Translation Processing Method for Securing Publication-Grade Quality Without External Expert Intervention by Internally Performing the Entire Process of Proofreading, Translation, Review, and Evaluation According to Field-Specific Evaluation Criteria, and Blocking Output and Billing When the Criteria Are Not Met

Abstract

The present invention relates to a translation processing method that internally evaluates the quality of proofreading / translation / review results to ensure quality at the level of official documents, and does not perform output or billing if the standard is not met. This is intended to solve the problem that, until now, AI translation or expert translation has been difficult to ensure translation quality at the level of official documents without external review, and users have been charged costs even for results with substandard translation quality. The method of the present invention includes the steps of determining the field of the original text (S20), generating a translation result (S30), calculating a comprehensive evaluation value using evaluation items and weights set differently according to the field (S40), and, if the value falls below a standard value, notifying the user without outputting or charging and re-performing upon the user's request (S50, S60). According to the present invention, translation time and costs are significantly reduced while maintaining a minimum quality standard without inspection by professional personnel. Furthermore, substandard results are not delivered to the user, and no costs are deducted. Translation is performed while maintaining a minimum quality standard even for multilingual pairs where it is difficult to secure professional translation personnel.

Description

Technology Field

[0001] The present invention relates to a translation processing method, and more specifically, to a method that evaluates translation results internally within a system by applying different evaluation criteria depending on the field of the source text, thereby ensuring quality at the level of official documents without the intervention of external experts, while achieving time and cost savings. It also enables smooth translation of language pairs where it is difficult to secure translation experts among multiple languages ​​while maintaining quality. Furthermore, if the evaluation result falls short of the criteria, the result is not displayed to the user, and no billing is performed. Background Technology

[0002] Previously, it was common practice to evaluate the quality of translation results retrospectively by users or separate reviewers. This presented difficulties for users due to the long translation times and high costs involved. Furthermore, there was the issue where results with inadequate translation quality were delivered directly to users, and in the case of paid services, users bore the cost for this.

[0003] Even with AI translation, while speed was fast, the quality was not sufficient for official documents, causing the inconvenience of having to deploy additional review personnel. There was a relationship where increasing processing speed led to a decline in quality, while ensuring quality required the deployment of reviewers, thereby extending the timeframe.

[0004] This problem intensifies as the complexity of the manuscript increases. This is because it is difficult to secure specialized personnel in the field, and the length of the review process increases, causing the required time to rise sharply.

[0005] Furthermore, there are areas where it is extremely difficult to secure specialized personnel capable of handling both specific language pairs and specific fields. Even if personnel with knowledge of the relevant language exist, those capable of translating original texts in a specific field to a publishable standard are a separate matter. Such specialists are few in number, and due to the lack of smooth generational transition, securing them becomes increasingly difficult over time.

[0006] On the other hand, in language pairs lacking specialized personnel, there is a problem in that there is also a lack of means to verify the quality of the output.

[0007] In the past, both AI translation and human expert translation were performed as a single process, leading to a problem where errors occurring in intermediate stages were carried over directly to the final output.

[0008] In addition, conventionally, due to limitations in the processing capacity of AI translation, it was common practice to divide the source text for processing; however, this resulted in a problem where consistency in terminology and style was not ensured between the divided sections.

[0009] Furthermore, ensuring translation quality required review by external experts, which resulted in increased costs and time.

[0010] Meanwhile, source texts subject to translation often contain information important to users, such as unpublished works, contracts, and research materials. However, conventional translation services typically store the source text and translation results in a temporary storage area during the processing stage; consequently, there is a constant risk that the content may be leaked externally or used for other purposes. The problem to be solved

[0011] The present invention aims to solve the above-mentioned problems and has the following objectives.

[0012] Evaluating the quality of translation results from various angles within the system to ensure that results falling short of standards are not delivered to users.

[0013] Ensuring that no costs are charged to the user for results that do not meet standards.

[0014] Ensuring quality tailored to the characteristics of each field by applying different evaluation criteria according to the field of the original text.

[0015] Ensuring that a minimum quality standard equivalent to that of official documents is maintained even in language pairs where external verification methods are absent.

[0016] Dividing the translation process into multiple stages and performing checks at each stage to ensure that intermediate errors do not carry over to the final output.

[0017] It completely prevents important documents from leaking externally by ensuring that user information and translation content do not remain within the system for any reason. means of solving the problem

[0018] A translation processing method according to the present invention for solving the above problem comprises the steps of: receiving a source text and determining a field corresponding to the source text (S20); generating a translation result for the source text (S30); calculating a comprehensive evaluation value for the translation result by applying a plurality of evaluation items set differently according to the determined field and a predetermined weight for each evaluation item (S40); comparing the comprehensive evaluation value with a predetermined standard value (S50); outputting the translation result and performing corresponding billing if the comprehensive evaluation value is greater than or equal to the standard value (S70); not outputting the translation result and not performing billing if the comprehensive evaluation value is less than the standard value, and notifying the user of the fact that it is less than the standard value (S60); and re-performing the step of generating the translation result (S30) upon the user's re-request.

[0019] The calculation of the above comprehensive evaluation value (S40) is characterized by being performed internally within the system without the intervention of external personnel.

[0020] The above billing (S70) is performed on a second server (200) that is physically separated from the first server (100) responsible for translation processing, and the first server (100) and the second server (200) are linked through a token (300). Effects of the invention

[0021] Since evaluation criteria corresponding to the field of the original text are applied, the quality is ensured to be suitable for use as an official document released externally, tailored to the characteristics of the relevant field, without the intervention of external experts.

[0022] Translation results that do not meet the standards are not delivered to the user.

[0023] Since no charges are made for results that do not meet the standards, the user does not bear the cost for results that do not meet the quality standards.

[0024] The minimum quality standard is maintained without the need for external expert review, and the time and cost required for translation are significantly reduced.

[0025] Breaking away from the conventional relationship where processing speed and quality conflict, the minimum quality limit is maintained while the timeline is shortened.

[0026] Even if the difficulty of the manuscript is high, only the number of reworks increases, so the total duration remains within a certain range.

[0027] Translation is performed even for language pairs where it is difficult to secure specialized personnel, and a minimum quality standard is maintained without external verification means.

[0028] Since checks are performed at each stage, intermediate errors do not carry over to the final product, and errors are eliminated through repeated checks.

[0029] Since the entire text is processed without being divided, inconsistencies in terminology and style between sections do not occur.

[0030] Since user information and translation content do not remain within the system, there is no subject for the document content to be leaked.

[0031] The present invention aims not to replace translation by professional personnel, but to provide results of a certain quality or higher that are suitable for use as external official documents disclosed in various professional fields, and to be widely utilized in areas where translation was not performed due to a lack of personnel. Brief explanation of the drawing

[0032] FIG. 1 is a flowchart of a translation processing method according to one embodiment of the present invention. FIG. 2 is a conceptual diagram showing the separate configuration of a first server (100) and a second server (200) according to an embodiment of the present invention. FIG. 3 is a conceptual diagram showing the configuration of field-specific evaluation criteria according to one embodiment of the present invention. FIG. 4 is a schematic diagram showing the main parts of a translation processing method according to one embodiment of the present invention. Specific details for implementing the invention

[0033] 1. Original text input step (S10)

[0034] The system receives the source text to be translated from the user.

[0035] 2. Field Decision Stage (S20)

[0036] The system determines the field corresponding to the input text. The field may be determined by the user's designation or based on the characteristics of the text.

[0037] 3. Translation result generation step (S30)

[0038] Generates translation results based on processing conditions corresponding to the determined field.

[0039] Translation processing is not performed as a single, all-in-one process, but rather through multiple iterative checks. These iterative checks are conducted internally within the system and are not exposed to the user. In each stage, the results of the previous stage are reorganized and verified, and the results are passed to the next stage.

[0040] The aforementioned reorganization and inspection are not a simple repetition of a single process. Steps proceed sequentially, multiple processes within a single step are executed in parallel through mutual referencing, and the process proceeds stepwise as the results of each process are reflected in other processes.

[0041] The reorganization and inspection of each of the above steps are performed according to pre-set processing standards for each field.

[0042] The above translation processing is performed on the entire text without dividing it, and each step is performed on the entire text.

[0043] The process of generating translation results sequentially includes the correction of notation errors, correction of contextual consistency, and verification of overall consistency. Conventionally, these processes were performed by separate personnel, resulting in handovers and waiting times at each stage; however, in the present invention, they are performed continuously within a single processing flow.

[0044] The entire process described above is performed internally within the system without the intervention of external professional personnel.

[0045] 4. Evaluation Phase (S40)

[0046] After all of the above multiple steps are completed, a comprehensive evaluation is performed on the final deliverable based on evaluation criteria by field.

[0047] For the generated translation result, an evaluation value is calculated for each of the multiple evaluation criteria. The aforementioned evaluation criteria may include, for example, the degree of semantic agreement with the original text, consistency of terminology, presence of omissions, grammatical accuracy, etc.

[0048] The composition of the aforementioned evaluation criteria and the weighting of each item are set differently depending on the field. This is because the magnitude of the impact on translation quality varies for each item within that field.

[0049] In one embodiment, the number of evaluation items is set to 100, and each item is assigned a different weight.

[0050] In this specification, "public document" refers to a document disclosed externally by field, and "quality of public document level" refers to a state in which the comprehensive evaluation value calculated based on evaluation items and weights corresponding to the field is equal to or greater than a predetermined standard value.

[0051] As shown in FIG. 3, the evaluation unit (400) selects evaluation criteria based on the field determination result (410), and calculates a comprehensive evaluation value by applying evaluation item groups and item-specific weights (421, 422, …, 42n) corresponding to the selected field (S40). The entire process of the evaluation unit (400) is performed within the system without the intervention of external personnel.

[0052] By performing intermediate checks and final evaluations in this manner, the occurrence of failure to meet standards at the final stage is reduced, and the number of times users have to retake the task is decreased.

[0053] 5. Comparison and branching step (S50)

[0054] The calculated comprehensive evaluation value is compared with a predetermined standard value.

[0055] If the value is greater than or equal to the threshold value (S70), the translation result is output to the user and the corresponding billing is performed.

[0056] If the threshold value is not met (S60), the user is notified, the translation result is not output, and no billing is performed. Subsequently, the user performs the translation result generation step (S30) again. If the threshold is not met even after the re-execution, the user is advised to re-execute after a short while, as there is a high possibility of excessive simultaneous usage due to a large number of users. In most cases, the output of the result is performed along with a cost deduction within the scope of performing the re-execution about once.

[0057] 6. Server Separation Configuration

[0058] As shown in FIG. 2, the first server (100) responsible for translation processing and the second server (200) responsible for billing are physically separated, and the two servers are linked through a token (300). The second server (200) does not possess information that can identify a user.

[0059] 7. Non-retention of information and translation content (S80)

[0060] The system does not retain user information, source texts for translation, or translation results after processing is complete. During the processing phase, source texts and intermediate results are maintained for re-execution, but they are removed once processing is complete and the results are delivered to the user. Subsequently, even if the same source text is input, previous processing results are not reused, and the process is restarted. The aforementioned removal applies to all areas within the system, including temporary storage areas.

[0061] 8. Examples of Implementation

[0062] Example 1 - General literary manuscript. It is intended for a manuscript of 500,000 characters. The process of proofreading, translation, and review by human labor typically takes more than one year. According to the present invention, based on a processing time of about 30 minutes per 10,000 characters, a total processing time of about 25 hours is required, and based on an 8-hour workday, the translation of 500,000 characters is completed in about 3 days.

[0063] Example 2 - Original text of a classical Chinese text. It is intended for an original text of a classical Chinese text of the same volume. Since it is extremely difficult to secure specialized personnel in this field, processing by manpower typically takes several years. According to the present invention, the process is completed within approximately 10 days, including the re-execution process.

[0064] Example 3 - Language pairs for which securing manpower is difficult. In the case of translating original Chinese texts into Arabic, it has been extremely difficult to secure specialized personnel capable of handling both the language pair and the field, making it difficult to perform the translation itself in the past. According to the present invention, translation is performed regardless of whether manpower is secured, and a minimum quality limit is maintained through field-specific evaluation.

[0065] The above estimated time is based on continuous translation without any quality substandard results during inspection; however, if the quality is substandard, additional processing time is added as the corresponding source text requested by the user is re-translated. Explanation of the symbols

[0066] 100: Server 1 (Translation Processing) 200: Server 2 (Billing Processing) 300: Token 400: Evaluation Department 410: Select Evaluation Criteria by Field 421, 422, 42n: Evaluation item groups by field and weights by item S10: Original text input step S20: Field Decision Stage S30: Translation result generation step S40: Calculation step of comprehensive evaluation value S50: Reference value comparison step S60: Stage where output and billing are not performed S70: Output and billing execution stage S80: Step to remove information and translation content

Claims

Claim 1 A translation processing method comprising: a step of determining a field corresponding to the original text; a step of generating a translation result for the original text; a step of calculating a comprehensive evaluation value for the translation result by applying a plurality of evaluation items set differently according to the field and a predetermined weight for each of the evaluation items; a step of comparing the comprehensive evaluation value with a predetermined standard value; and a step of, if the comprehensive evaluation value falls short of the standard value, not outputting the translation result and not charging, notifying the user of this, and re-performing the step of generating the translation result upon the user's re-request. Claim 2 A translation processing method according to claim 1, characterized in that the calculation of the comprehensive evaluation value is performed internally within the system without the intervention of external personnel. Claim 3 A translation processing method according to claim 1, further comprising the step of outputting the translation result and performing corresponding billing when the comprehensive evaluation value is greater than or equal to the reference value. Claim 4 A translation processing method according to claim 1, wherein the generation of the translation result is performed on a first server and the billing is performed on a second server physically separated from the first server, and the first server and the second server are linked through a token. Claim 5 A translation processing method according to claim 1, characterized in that user information, the source text to be translated, and the translation result are not stored within the system after processing is completed, and do not remain in any form, including a temporary storage area. Claim 6 A translation processing method according to claim 1, characterized in that the step of generating the translation result includes the step of sequentially performing correction of notation errors, correction of contextual consistency, and verification of overall consistency. Claim 7 A translation processing method according to claim 1, wherein the step of generating the translation result comprises a step of dividing the translation processing into a plurality of steps, and a step of performing reorganization and verification of the result of the previous step in each step within the system and transmitting the result to the next step, and wherein the calculation of the comprehensive evaluation value is performed on the result of the completion of the plurality of steps. Claim 8 A translation processing method according to claim 7, characterized in that the above-mentioned reorganization and inspection are performed in parallel with multiple processes referencing each other, and are repeated while the results of each process are reflected in other processes. Claim 9 A translation processing method according to claim 1, characterized in that the translation processing is performed on the entire text without dividing the original text. Claim 10 A computer program stored on a computer-readable recording medium to execute a translation processing method according to any one of paragraphs 1 through 9.

Images