Interpreting evaluation method and system based on multi-dimensional competence model

Abstract

This application provides a method and system for interpreting assessment based on a multi-dimensional competence model, belonging to the field of interpreting assessment technology. The method first collects original speech, text transcription, and preset multi-dimensional competence benchmark data for the interpreter to be assessed. Then, it performs cross-modal feature alignment, integrating speech fluency, language accuracy, logical completeness, on-the-spot adaptability, and cross-cultural adaptability to construct a hierarchical assessment model. Next, it calibrates the weights of each dimension index using a dynamic weight adaptive allocation mechanism, and introduces a feature bias normalization correction algorithm to eliminate interference from cross-source data heterogeneity. Finally, based on the calibration results and the model's hierarchical mapping rules, it outputs a comprehensive assessment score, a map of competence deficiencies in each dimension, and conclusions on competence improvement and optimization. This invention can comprehensively and accurately assess interpreting ability, adapt to different scenarios, and provide targeted improvement suggestions for interpreters.

Description

Technical Field

[0001] This application relates to the field of interpreting assessment technology, and more specifically, to interpreting assessment methods and systems based on multidimensional competence models. Background Technology

[0002] In the field of interpreting, accurately assessing interpreters' abilities is crucial, as it not only affects the quality of interpreting tasks but also influences their career development and training direction. However, traditional interpreting assessment methods have several limitations. On the one hand, traditional assessments often rely on single indicators, such as focusing solely on language accuracy while neglecting other important dimensions like fluency, logical completeness, adaptability, and cross-cultural suitability, making it difficult to comprehensively and objectively reflect an interpreter's overall capabilities. On the other hand, different interpreting scenarios have varying requirements for each capability dimension. For example, business negotiation interpreting emphasizes language accuracy and logical completeness, while cultural exchange interpreting places higher demands on cross-cultural suitability, but traditional methods cannot dynamically adjust the weighting of each dimension indicator according to the scenario. Furthermore, in actual assessments, the collected speech and text data come from different sources, resulting in heterogeneous interference, which affects the accuracy and consistency of the data, thereby reducing the reliability of the assessment results. Therefore, developing a scientific, comprehensive, accurate, and scenario-adaptable interpreting assessment method and system is of significant practical importance. Summary of the Invention

[0003] This invention proposes a method and system for interpreting assessment based on a multidimensional competence model to address the shortcomings of traditional interpreting assessment methods. The method first collects data, acquiring raw speech data to be assessed, including continuous speech frame temporal sequence, articulation prosodic features, and fundamental frequency fluctuations. Simultaneously, it collects transcribed text data and pre-set multidimensional competence benchmark data. Next, it performs cross-modal feature alignment processing, collecting semantic entities, sentence structure, and grammatical data from the transcribed text. Based on rules, it performs text parsing and feature fusion to obtain text-dimensional feature data and completes the association matching between speech and text features. On this basis, it integrates multiple dimensions such as speech fluency to construct a hierarchical multidimensional competence assessment model. Finally, it incorporates a dynamic weight adaptive allocation mechanism. This system collects historical evaluation data from different interpreting scenarios, constructs a weighted iterative update function, and dynamically adjusts the initial weight coefficients of each capability dimension. Simultaneously, a feature bias normalization correction algorithm is introduced to standardize cross-source heterogeneous data and eliminate data interference. Finally, based on the calibrated multi-dimensional index fusion calculation results and the capability model hierarchical mapping rules, the system outputs the comprehensive interpreting capability evaluation score and dimension-specific capability defect map of the evaluated object, clarifying the causes, scope of impact, and attribute classification information of defects. It also provides capability improvement optimization conclusions including phased capability improvement suggestions, specialized training task data, and long-term capability planning data. The system includes modules for data acquisition, cross-modal feature alignment, model calibration, and evaluation output. It can also implement the above methods through processor execution, achieving comprehensive and accurate interpreting capability assessment and personalized improvement suggestion output.

[0004] This application provides an interpreting assessment method based on a multidimensional competence model, including the following steps: Collect raw speech data, translated text data, and pre-set multi-dimensional benchmark indicators of interpreting skills to be evaluated; Cross-modal feature alignment is performed based on speech temporal features and text semantic features, and a hierarchical multidimensional ability evaluation model is constructed by integrating speech fluency, language accuracy, logical integrity, on-the-spot adaptability and cross-cultural adaptability. By combining a dynamic weight adaptive allocation mechanism, the weights of the capability indicators of each dimension are iteratively calibrated, and a feature bias normalization correction algorithm is introduced to eliminate cross-source data heterogeneity interference. Based on the calibrated multidimensional index fusion calculation results and the hierarchical mapping rules of the capability model, the comprehensive interpretation capability assessment score, the capability defect map of each dimension, and the capability improvement and optimization conclusions of the subject to be evaluated are obtained.

[0005] In the interpreting assessment method based on a multidimensional competence model described in this application, the collection of raw speech data of the interpreter to be assessed specifically includes: Collect sequential data of continuous speech frames, prosodic feature data, and fundamental frequency fluctuation data of speech during the interpretation process to be evaluated; Based on the speech temporal feature decomposition rules, segmented feature extraction is performed on continuous speech frames. Combined with the speech prosody distortion discrimination criteria, the underlying feature quantification and analysis of fluency index is completed. Based on the preset speech distortion threshold, speech feature discrimination of stuttering, redundant sentence breaks, and speech rate imbalance in the interpretation process is completed. Based on the above quantitative analysis and feature discrimination results of speech features, the underlying feature data of the speech fluency dimension to be evaluated in interpreting is obtained.

[0006] In the interpreting evaluation method based on a multidimensional competence model described in this application, the cross-modal feature alignment processing based on speech temporal features and text semantic features specifically includes: Collect semantic entity data, sentence and grammatical structure data, terminology translation accuracy data, and discourse logical connection data from the translated text. Based on the text semantic parsing rules, entity extraction, grammatical regulation and logical link reconstruction are carried out on the transcribed text. Multiple text evaluation sub-dimensions are divided, and text feature fusion processing is completed by combining preset weight factors. Based on the multi-dimensional analysis results of the text and the mapping rules in the middle layer of the model, the dimensional feature data of the interpreted text are obtained, and the cross-modal feature association matching between the speech temporal features and the text semantic features is completed.

[0007] In the interpreting assessment method based on a multidimensional competence model described in this application, the hierarchical multidimensional competence assessment model constructed by integrating speech fluency, language accuracy, logical completeness, on-the-spot adaptability, and cross-cultural adaptability is specifically as follows: Collect multidimensional capability-level basic feature data, mid-level sub-capability feature data, and top-level comprehensive capability level classification data; Based on the hierarchical mapping rules of the model, the hierarchical association mapping of the bottom features, the middle indicators, and the top level is completed, and a hierarchical multidimensional capability evaluation model is built. Feature aggregation processing, indicator calculation processing, and level classification processing are carried out in sequence. Based on the results of the hierarchical mapping operation and the grading threshold judgment criteria, the data of each sub-item capability level of the object to be evaluated are obtained.

[0008] In the interpreting assessment method based on a multidimensional competence model described in this application, the step of combining a dynamic weight adaptive allocation mechanism to iteratively calibrate the weights of each competence indicator specifically involves: Collect historical evaluation data of different interpretation scenarios, historical data of various ability indicators, and model evaluation error feedback data; Based on the dynamic adjustment rules of the weights corresponding to the scene attributes, and combined with the convergence rules of historical data, a weight iteration update function is constructed to dynamically adjust the initial weight coefficients of each capability dimension according to different interpretation application scenarios. Based on the dynamic adjustment calculation results of the weights, the target weight data of each dimension of capability indicators after iterative calibration is obtained.

[0009] In the interpreting assessment method based on a multidimensional competence model described in this application, the introduction of a feature bias normalization correction algorithm to eliminate cross-source data heterogeneity interference specifically includes: Collect feature difference data of cross-source heterogeneous data, dimensional deviation data of speech and text data, and feature distribution dispersion data; Data standardization processing is carried out based on the feature deviation normalization correction algorithm. The dimensionality of speech time-domain feature data and text semantic quantization data is uniformly converted, and the feature distribution interval regularization processing corresponding to cross-source data heterogeneity interference is completed. Based on the dimensionally unified feature data and the multidimensional model fusion input rules, cross-modal fusion feature data of uniform specifications is obtained.

[0010] In the interpreting assessment method based on a multidimensional competence model described in this application, obtaining the dimensional component competence defect map specifically involves: Collect feature data on the specific capabilities of the object to be evaluated, quantitative data on the shortcomings of capabilities in each dimension, and benchmark data on the capability indicators of similar samples. Based on the defect feature comparison analysis rules, the difference between individual evaluation indicators and benchmark indicators is compared and processed, and defect levels are classified in combination with the defect level classification rules. Based on the correlation analysis results between dimensional defect comparison data and capability shortcomings, dimensional capability defect map data containing information on defect causes, impact scope, and attribute classification are obtained.

[0011] In the interpreting assessment method based on a multidimensional competence model described in this application, the achievement of competence improvement optimization conclusions specifically refers to: Collect data on the generated comprehensive evaluation scores, the capability deficiency map of each dimension, and the knowledge material data for improving capabilities in each dimension; Based on the corresponding supplementation strategies and personalized matching rules for the defects in the capability dimensions, and combined with the data on the shortcomings of each capability dimension, corresponding training content and improvement paths are matched, and a hierarchical optimization scheme is constructed according to the priority of the defects. Based on the scheme matching calculation results and regularization output rules, the capability improvement optimization conclusions are obtained, including phased capability improvement suggestions, special training task data, and long-term capability planning data.

[0012] Secondly, this application provides an interpreting assessment system based on a multidimensional competence model, including: The data acquisition module is used to collect the original speech data, the transcribed text data, and the preset multi-dimensional capability benchmark data of the interpreters to be evaluated. The cross-modal feature alignment module is used to perform cross-modal feature alignment processing based on speech temporal features and text semantic features, and integrates speech fluency, language accuracy, logical integrity, on-the-spot adaptability and cross-cultural adaptability to build a hierarchical multi-dimensional ability evaluation model. The model calibration module is used to perform iterative calibration of the capability indicators of each dimension by combining a dynamic weight adaptive allocation mechanism, and at the same time introduces a feature bias normalization correction algorithm to eliminate cross-source data heterogeneity interference. The evaluation output module is used to output the comprehensive interpreting ability evaluation score, the dimension-specific ability defect map, and the ability improvement and optimization conclusions of the evaluated object based on the fusion calculation results of the multi-dimensional indicators after calibration and the hierarchical mapping rules of the ability model.

[0013] The system also includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the interpretation assessment methods based on the multidimensional competence model.

[0014] As described above, this invention discloses an interpretation evaluation method and corresponding system based on a multi-dimensional capability model. It comprehensively and meticulously collects the original speech data (including continuous speech frame temporal sequence, pronunciation prosody, etc.), text transcription data, and preset multi-dimensional capability benchmark data of the interpretation to be evaluated. Cross-modal feature alignment is performed, and text semantic entity data is processed in depth to complete text feature fusion, which is then associated and matched with speech temporal features. A hierarchical evaluation model is constructed by integrating multiple dimensions such as speech fluency, and multi-dimensional data is collected to complete hierarchical mapping to obtain sub-item capability level data. Combined with a dynamic weight adaptive allocation mechanism, multiple types of data are collected to construct a weight iterative update function to calibrate the weights of each dimension indicator. Simultaneously, a feature deviation normalization correction algorithm is introduced, and data such as cross-source heterogeneous data feature differences are collected. The dimensional transformation and feature distribution interval regularization of speech temporal and text semantic quantification data are performed to eliminate cross-source data heterogeneity interference. Finally, based on the calibration results and model hierarchical mapping rules, a comprehensive evaluation score, a dimensional sub-item capability defect map (including defect causes and other information), and capability improvement optimization conclusions (including phased suggestions) are output.

[0015] Other features and advantages of this application will be set forth in the following description and will be apparent in part from the description or may be learned by practicing embodiments of this application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A high-level flowchart of the interpreting assessment method based on a multidimensional competence model provided in the embodiments of this application, used for interpreting assessment; Figure 2 A flowchart illustrating the interpreting assessment method based on a multidimensional competence model provided in this application embodiment; Figure 3 The structural block diagram of the interpreting assessment system based on a multidimensional competence model provided in the embodiments of this application is shown. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0019] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first" and "second" are used only to distinguish descriptions and should not be construed as indicating or implying relative importance. It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0020] Please refer to Figure 1 , Figure 1This is a high-level flowchart of an interpreting assessment method based on a multi-dimensional competence model, as described in some embodiments of this application. The high-level flowchart can be summarized as follows: First, collect the original speech, transcribed text, and preset multi-dimensional competence benchmark data of the interpreter to be assessed; then, perform cross-modal feature alignment based on the temporal and semantic features of the speech and text, and fuse multiple dimensions to construct a hierarchical assessment model; next, iteratively calibrate the weights of each dimension index using a dynamic weight adaptive allocation mechanism, and eliminate cross-source data interference using a feature deviation normalization correction algorithm; finally, based on the calibrated multi-dimensional index fusion results and hierarchical mapping rules, derive the comprehensive assessment score, sub-item competence defect map, and competence improvement conclusions.

[0021] Please refer to Figure 2 , Figure 2 This is a flowchart of an interpretation assessment method based on a multidimensional competence model in some embodiments of this application.

[0022] The first aspect of this invention discloses an interpreting assessment method based on a multidimensional competence model for use in terminal devices, such as computers and mobile phones. This interpreting assessment method based on a multidimensional competence model includes the following steps: S201. Collect the original speech data of the interpreter to be evaluated, the transcription data of the interpreter text, and the preset multi-dimensional ability benchmark data. S202. Based on the speech temporal features and text semantic features, perform cross-modal feature alignment processing, and integrate speech fluency, language accuracy, logical integrity, on-the-spot adaptability and cross-cultural adaptability to construct a hierarchical multi-dimensional ability evaluation model. S203. Combine the dynamic weight adaptive allocation mechanism to perform weight iterative calibration of the capability indicators of each dimension, and introduce the feature deviation normalization correction algorithm to eliminate cross-source data heterogeneity interference. S204. Based on the fusion calculation results of the multi-dimensional indicators after calibration and the hierarchical mapping rules of the capability model, the comprehensive interpretation capability assessment score, the capability defect map of each dimension, and the capability improvement and optimization conclusions of the subject to be evaluated are obtained.

[0023] The process involves several key steps. First, data collection is conducted to accurately acquire the raw audio data of the interpreters to be evaluated. This data includes temporal information of continuous audio frames and prosodic features. Corresponding transcripts are also collected to fully represent the interpreting content. Furthermore, pre-defined multi-dimensional competency benchmark data are gathered to provide a reference standard for subsequent evaluation. Next, cross-modal feature alignment processing is implemented based on the temporal features of the audio and the semantic features of the text, deeply associating audio and text information. On this basis, a hierarchical, multi-dimensional competency evaluation model is constructed by integrating multiple key dimensions such as fluency, accuracy, logical completeness, adaptability, and cross-cultural suitability to comprehensively and meticulously assess interpreting skills. Subsequently, a dynamic weight adaptive allocation mechanism is used to iteratively calibrate the weights of each competency indicator based on different scenarios and actual data conditions, ensuring the scientific and reasonable allocation of weights. Simultaneously, a feature bias normalization correction algorithm is introduced to effectively eliminate heterogeneous interference caused by different data sources, ensuring data accuracy and consistency. Finally, based on the calibrated multidimensional index fusion calculation results and the hierarchical mapping rules of the capability model, the comprehensive interpreting ability assessment score of the subject to be evaluated is obtained, which intuitively reflects its overall interpreting level; a capability defect map of each dimension is generated, which clearly presents the weak links of each dimension; and capability improvement and optimization conclusions are given, providing targeted suggestions for the improvement of interpreting ability, and providing a scientific, accurate and comprehensive method for interpreting assessment.

[0024] According to an embodiment of the present invention, the collection of raw speech data for the interpretation to be evaluated specifically includes: Collect sequential data of continuous speech frames, prosodic feature data, and fundamental frequency fluctuation data of speech during the interpretation process to be evaluated; Based on the speech temporal feature decomposition rules, segmented feature extraction is performed on continuous speech frames. Combined with the speech prosody distortion discrimination criteria, the underlying feature quantification and analysis of fluency index is completed. Based on the preset speech distortion threshold, speech feature discrimination of stuttering, redundant sentence breaks, and speech rate imbalance in the interpretation process is completed. Based on the above quantitative analysis and feature discrimination results of speech features, the underlying feature data of the speech fluency dimension to be evaluated in interpreting is obtained.

[0025] Among these efforts, meticulous and comprehensive operations were conducted focusing on the crucial step of collecting raw audio data for the interpretation tests. This involved collecting a wealth of audio information from the interpretation process, including sequential data of continuous audio frames. This data accurately presents the order and distribution of speech over time, providing a foundation for subsequent analysis of rhythm and coherence. It also included prosodic feature data, reflecting variations in pitch, intensity, and duration, revealing the interpreter's style and emotional state, and fundamental frequency fluctuation data. Fundamental frequency fluctuations are closely related to pitch changes and are crucial for accurately grasping audio characteristics. Subsequently, based on pre-defined temporal feature decomposition rules, segmented feature extraction was performed on continuous audio frames. This rule acts as a precise ruler, rationally dividing continuous audio frames to deeply explore the unique features of each segment. Building upon this, a prosodic distortion discrimination criterion was used to quantitatively analyze the underlying features of the fluency index. This criterion compares the differences between normal and actual prosodic rhythms to determine the presence of prosodic distortions, thus providing a quantitative basis for fluency assessment. Simultaneously, based on a preset speech distortion threshold, the system accurately identifies speech features such as pauses, redundant phrasing, and unbalanced speaking rates during interpretation. These speech distortion phenomena severely affect the quality and fluency of interpretation; by setting reasonable thresholds, these problems can be objectively and accurately identified. Combining the above-mentioned quantitative analysis and feature discrimination results of speech features, the underlying feature data for the speech fluency dimension of the interpreter to be evaluated is obtained. This data serves as a crucial foundation for subsequently constructing a hierarchical, multi-dimensional competence evaluation model, providing detailed and reliable evidence for accurately assessing the interpreter's speech fluency and ensuring the scientific rigor and accuracy of the entire interpretation evaluation technology solution.

[0026] According to an embodiment of the present invention, the cross-modal feature alignment processing based on speech temporal features and text semantic features specifically includes: Collect semantic entity data, sentence and grammatical structure data, terminology translation accuracy data, and discourse logical connection data from the translated text. Based on the text semantic parsing rules, entity extraction, grammatical regulation and logical link reconstruction are carried out on the transcribed text. Multiple text evaluation sub-dimensions are divided, and text feature fusion processing is completed by combining preset weight factors. Based on the multi-dimensional analysis results of the text and the mapping rules in the middle layer of the model, the dimensional feature data of the interpreted text are obtained, and the cross-modal feature association matching between the speech temporal features and the text semantic features is completed.

[0027] The key step of cross-modal feature alignment based on speech temporal features and text semantic features involves the comprehensive collection of various key data from the translated text. This includes semantic entity data, which accurately identifies core concepts and key information; sentence structure and grammar data, which helps to clearly understand the sentence structure and adherence to grammatical rules; terminology translation accuracy data, crucial for assessing the accuracy of interpreting in conveying professional terminology; and discourse logic data, reflecting the overall logical coherence and coherence of the text. Based on pre-defined text semantic parsing rules, the translated text undergoes in-depth processing. Entity extraction accurately extracts semantic entities from the text, providing foundational elements for subsequent analysis; grammatical regulation makes the sentence and grammatical structures more standardized, reasonable, and consistent with language expression habits; and logical link reconstruction is carried out to clarify and strengthen the logical connections between different parts of the text, making the text's logical flow clearer. Based on this, multiple sub-dimensions for text evaluation are defined, such as entity accuracy, grammatical correctness, and logical rationality. These sub-dimensions are then fused using pre-defined weighting factors to comprehensively assess the text's performance across various aspects. Based on the analysis results of these multi-dimensional text features and the model's layer mapping rules, dimensional feature data for the interpreting text are obtained. This data comprehensively and accurately reflects the text's semantic, grammatical, and logical characteristics. Simultaneously, cross-modal feature association matching between speech temporal features and text semantic features is achieved, closely linking the speech's temporal features with the text's semantic features. This lays a solid foundation for subsequently constructing a hierarchical, multi-dimensional competence evaluation model and comprehensively and accurately assessing interpreting abilities.

[0028] According to an embodiment of the present invention, the hierarchical multidimensional ability assessment model constructed by integrating speech fluency, language accuracy, logical completeness, on-the-spot adaptability, and cross-cultural adaptability is specifically as follows: Collect multidimensional capability-level basic feature data, mid-level sub-capability feature data, and top-level comprehensive capability level classification data; Based on the hierarchical mapping rules of the model, the hierarchical association mapping of the bottom features, the middle indicators, and the top level is completed, and a hierarchical multidimensional capability evaluation model is built. Feature aggregation processing, indicator calculation processing, and level classification processing are carried out in sequence. Based on the results of the hierarchical mapping operation and the grading threshold judgment criteria, the data of each sub-item capability level of the object to be evaluated are obtained.

[0029] To construct a scientific and reasonable hierarchical multidimensional ability assessment model, a comprehensive and meticulous process was conducted to integrate key dimensions such as speech fluency, language accuracy, logical completeness, on-the-spot adaptability, and cross-cultural adaptability. First, a wide range of basic feature data on multidimensional abilities were collected. This data originated from in-depth analysis of speech and text. For example, the basic feature data for speech fluency includes features such as stuttering, redundant sentence breaks, and unbalanced speech rate, accurately reflecting the fluency of interpreting speech. Simultaneously, mid-level sub-ability feature data was collected, which is further refined based on the basic features. For example, the sub-ability feature data for language accuracy reflects the accuracy of interpreting in terms of vocabulary and grammar. Finally, top-level comprehensive ability level classification data was collected to provide a basis for the final assessment of the comprehensive interpreting ability level. Based on pre-defined hierarchical mapping rules, a step-by-step mapping of basic features, mid-level indicators, and top-level levels was completed. This rule clarifies the correspondence and conversion methods between each level, ensuring that data can be accurately and reasonably transmitted and converted between different levels, thereby building a hierarchical multidimensional ability assessment model. Within this model framework, feature aggregation is performed sequentially to integrate and summarize scattered low-level features; index calculation is conducted, using specific algorithms to calculate and analyze the mid-level sub-skill feature data; and level classification is performed, categorizing interpreting abilities into the corresponding top-level levels based on the calculation results. By comparing the hierarchical mapping calculation results with the grading threshold judgment criteria—which are reasonable boundary values ​​determined through extensive practice and data analysis—the calculation results can be accurately compared to obtain the level data of each sub-skill of the subject being evaluated, providing a clear and accurate basis for a comprehensive and objective assessment of interpreting abilities.

[0030] According to an embodiment of the present invention, the step of combining a dynamic weight adaptive allocation mechanism to perform weight iterative calibration of the capability indicators of each dimension specifically involves: Collect historical evaluation data of different interpretation scenarios, historical data of various ability indicators, and model evaluation error feedback data; Based on the dynamic adjustment rules of the weights corresponding to the scene attributes, and combined with the convergence rules of historical data, a weight iteration update function is constructed to dynamically adjust the initial weight coefficients of each capability dimension according to different interpretation application scenarios. Based on the dynamic adjustment calculation results of the weights, the target weight data of each dimension of capability indicators after iterative calibration is obtained.

[0031] To ensure the scientific and rational allocation of weights for each capability indicator and thus improve the accuracy of interpreting assessments, a method combining dynamic weight adaptive allocation mechanism for iterative calibration of weights for each capability indicator was adopted. Comprehensive and detailed historical assessment data from different interpreting scenarios were collected. This data covers assessment results in various actual interpreting scenarios, accurately reflecting the specific needs of different scenarios for interpreting skills. Historical data for each capability indicator was also collected; analysis of this data reveals the performance of each indicator in different periods and scenarios. Additionally, model assessment error feedback data was collected, accurately identifying deviations in the model during assessment and providing crucial information for weight adjustments. Based on the dynamic weight adjustment rules corresponding to scenario attributes, this rule fully considers the uniqueness of different interpreting scenarios. For example, business negotiation scenarios emphasize language accuracy and on-the-spot adaptability, while cultural exchange scenarios have higher requirements for cross-cultural adaptability. A weight iterative update function was constructed based on the convergence pattern of historical data—that is, the pattern of data gradually stabilizing and concentrating as the amount of data increases. This function can flexibly and dynamically adjust the initial weight coefficients of each capability dimension according to the actual situation of different interpreting application scenarios, ensuring that the initial weights match the needs of the actual scenario. Based on the dynamic adjustment calculation results, after a series of scientific and reasonable calculations and analyses, the target weight data of each capability indicator after iterative calibration is obtained. This target weight data will provide accurate weight basis for subsequent interpreting assessments, making the assessment results more objective, fair, and accurate, and effectively improving the quality and reliability of the entire interpreting assessment technology solution.

[0032] According to an embodiment of the present invention, the introduction of a feature bias normalization correction algorithm to eliminate cross-source data heterogeneity interference specifically includes: Collect feature difference data of cross-source heterogeneous data, dimensional deviation data of speech and text data, and feature distribution dispersion data; Data standardization processing is carried out based on the feature deviation normalization correction algorithm. The dimensionality of speech time-domain feature data and text semantic quantization data is uniformly converted, and the feature distribution interval regularization processing corresponding to cross-source data heterogeneity interference is completed. Based on the dimensionally unified feature data and the multidimensional model fusion input rules, cross-modal fusion feature data of uniform specifications is obtained.

[0033] One challenge is the heterogeneity of the collected speech and text data, stemming from different sources. This heterogeneity can interfere with subsequent evaluation model construction and accurate assessment. To address this, a feature bias normalization correction algorithm is introduced to eliminate cross-source data heterogeneity interference. This algorithm comprehensively and accurately collects feature difference data from heterogeneous cross-source data, clearly revealing the differences between speech and text data at the feature level. Simultaneously, it collects dimensional deviation data for speech and text data. Differences in dimensions lead to inconsistencies in numerical values ​​and measurement standards, affecting comprehensive data analysis. Additionally, it collects feature distribution dispersion data, reflecting the degree of data dispersion; directly fusing data with different dispersion levels can degrade data quality. Next, data standardization is performed using the feature bias normalization correction algorithm. This algorithm performs a unified dimensional transformation on speech temporal feature data and text semantic quantization data, converting data with different dimensions to a unified measurement standard, eliminating the impact of dimensional differences. Furthermore, it performs feature distribution interval normalization processing corresponding to cross-source data heterogeneity interference, making the feature distribution intervals of data from different sources more consistent, further reducing the interference of data heterogeneity. Finally, based on the unified feature data and the multidimensional model fusion input rules, these rules clarify the format, order and other requirements of different feature data during fusion, thereby obtaining cross-modal fusion feature data of unified specifications, ensuring that the entire interpretation evaluation technology solution can be carried out accurately and effectively.

[0034] According to an embodiment of the present invention, obtaining the dimensional item capability defect map specifically involves: Collect feature data on the specific capabilities of the object to be evaluated, quantitative data on the shortcomings of capabilities in each dimension, and benchmark data on the capability indicators of similar samples. Based on the defect feature comparison analysis rules, the difference between individual evaluation indicators and benchmark indicators is compared and processed, and defect levels are classified in combination with the defect level classification rules. Based on the correlation analysis results between dimensional defect comparison data and capability shortcomings, dimensional capability defect map data containing information on defect causes, impact scope, and attribute classification are obtained.

[0035] To accurately pinpoint the deficiencies of the candidates in each dimension of their abilities, a defect map of each dimension's capabilities was obtained. This involved comprehensively collecting characteristic data on the defects of the candidates' abilities across various dimensions, including speech, language, and logic, during interpreting. Examples of defects included potential stuttering or uneven speaking speed in terms of fluency. Simultaneously, quantitative data on shortcomings in each dimension was collected, quantifying these shortcomings using scientific methods, such as specific numerical values ​​for vocabulary and grammar error rates in the language accuracy dimension. Furthermore, benchmark data on the ability indicators of similar samples was collected. This data was derived from the analysis and summarization of numerous similar interpreting samples, providing an objective and unified reference standard for subsequent comparisons. Based on the rules for comparative analysis of defect characteristics, the individual evaluation indicators of the candidates were compared with the benchmark indicators of similar samples to clearly identify the gap between the individual and the benchmark. Building upon this, and combining the rules for classifying defect levels, defect levels were determined based on factors such as the magnitude and nature of the difference, such as classifying defects as minor, moderate, and severe, to more accurately assess the severity of the defects. Based on the correlation analysis results between dimensional defect comparison data and capability shortcomings, this study delves into the intrinsic connections between defects in each dimension and their relationship with overall capability shortcomings, resulting in a dimensional capability defect map data that includes information on defect causes, scope of impact, and attribute grading. This map data can intuitively and comprehensively present the defects of the evaluated object in each dimension, providing a strong basis for subsequently developing targeted improvement strategies and solutions, and ensuring that the entire interpreting assessment technology solution can effectively help the evaluated object improve its interpreting skills.

[0036] According to an embodiment of the present invention, the process of obtaining the capability enhancement optimization conclusion specifically includes: Collect data on the generated comprehensive evaluation scores, the capability deficiency map of each dimension, and the knowledge material data for improving capabilities in each dimension; Based on the corresponding supplementation strategies and personalized matching rules for the defects in the capability dimensions, and combined with the data on the shortcomings of each capability dimension, corresponding training content and improvement paths are matched, and a hierarchical optimization scheme is constructed according to the priority of the defects. Based on the scheme matching calculation results and regularization output rules, the capability improvement optimization conclusions are obtained, including phased capability improvement suggestions, special training task data, and long-term capability planning data.

[0037] To ultimately arrive at scientific, practical, and personalized assessment conclusions for enhancing interpreting skills, and to genuinely assist the assessed individuals in improving their interpreting abilities, the following steps were taken: A comprehensive evaluation score was collected, which considers the individual's overall performance across multiple dimensions, including fluency, accuracy, logical completeness, adaptability, and cross-cultural compatibility, providing a clear picture of their comprehensive interpreting ability. Simultaneously, a deficiency map of each dimension's capabilities was acquired. This map details the causes, scope, and attribute classification of each individual's deficiencies, providing a strong basis for accurately identifying weaknesses. Furthermore, knowledge materials for improving capabilities across each dimension were collected. These materials encompass rich theoretical knowledge, practical skills, and classic cases, representing crucial resources for enhancing interpreting abilities. Based on corresponding remediation strategies and personalized matching rules for each capability dimension, the rules fully consider the characteristics of different capability deficiencies and the individual differences of the assessed individuals. Combining this with the data on weaknesses in each dimension, the corresponding training content and improvement paths were precisely matched. For example, if the candidate exhibits inaccurate vocabulary usage in the language accuracy dimension, relevant vocabulary learning materials and specialized training tasks are provided; if there is ambiguity in logical completeness, logical training methods and case studies are offered. Furthermore, based on the priority of deficiencies—that is, ranking them according to their impact on overall interpreting ability—a tiered optimization plan is constructed to ensure that the most impactful deficiencies are addressed first. Based on the matching calculation results and output rules, the format and structure of the output content are standardized to ensure clear and readable information. After a series of calculations and processing, capability improvement optimization conclusions are obtained, including phased capability improvement suggestions. Based on the candidate's actual situation and capability improvement goals, the candidate is divided into different stages, with specific improvement goals and action plans for each stage; specialized training task data clarifies the specific training tasks and requirements for each dimension of capability deficiency; and long-term capability planning data develops a long-term interpreting ability development blueprint for the candidate, enabling them to improve their interpreting skills systematically.

[0038] According to an embodiment of the present invention, it further includes: Collect raw input data, intermediate feature processing data, model calculation process data, and final evaluation output data at each stage; Based on the full-link data traceability rules, the correlation of data flow nodes at each stage is verified. Combined with the data deviation backtracking and location rules, the source investigation of abnormal data nodes is completed. Based on the data integrity verification standards, the validity of data at each stage is verified. Based on the results of the full-process data traceability and verification, data credibility parameters for each stage and reference criteria for correcting abnormal data are obtained.

[0039] To ensure the reliability, accuracy, and completeness of data throughout the entire evaluation process, and thus guarantee the objectivity and fairness of the evaluation results, a full-process data traceability and verification step has been added. This involves comprehensively and meticulously collecting raw input data from each stage. This data serves as the starting information for the entire evaluation process, covering key content such as the basic information of the object being evaluated and initial interpretation materials, forming the basis for all subsequent processing and analysis. Simultaneously, intermediate feature processing data is collected. During the interpretation evaluation process, a series of feature extraction and processing operations are performed on the raw data. The data generated at this stage reflects the intermediate states of the data processing. Model computation data is also indispensable, recording various parameters and calculation results of the evaluation model during its operation, which is crucial for understanding the model's working mechanism and evaluating its performance. The final evaluation output data is the final result of the entire process, directly reflecting the evaluation conclusions for the object being evaluated. Based on the full-link data traceability rules, the correlation of data flow nodes at each stage is verified. These rules clarify the transmission relationships and logical connections between different stages. Verification ensures that data maintains continuity and consistency during the flow process, avoiding data loss or incorrect correlation. Building upon this foundation, and incorporating data deviation backtracking and location rules, when data anomalies are detected, the system can trace back along the data flow path to pinpoint the anomaly, providing a clear direction for subsequent problem-solving. Furthermore, based on data integrity verification standards, the validity of data at each stage is validated. These standards specify the complete elements and format requirements that data should possess, allowing for the timely identification of missing, erroneous, or non-standard data. Based on the full-process data traceability verification results, and through comprehensive analysis of data correlation, deviation location, and integrity, data reliability parameters for each stage are obtained. These parameters directly reflect the reliability of data at each stage, providing crucial information for evaluating the quality of the entire assessment process. Simultaneously, reference data for anomaly correction is obtained, providing specific correction methods and suggestions for anomaly nodes identified through traceability, ensuring timely and accurate data correction. This guarantees the smooth implementation of the entire interpreting assessment technology solution and the accuracy and reliability of the assessment results.

[0040] According to an embodiment of the present invention, it further includes: Collect batch evaluation results data of multiple sets of objects to be evaluated, cumulative sub-item capability statistics data, and updated historical data of benchmark indicators; Based on the batch data statistical analysis rules, multi-dimensional indicator clustering and summarization are carried out, and the preset multi-dimensional capability benchmark indicator thresholds are corrected in combination with the dynamic update rules of the indicator library. Based on the results of batch data statistical calculations and indicator updates, a dynamically optimized multi-dimensional capability benchmark indicator dataset is obtained.

[0041] To continuously improve the accuracy and adaptability of the assessment, and to ensure it keeps pace with changes in real-world application scenarios and the development trends of the assessed individuals' abilities, a dynamic optimization mechanism for multi-dimensional competency benchmark indicators has been introduced. First, a comprehensive and systematic collection of batch assessment results data from multiple groups of assessed individuals is conducted. This data covers the overall performance of different batches and types of assessed individuals in interpreting assessments, reflecting the overall ability distribution. Simultaneously, cumulative sub-item competency statistics are collected. Through long-term statistics of each competency indicator, the development trends and changing patterns of different dimensions are clearly presented. Additionally, historical data of updated benchmark indicators is collected, recording adjustments made to the benchmark indicators at different times, providing historical reference for subsequent indicator optimization. Based on batch data statistical analysis rules, the collected data are clustered and summarized using multi-dimensional indicators. This rule uses scientific statistical methods to group data with similar characteristics into one category, thereby deeply exploring the potential information and patterns behind the data. On this basis, combined with the dynamic update rules of the indicator library, which fully consider the development dynamics of the interpreting industry, changes in actual needs, and the improvement of the assessed individuals' abilities, the preset thresholds for multi-dimensional competency benchmark indicators are adjusted. This approach ensures that benchmark indicators reflect the actual level and requirements of current interpreting skills in a timely manner. Based on batch data statistical calculations and indicator update results, a series of rigorous calculations and analyses yield a dynamically optimized multi-dimensional competency benchmark indicator dataset. This dataset not only includes the updated thresholds for each dimension of competency benchmark indicators but also reflects the interrelationships and overall structure between the indicators. It provides a more scientific, reasonable, and accurate reference standard for subsequent interpreting assessments, making the assessment results more convincing and practical.

[0042] Please refer to Figure 3 , Figure 3 This is a structural block diagram of the interpreting assessment system based on a multidimensional competence model provided in the embodiments of this application.

[0043] A second aspect of the present invention also discloses an interpreting assessment system based on a multidimensional competence model, comprising: The data acquisition module 301 is used to collect the original speech data of the interpreter to be evaluated, the transcribed data of the interpreter text, and the preset multi-dimensional ability benchmark index data. The cross-modal feature alignment module 302 is used to perform cross-modal feature alignment processing based on speech temporal features and text semantic features, and to integrate speech fluency, language accuracy, logical integrity, on-the-spot adaptability and cross-cultural adaptability to construct a hierarchical multi-dimensional ability evaluation model. The model calibration module 303 is used to perform weight iteration calibration of the capability indicators of each dimension by combining a dynamic weight adaptive allocation mechanism, and at the same time introduces a feature bias normalization correction algorithm to eliminate cross-source data heterogeneity interference. The evaluation output module 304 is used to output the comprehensive interpretation ability evaluation score, the dimension sub-item ability defect map, and the ability improvement and optimization conclusion of the subject to be evaluated based on the fusion calculation results of the multi-dimensional indicators after calibration and the hierarchical mapping rules of the ability model.

[0044] The system also includes a memory and a processor, wherein the memory includes a program for an interpreting assessment method based on a multidimensional competence model, and when the program for the interpreting assessment method based on the multidimensional competence model is executed by the processor, it implements the steps of the interpreting assessment method based on the multidimensional competence model as described in any one of the first aspects.

[0045] The present invention discloses an interpreting evaluation method and system based on a multi-dimensional competence model. The method first collects raw speech data to be evaluated (including continuous speech frame temporal data, from which low-level feature data for speech fluency dimension is obtained), translated text data, and preset multi-dimensional competence benchmark data. Next, cross-modal feature alignment processing is performed based on speech temporal features and text semantic features (collecting relevant text data, completing parsing, fusion, and association matching to obtain translated text dimension feature data). A hierarchical multi-dimensional competence evaluation model is constructed by integrating multiple dimensions such as speech fluency (collecting relevant data, completing hierarchical association mapping, and processing to obtain the competence level data for each sub-item). Finally, a dynamic weight adaptive allocation mechanism is used to evaluate the competence levels of each dimension. The system performs iterative calibration of the performance indicators (collecting relevant data to construct an update function and dynamically adjust the initial weight coefficients to obtain target weight data). A feature bias normalization correction algorithm is introduced to eliminate cross-source data heterogeneity interference (collecting relevant data and performing standardization processing to obtain cross-modal fusion feature data of uniform specifications). Finally, based on the calibrated multi-dimensional indicator fusion calculation results and the hierarchical mapping rules of the capability model, the system obtains the comprehensive interpreting ability evaluation score of the subject to be evaluated, the dimensional item capability defect map (collecting relevant data to complete difference comparison and defect level classification to obtain map data containing multi-faceted information), and capability improvement optimization conclusions (collecting relevant data to match training content and improvement paths to construct a hierarchical optimization scheme and obtain relevant conclusions). The system includes a data acquisition module, a cross-modal feature alignment module, a model calibration module, and an evaluation output module. It also includes a memory, a processor, and a computer program. The processor executes the program to implement the above methods. In addition, data from each stage is collected, and correlation verification, abnormal data node tracing and investigation, and data validity verification are carried out according to the full-link data traceability rules to obtain data credibility parameters and abnormal data correction reference basis for each stage; multiple sets of batch evaluation data of the objects to be evaluated are collected, and the preset multi-dimensional capability benchmark index thresholds are corrected according to the batch data statistical analysis rules to obtain a dynamically optimized multi-dimensional capability benchmark index dataset.

[0046] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0047] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0048] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0049] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, read-only memory, random access memory, magnetic disks, or optical disks.

[0050] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

Claims

1. An interpreting assessment method based on a multidimensional competence model, characterized in that, Includes the following steps: Collect raw speech data, translated text data, and pre-set multi-dimensional benchmark indicators of interpreting skills to be evaluated; Cross-modal feature alignment is performed based on speech temporal features and text semantic features, and a hierarchical multidimensional ability evaluation model is constructed by integrating speech fluency, language accuracy, logical integrity, on-the-spot adaptability and cross-cultural adaptability. By combining a dynamic weight adaptive allocation mechanism, the weights of the capability indicators of each dimension are iteratively calibrated, and a feature bias normalization correction algorithm is introduced to eliminate cross-source data heterogeneity interference. Based on the calibrated multidimensional index fusion calculation results and the hierarchical mapping rules of the capability model, the comprehensive interpretation capability assessment score, the capability defect map of each dimension, and the capability improvement and optimization conclusions of the subject to be evaluated are obtained.

2. The interpreting assessment method based on a multidimensional competence model according to claim 1, characterized in that, The collection of raw speech data for the interpretation to be evaluated specifically includes: Collect sequential data of continuous speech frames, prosodic feature data, and fundamental frequency fluctuation data of speech during the interpretation process to be evaluated; Based on the speech temporal feature decomposition rules, segmented feature extraction is performed on continuous speech frames. Combined with the speech prosody distortion discrimination criteria, the underlying feature quantification and analysis of fluency index is completed. Based on the preset speech distortion threshold, speech feature discrimination of stuttering, redundant sentence breaks, and speech rate imbalance in the interpretation process is completed. Based on the above quantitative analysis and feature discrimination results of speech features, the underlying feature data of the speech fluency dimension to be evaluated in interpreting is obtained.

3. The interpreting assessment method based on a multidimensional competence model according to claim 1, characterized in that, The cross-modal feature alignment process based on speech temporal features and text semantic features specifically involves: Collect semantic entity data, sentence and grammatical structure data, terminology translation accuracy data, and discourse logical connection data from the translated text. Based on the text semantic parsing rules, entity extraction, grammatical regulation and logical link reconstruction are carried out on the transcribed text. Multiple text evaluation sub-dimensions are divided, and text feature fusion processing is completed by combining preset weight factors. Based on the multi-dimensional analysis results of the text and the mapping rules in the middle layer of the model, the dimensional feature data of the interpreted text are obtained, and the cross-modal feature association matching between the speech temporal features and the text semantic features is completed.

4. The interpreting assessment method based on a multidimensional competence model according to claim 1, characterized in that, The hierarchical, multi-dimensional competency assessment model, which integrates speech fluency, language accuracy, logical completeness, on-the-spot adaptability, and cross-cultural adaptability, is as follows: Collect multidimensional capability-level basic feature data, mid-level sub-capability feature data, and top-level comprehensive capability level classification data; Based on the hierarchical mapping rules of the model, the hierarchical association mapping of the bottom features, the middle indicators, and the top level is completed, and a hierarchical multidimensional capability evaluation model is built. Feature aggregation processing, indicator calculation processing, and level classification processing are carried out in sequence. Based on the results of the hierarchical mapping operation and the grading threshold judgment criteria, the data of each sub-item capability level of the object to be evaluated are obtained.

5. The interpreting assessment method based on a multidimensional competence model according to claim 1, characterized in that, The method of combining dynamic weight adaptive allocation mechanism to perform weight iterative calibration of capability indicators in each dimension is as follows: Collect historical evaluation data of different interpretation scenarios, historical data of various ability indicators, and model evaluation error feedback data; Based on the dynamic adjustment rules of the weights corresponding to the scene attributes, and combined with the convergence rules of historical data, a weight iteration update function is constructed to dynamically adjust the initial weight coefficients of each capability dimension according to different interpretation application scenarios. Based on the dynamic adjustment calculation results of the weights, the target weight data of each dimension of capability indicators after iterative calibration is obtained.

6. The interpreting assessment method based on a multidimensional competence model according to claim 1, characterized in that, The introduced feature bias normalization correction algorithm eliminates cross-source data heterogeneity interference, specifically as follows: Collect feature difference data of cross-source heterogeneous data, dimensional deviation data of speech and text data, and feature distribution dispersion data; Data standardization processing is carried out based on the feature deviation normalization correction algorithm. The dimensionality of speech time-domain feature data and text semantic quantization data is uniformly converted, and the feature distribution interval regularization processing corresponding to cross-source data heterogeneity interference is completed. Based on the dimensionally unified feature data and the multidimensional model fusion input rules, cross-modal fusion feature data of uniform specifications is obtained.

7. The interpreting assessment method based on a multidimensional competence model according to claim 1, characterized in that, The obtained dimensional component capability defect map is specifically as follows: Collect feature data on the specific capabilities of the object to be evaluated, quantitative data on the shortcomings of capabilities in each dimension, and benchmark data on the capability indicators of similar samples. Based on the defect feature comparison analysis rules, the difference between individual evaluation indicators and benchmark indicators is compared and processed, and defect levels are classified in combination with the defect level classification rules. Based on the correlation analysis results between dimensional defect comparison data and capability shortcomings, dimensional capability defect map data containing information on defect causes, impact scope, and attribute classification are obtained.

8. The interpreting assessment method based on a multidimensional competence model according to claim 1, characterized in that, The obtained capability enhancement and optimization conclusions are as follows: Collect data on the generated comprehensive evaluation scores, the capability deficiency map of each dimension, and the knowledge material data for improving capabilities in each dimension; Based on the corresponding supplementation strategies and personalized matching rules for the defects in the capability dimensions, and combined with the data on the shortcomings of each capability dimension, corresponding training content and improvement paths are matched, and a hierarchical optimization scheme is constructed according to the priority of the defects. Based on the scheme matching calculation results and regularization output rules, the capability improvement optimization conclusions are obtained, including phased capability improvement suggestions, special training task data, and long-term capability planning data.

9. An interpreting assessment system based on a multidimensional competence model, characterized in that, include: The data acquisition module is used to collect the original speech data, the transcribed text data, and the preset multi-dimensional capability benchmark data of the interpreters to be evaluated. The cross-modal feature alignment module is used to perform cross-modal feature alignment processing based on speech temporal features and text semantic features, and integrates speech fluency, language accuracy, logical integrity, on-the-spot adaptability and cross-cultural adaptability to build a hierarchical multi-dimensional ability evaluation model. The model calibration module is used to perform iterative calibration of the capability indicators of each dimension by combining a dynamic weight adaptive allocation mechanism, and at the same time introduces a feature bias normalization correction algorithm to eliminate cross-source data heterogeneity interference. The evaluation output module is used to output the comprehensive interpreting ability evaluation score, the dimension-specific ability defect map, and the ability improvement and optimization conclusions of the evaluated object based on the fusion calculation results of the multi-dimensional indicators after calibration and the hierarchical mapping rules of the ability model.

10. An interpreting assessment system based on a multidimensional competence model, characterized in that, The system also includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the interpreting assessment method based on the multidimensional competence model as described in any one of claims 1 to 8.

Images