission and reception of data between the information processing device 10 and the information terminal 20.

[0158] As shown in FIG. 7B, the information processing device 10 has a function to perform processing of step S301, a function to perform processing of step S302, a function to perform processing of step S303, a function to perform processing of step S304, a function to perform processing of step S122, a function to perform processing of step S131, and a function to perform processing of step S141.

[0159] As shown in FIG. 7B, the information processing device 40 has a function to perform the process of step S132 and a function to perform the process of step S142.

[0160] Also, as shown in FIG. 7B, the information terminal 20 has a function to perform processing of step S101, a function to perform processing of step S111, a function to perform processing of step S123, a function to perform processing of step S133, a function to perform processing of step S134, a function to perform processing of step S143, and a function to perform processing of step S144.

[0161] In the configuration example 1 of the information processing system of the first embodiment, "Step S121: Determine whether the translation contains a free translation" can be processed using the information processing device 10 and the information processing device 40. In contrast, in the configuration example 2 of the information processing system of the second embodiment, "Step S121: Determine whether the translation contains a free translation" can be processed only by the information processing device 10.

[0162] The information processing system, information processing method, and information processing device described above enable the work of checking and correcting translations to be carried out efficiently.

[0163] 10: Information processing device, 20a: Desktop computer, 20b: Notebook computer, 20c: Smartphone, 20d: Tablet computer, 20: Information terminal, 21: Housing, 30: Network, 31: Network, 40: Information processing device, 110: Input unit, 120: Storage unit, 130: Processing unit, 140: Output unit, 150: Transmission path

Claims

A first information processing device and a second information processing device are included, The first information processing device, accepting input of a first document in a first language and a second document in a second language that is a translation of the first document; a function of transmitting a prompt including an instruction for determining whether the second document includes a free translation of the first document, the first document, and the second document to the second information processing device, thereby obtaining a result of the determination; a function of acquiring the correspondence between a phrase included in the second document, which has been determined not to include the free translation in the determination, and a phrase included in the first document, by transmitting a prompt including the first document and the second document to the second information processing device; and a function of outputting the correspondence relationship. Information processing system. 　 In claim 1, The first information processing device, a function of accepting a selection as to whether or not to allow the second document to include a free translation; a function of transmitting, to the second information processing device, an instruction sentence for performing a synonymity determination between the second document selected in the selection as allowing the inclusion of a free translation and the first document, and a prompt including the first document and the second document, thereby acquiring a result of the synonymity determination; Information processing system. 　 In claim 1, The first information processing device, a function of accepting corrections to the second document determined to include the free translation, and repeating the acceptance of the corrections, the transmission of the prompt for the determination, and the acquisition of the result of the determination until the second document is determined to not include the free translation; Information processing system. 　 In claim 1, The first information processing device, a function of accepting a correction of the second document when there is an error in the correspondence between the words included in the second document and the words included in the first document; Information processing system. 　 In claim 2, The first information processing device, a function of accepting a correction of the second document when an error is found in the synonymity determination between the second document and the first document; Information processing system. 　 A first information processing device and a second information processing device are included, The first information processing device, accepting input of a first document in a first language and a second document in a second language that is a translation of the first document; generating a direct translation of the first document in the second language; generating a free translation of the first document in the second language; A function of calculating a similarity between the second document and the literal translation; A function of calculating a similarity between the second document and the free translation; a function of comparing a similarity between the literal translation and the free translation, and determining that the second document does not contain a free translation of the first document if the similarity between the literal translation and the free translation is higher; a function of acquiring the correspondence between a phrase included in the second document, which has been determined not to include the free translation in the determination, and a phrase included in the first document, by transmitting a prompt including the first document and the second document to the second information processing device; Information processing system. 　 accepting input of a first document in a first language and a second document in a second language that is a translation of the first document; creating a first prompt including an instruction for determining whether the second document includes a paraphrase of the first document, the first document, and the second document; making said determination in accordance with said first prompt; creating a second prompt including an instruction for outputting a correspondence between a phrase included in the second document and a phrase included in the first document, the first document, and the second document when it is determined that the free translation is not included in the second document; and outputting the correspondence relationship in accordance with the second prompt. Information processing methods. 　 In claim 7, receiving a selection as to whether or not to allow the second document to include a free translation; creating a third prompt including an instruction sentence for performing a synonymity determination between the second document selected in the selection as being allowed to include a free translation and the first document, the first document, and the second document; and performing the synonymity determination in accordance with the third prompt. Information processing methods. 　 In claim 7, The determination is made by a large-scale language model; The output of the correspondence is performed by the large-scale language model. Information processing methods. 　 In claim 8, The determination is made by a large-scale language model; The output of the correspondence is performed by the large-scale language model; The synonymity determination is performed by the large-scale language model. Information processing methods.