Intelligent quantitative evaluation system and method for personal competitiveness of overseas study application

Abstract

The invention discloses an intelligent quantitative evaluation system and method for personal competitiveness of an overseas study application, and belongs to the technical field of text processing, and the method specifically comprises the steps: receiving a personal statement text and a recommended letter text of a user; capturing public recommendation information samples historically written by the recommender to construct an evaluation database; the occurrence frequency of the current evaluation phrase in the historical evaluation database is compared, and an evaluation intensity value is generated; the individual statement text is segmented into a plurality of narrative units, and the narrative units comprise a challenge description module, an action decision module and a result statement module which are arranged in sequence; detecting causal associated word density between the challenge description module and the result declaration module, and activating a logic closed-loop mark when a question mentioned in the challenge description module is solved by a quantitative index in the result declaration module; based on the evaluation intensity value and the logic closed-loop mark number distribution, outputting a hidden competitiveness index; according to the method, the precision and individuation requirements of overseas study application document evaluation are effectively met.

Description

Technical Field

[0001] This invention relates to the field of text processing technology, specifically to an intelligent quantitative assessment system and method for evaluating personal competitiveness in study abroad applications. Background Technology

[0002] In the context of intensifying global competition for educational resources, college application essays, as the core vehicle for showcasing applicants' comprehensive qualities, convey "implicit competitiveness" such as problem-solving abilities, innovative potential, and leadership skills, which have become key factors in university admissions decisions. Compared to quantitative indicators such as standardized test scores, the unstructured information contained in these essays better reflects the applicant's characteristics and potential. Therefore, in-depth analysis and quantitative evaluation of these texts have become an important technical direction for improving application accuracy and assisting universities in efficient screening.

[0003] In existing technologies, natural language processing (NLP) techniques have been gradually incorporated into the evaluation of college application documents. These techniques utilize keyword extraction, sentiment analysis, and topic clustering to achieve preliminary structuring of the document content. Simultaneously, some methods incorporate historical admission cases to establish correlation models between document characteristics and admission results, providing data support for the evaluation. These technologies, to some extent, reduce the subjectivity of manual evaluation, improve document screening efficiency, and have become the foundational technical solutions for college application support tools.

[0004] However, existing technologies still have significant limitations in analyzing the deep semantics of unstructured text in application documents: First, the interpretation of evaluative phrases in recommendation letters lacks personalized calibration. The actual effectiveness of the same evaluative phrase varies greatly depending on the recommender's expression habits, and existing methods mostly use fixed weight assignments, which are difficult to reflect the true strength of the evaluation. Second, there is insufficient analysis of the narrative logic of personal statements, making it impossible to accurately distinguish between roles such as "independent breakthrough" and "team support." Relying solely on keyword extraction is insufficient to determine the actual scenario and depth of ability. Third, existing natural language processing technologies mostly remain at the level of surface text feature matching, failing to combine in-depth analysis with the recommender's historical evaluation style and the completeness of the logical chain of the personal statement. This results in most implicit information in the text not being effectively utilized, ultimately causing the evaluation results to deviate from the university's true consideration of implicit competitiveness, making it difficult to meet the needs of precise and personalized assistance in study abroad applications. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent quantitative assessment system and method for personal competitiveness in study abroad applications, addressing the following technical problems:

[0006] Existing technologies lack personalized calibration of evaluation phrases in recommendation letters, struggle to analyze differences in narrative logic in personal statements, and merely rely on surface-level feature matching, resulting in the ineffective utilization of implicit information. Ultimately, this leads to a discrepancy between the evaluation results and the institutions' true assessment of implicit competitiveness.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] A method for intelligently quantifying the personal competitiveness of students applying for study abroad includes the following steps:

[0009] S1. Receive an unstructured text dataset submitted by a user, wherein the unstructured text dataset includes personal statement text and recommendation letter text from recommenders;

[0010] S2. Capture samples of publicly written recommendation letters from the recommender's past work, extract all evaluative phrases, and construct a database of the recommender's historical evaluations.

[0011] S3. Identify the evaluation phrases in the current recommendation letter, compare the frequency of the evaluation phrases with the frequency of occurrence in the recommender's historical evaluation database, and generate an evaluation intensity value that is inversely proportional to the frequency of occurrence.

[0012] S4. Divide the personal statement text into several narrative units, each containing a sequentially arranged challenge description module, action decision module, and outcome statement module;

[0013] S5. Detect the density of causal association words between the challenge description module and the result statement module. When the problem mentioned in the challenge description module is solved by quantitative indicators in the result statement module, activate the logical closed loop marker.

[0014] S6. Count the number of activated logical closed-loop markers in all narrative units, and output the implicit competitiveness index based on the evaluation intensity value and the distribution of the number of logical closed-loop markers.

[0015] As a further aspect of the present invention: in step S2, the process of constructing the recommender's historical evaluation database is as follows:

[0016] A grading system for evaluation phrases is established, which divides evaluation phrases into three levels according to their effectiveness: baseline phrases, enhanced phrases, and top-level phrases. The frequency percentage of enhanced phrases appearing in the recommender's historical samples is calculated. When the frequency percentage is lower than a preset threshold, the recommender's enhanced phrase is automatically upgraded to a top-level phrase. When a recommender adds new recommendation letter samples, the distribution ratio of phrases at each level is recalculated and the evaluation intensity value calculation rules are updated synchronously.

[0017] As a further aspect of the present invention: in step S3, the process of generating the evaluation intensity value is as follows:

[0018] The effectiveness level of the current evaluation phrase is retrieved from the recommender's historical evaluation database. If the current evaluation phrase is an enhanced phrase and its historical usage frequency is below the threshold, the evaluation strength value is calculated using the effectiveness coefficient of the top-level phrase. When the distribution of levels changes due to the addition of new recommendation letter samples, the evaluation strength value is recalculated using the updated effectiveness coefficient.

[0019] As a further aspect of the present invention: in step S4, the process of dividing the personal statement text into several narrative units is as follows:

[0020] Context boundary recognition is performed on personal statement texts to detect the set of transitional conjunctions and conclusion indicators in paragraphs; based on the location tags of transitional conjunctions and conclusion indicators, the text is segmented into continuous text segments, and each text segment is used as a candidate narrative unit;

[0021] Scan each candidate to see if it contains the set of noun keywords corresponding to the challenge description module, the set of verb keywords corresponding to the action decision module, and the set of quantitative keywords corresponding to the result statement module;

[0022] When a narrative unit candidate is missing the keyword set of any module, a predefined template framework is inserted at the module position. The template framework includes the standard question sentence of the challenge description module, the typical solution structure of the action decision module, and the quantitative indicator placeholders of the result statement module. The narrative unit candidate that completes the module is marked as a qualified narrative unit and output to step S5 for logical loop detection.

[0023] As a further aspect of the present invention: the specific process of activating the logic closed-loop marker in S5 is as follows:

[0024] In the challenge description module, problem type tags are identified. Problem type tags include three basic categories: resource scarcity, technical obstacles, and team conflict. In the action decision module, solution feature words are extracted and solution adaptation values ​​are generated based on the matching degree between solution feature words and problem type tags. In the result statement module, the occurrence status of numerical descriptive words is detected when verifying quantitative indicators.

[0025] When there is quantifiable improvement rate growth rate or percentage data, the result credibility marker is activated. When there are both a scheme adaptation value higher than the preset threshold and a result credibility marker in the same narrative unit, a logical closed loop marker is generated.

[0026] As a further aspect of the present invention: the process for generating the adaptation value of the solution is as follows:

[0027] A problem-solution mapping knowledge base is established, which stores the terms for fundraising and manpower expansion solutions corresponding to resource-scarce problems, and the terms for prototype iterative algorithm optimization solutions corresponding to technical obstacle problems. When the solution feature words in the action decision module match the preset solution words in the problem-solution mapping knowledge base, a basic fit score is assigned. When identifying the innovative description of the solution, an innovative identifier word set is detected. The innovative identifier word set includes pioneering reconstruction of cross-domain application vocabulary. An innovation bonus coefficient is added according to the level of the innovative identifier word in the innovation level table. The product of the basic fit score and the innovation bonus coefficient constitutes the solution fit value.

[0028] As a further aspect of the present invention: in S5, the proportion of units with activated logical closed-loop markers in all narrative units is counted as the effective event chain proportion value; narrative units without activated logical closed-loop markers are identified as narrative units to be optimized; and the original solution feature words of the action decision module are located in the narrative units to be optimized.

[0029] From the problem solution mapping knowledge base, retrieve a set of candidate solution words that match the problem type label of the narrative unit to be optimized, and select the candidate solution word with the highest solution fit value in the candidate solution word set; replace the original solution feature word with the candidate solution word with the highest solution fit value to generate the updated narrative unit; and re-perform logical loop detection on the updated narrative unit.

[0030] As a further aspect of the present invention: In step S6, text confidence verification is performed before outputting the implicit competitiveness index.

[0031] A cross-validation channel is established between the personal statement text and the recommendation letter text. When a certain ability mentioned in the recommendation letter is used as the core action basis in the event chain of the personal statement, the logical closed loop marker corresponding to the event chain is located. The parsing details of the logical closed loop marker are extracted and semantic consistency is checked with the ability description in the recommendation letter. The confidence weight of the implicit competitiveness index is adjusted according to the semantic consistency detection results.

[0032] As a further aspect of the present invention: the semantic consistency detection process is as follows:

[0033] Action verbs and recipients are extracted from the description of abilities in the recommendation letter to form ability triples. These ability triples include the action recipient and the effect element. When searching for matching ability triples in the action decision module of the event chain corresponding to the personal statement, the action verbs must be in the same tense and the recipient categories must overlap. When there are numerical differences in the effect descriptions, the recommendation letter value is used as the benchmark to calibrate the personal statement result data. The product of the number of successful matching of ability triples and the numerical calibration magnitude determines the semantic consistency score.

[0034] This invention also includes an intelligent quantitative assessment system for personal competitiveness in study abroad applications, used to implement the aforementioned intelligent quantitative assessment method for personal competitiveness in study abroad applications, comprising:

[0035] The data acquisition module is used to receive unstructured text datasets submitted by users, which include personal statement texts and recommendation letters from recommenders.

[0036] The database construction module is used to crawl samples of publicly written recommendation letters from recommenders in the past and extract all evaluative phrases to build a database of historical evaluations of recommenders;

[0037] The evaluation optimization module is used to identify evaluation phrases in the current recommendation letter, compare the frequency of occurrence of the evaluation phrases in the recommender's historical evaluation database, and generate an evaluation intensity value that is inversely proportional to the frequency of occurrence.

[0038] The text analysis module is used to segment the personal statement text into several narrative units, each containing a sequentially arranged challenge description module, action decision module, and outcome statement module;

[0039] The logic analysis module is used to detect the density of causal association words between the challenge description module and the result statement module. When the problem mentioned in the challenge description module is solved by quantitative indicators in the result statement module, the logic closed loop marker is activated.

[0040] The competitiveness output module is used to count the number of activated logical closed-loop markers in all narrative units, and output the implicit competitiveness index based on the evaluation intensity value and the distribution of the number of logical closed-loop markers.

[0041] The beneficial effects of this invention are:

[0042] This invention addresses the issue of misjudgment of the effectiveness of the same evaluation phrase in recommendation letters due to different recommender habits by constructing a historical evaluation database of recommenders and a personalized evaluation lexicon. It combines the frequency of evaluation phrase occurrences with a grading system to dynamically calibrate evaluation intensity values, achieving precise quantification of evaluation intensity. By segmenting personal statements into narrative units containing challenge, action, and result modules and supplementing the template framework, it activates logical loop markers using problem type tags, solution fit values, and result credibility markers, solving the problem of insufficient narrative logic analysis in personal statements and accurately identifying the complete event chain of "challenge-action-result." By using a problem-solution mapping knowledge base and innovative identifier words to calculate solution fit values, it improves the matching degree between solutions and problems and the accuracy of innovativeness assessment. Through text confidence verification and semantic consistency detection of ability triples, it strengthens the cross-validation of personal statements and recommendation letters, improving the credibility of the evaluation results. By identifying narrative units to be optimized and replacing them with high-fit-value candidate solution words, it optimizes the narrative density and quality of the documents. The final output implicit competitiveness index accurately reflects the applicant's implicit competitiveness, effectively meeting the precise and personalized needs of study abroad application document evaluation. Attached Figure Description

[0043] The invention will now be further described with reference to the accompanying drawings.

[0044] Figure 1 This is a flowchart illustrating an intelligent quantitative assessment method for personal competitiveness in study abroad applications, as described in this invention.

[0045] Figure 2 This is a schematic diagram of the structure of an intelligent quantitative assessment system for personal competitiveness in study abroad applications according to the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] Please see Figure 1 As shown, this invention is an intelligent quantitative assessment method for personal competitiveness in study abroad applications, comprising the following steps:

[0048] S1. The first step is to receive unstructured text datasets submitted by users. During this process, applicants need to upload their personal statement and letters of recommendation. The personal statement typically includes the applicant's academic background, research experience, and extracurricular activities, serving as an important vehicle for showcasing the applicant. Letters of recommendation, written by the recommenders, contain evaluations of the applicant's academic abilities, professional competence, and personal qualities. These two types of texts together constitute the foundational data for the evaluation.

[0049] S2. Next, sample publicly available recommendation letters written by the recommender in the past are collected. By collecting these letters, all evaluative phrases are extracted to construct a historical evaluation database of the recommender. This database reflects the recommender's evaluation style and habits, providing a reference for calculating the subsequent evaluation strength value.

[0050] S3. Next, identify the evaluative phrases in the current recommendation letter. Compare the evaluative phrases in the current recommendation letter with phrases in the recommender's historical evaluation database, and calculate the frequency of each evaluative phrase in the historical samples. Generate an evaluation strength value based on the frequency of occurrence; the higher the frequency, the lower the evaluation strength value, and vice versa. This inverse relationship can more accurately reflect the recommender's true evaluation of the applicant.

[0051] S4. Next, divide the personal statement text into several narrative units. Each narrative unit contains a challenge description module, an action decision module, and a result statement module arranged in sequence. The challenge description module describes the difficulties and challenges faced by the applicant in a project or activity; the action decision module explains the specific measures and methods taken by the applicant to solve the problems; and the result statement module shows the results and gains achieved by the applicant through the actions.

[0052] S5. Next, the density of causal relationship words between the challenge description module and the outcome statement module is detected. When the problem mentioned in the challenge description module is solved by quantifiable indicators in the outcome statement module, a logical closure mark is activated. This process verifies the completeness of the applicant's narrative logic and ensures that the described experience is authentic and credible.

[0053] S6. Finally, count the number of activated logical loop markers in all narrative units. Based on the evaluation intensity value and the distribution of the number of logical loop markers, calculate and output the implicit competitiveness index using a weighted average. This index comprehensively reflects the applicant's academic ability, practical ability, innovation ability, and other qualities, providing an objective and scientific evaluation basis for study abroad applications.

[0054] In step S2, the process of constructing the recommender's historical evaluation database is as follows:

[0055] First, a tiered system of evaluative phrases should be established, dividing all evaluative phrases extracted from the recommender's historical recommendation letters into three levels based on their strength. The baseline level consists of the most commonly used basic evaluative expressions in recommendations, such as "good performance" and "responsible." These phrases appear frequently in most recommendation letters and have relatively basic effectiveness. The reinforcement level consists of more affirmative expressions than the baseline level, such as "outstanding contributions" and "exceptional abilities," and their strength is higher than the baseline level. The top-level phrases are rarely used but carry significant weight in recommendations, such as "top talent" and "rare find," representing the recommender's highest level of recognition.

[0056] Based on the grading, the frequency ratio of reinforcing phrases in all historical recommendation letter samples of the recommender needs to be calculated, that is, the ratio of the total number of times reinforcing phrases appeared to the total number of times all evaluative phrases of the recommender appeared. When this ratio is lower than a pre-set threshold, it indicates that the recommender rarely uses reinforcing phrases, and the actual effectiveness of their reinforcing phrases is close to that of top-level phrases. In this case, the recommender's reinforcing phrases should be automatically upgraded to the effectiveness of top-level phrases to match their actual evaluation habits.

[0057] Meanwhile, the database needs to be dynamically updated. When a recommender adds new recommendation letter samples, all historical samples must be reviewed again, and the distribution ratios of baseline, reinforcement, and top-level phrases in the total evaluation phrases must be recalculated. Based on the new distribution ratios, the calculation rules for evaluation intensity values ​​should be updated synchronously to ensure that the database always reflects the recommender's latest evaluation style and avoids distortion of evaluation standards due to outdated samples.

[0058] In step S3, the process of generating the evaluation intensity value is as follows:

[0059] Based on a database of historical recommender evaluations, this system achieves precise quantification of evaluation phrases in current recommendation letters. First, it retrieves the effectiveness level of each evaluation phrase in the current recommendation letter from the database, clarifying its initial evaluation strength benchmark. If the search reveals that the current evaluation phrase belongs to the reinforcement level, and historical data shows that the phrase's usage frequency in the recommender's past evaluations is below a preset threshold, then according to previously established rules, its evaluation strength value is calculated based on the effectiveness coefficient of top-level phrases, thus reflecting the actual weight of reinforcement-level phrases in the recommender's evaluation.

[0060] Furthermore, the calculation of evaluation intensity values ​​needs to be dynamically adjusted. When a recommender adds new recommendation letter samples, causing changes in the distribution ratio of baseline, reinforcement, and top-level phrases, the effectiveness coefficients at each level used to calculate the evaluation intensity values ​​must also be updated accordingly. In this case, the updated effectiveness coefficients must be used to recalculate the evaluation intensity values ​​for the evaluation phrases in the current recommendation letters, ensuring that the evaluation results match the recommender's latest evaluation tendencies in real time. This dynamic calibration mechanism avoids misjudgments caused by changes in the recommender's evaluation style and ensures that the evaluation intensity values ​​always remain consistent with the recommender's true level of approval, providing a reliable basis for subsequent implicit competitiveness assessments.

[0061] In S4, the process of dividing the personal statement text into several narrative units is as follows:

[0062] First, contextual boundary identification is performed on the personal statement text. Natural dividing points are located by retrieving sets of transitional conjunctions and conclusion indicators within paragraphs. Transitional conjunctions include words such as "however," "in addition," and "on the other hand," which often mark shifts in narrative scenes or events. Conclusion indicators include words such as "finally," "therefore," and "the result is," which are mostly used to summarize the achievements of a certain stage of action. Based on the location markers of these words, the text is segmented into continuous text segments, each of which serves as a candidate narrative unit, thus initially achieving a structured decomposition of complex narratives.

[0063] Subsequently, a module integrity check is performed on each candidate. Specifically, the candidate is scanned to see if it contains three sets of keywords: a set of noun keywords corresponding to the challenge description module, such as "problem," "dilemma," "limitation," and "obstacle," used to identify the specific challenges faced by the applicant; a set of verb keywords corresponding to the action decision module, such as "implement," "design," "coordinate," and "optimize," used to reflect the measures taken by the applicant; and a set of quantitative keywords corresponding to the result statement module, such as "improve," "reduce," "achieve," and "realize," used to explain the specific results brought about by the action.

[0064] When a narrative unit candidate lacks the keyword set for any module, it needs to be supplemented using a predefined template framework. Standard question formats for the challenge description module include phrases like "During the XX process, we faced limitations in XX aspects" or "Due to XX factors, XX type of problem arose," helping applicants to standardize their description of the challenge background. Typical solution structures for the action decision module include phrases like "Through XX method, we took XX specific measures" or "For XX problem, we formulated and implemented XX solution," guiding a clear presentation of the action logic. Quantitative indicator placeholders for the results statement module include phrases like "Ultimately, we achieved XX improvement" or "This resulted in XX efficiency changes," reserving space for filling in specific data. After template supplementation, ensuring each narrative unit candidate contains all three modules, it is then marked as a qualified narrative unit and output to the next step for logical loop detection.

[0065] In S5, the specific process of activating the logic closed-loop marker is as follows:

[0066] First, the challenge description module identifies problem type tags, specifically divided into three basic categories: resource-scarce, such as descriptions involving "insufficient funds," "manpower shortage," and "limited equipment"; technical obstacle, such as those including "system vulnerabilities," "algorithm bottlenecks," and "technical barriers"; and team conflict, such as mentioning scenarios like "disagreements," "inefficient collaboration," and "communication barriers." This tagging process clarifies the core attributes of each challenge, providing a basis for subsequent solution matching.

[0067] Next, solution feature words are extracted from the action decision module, and solution adaptation values ​​are generated based on the matching degree between these feature words and problem type labels. This process relies on a pre-established problem-solution mapping knowledge base, which stores typical solution words corresponding to different problem types: resource-scarce problems correspond to solution words such as "fundraising," "human resource expansion," and "resource integration"; technical obstacle problems correspond to solution words such as "prototype iteration," "algorithm optimization," and "technical breakthrough." When the solution feature words in the action decision module match the preset solution words in the knowledge base, a basic adaptation score is assigned, and the basic score is dynamically adjusted according to the accuracy of the match.

[0068] Simultaneously, it is necessary to identify innovative descriptions within the solution and add an innovation bonus coefficient by detecting a set of innovative identifiers. This set of innovative identifiers includes terms such as "first-of-its-kind," "restructuring," "cross-domain application," and "breakthrough," which are assigned different levels in the innovation level table, with "first-of-its-kind" corresponding to the highest level, followed by "cross-domain application." Based on the level of the innovative identifier, a corresponding innovation bonus coefficient is added to the basic fit score; the product of the two is the final solution fit value, comprehensively reflecting the solution's relevance and innovation.

[0069] Finally, the validity of quantitative indicators is verified in the results presentation module by detecting the presence of numerical descriptive terms to determine the credibility of the results. When quantifiable improvement rates, growth rates, or percentage data are present in the text, such as "improving efficiency by XX", "reducing costs by XX", or "achieving market share of XX", a result credibility marker is activated to confirm the specificity and verifiability of the results. When both the solution fit value and the result credibility marker exist within the same narrative unit, it indicates that the unit fully presents the closed-loop logic of "encountering a problem - taking measures - achieving results". At this time, a logical closed-loop marker is generated to identify the narrative unit as a valid event chain, which can serve as an important basis for evaluating the applicant's implicit competitiveness.

[0070] S5 further includes counting the number of units with activated logical closed-loop markers in all narrative units, calculating the ratio of these units to the total number of narrative units, and using this ratio as the effective event chain ratio to measure the proportion of complete and persuasive narratives in the personal statement. Simultaneously, narrative units without activated logical closed-loop markers are identified and marked as narrative units to be optimized, as these units often have incomplete narrative logic or insufficient persuasiveness.

[0071] For the narrative unit to be optimized, the original solution feature words in its action decision module are further located, which are the core words originally used to describe the response measures. Then, the set of candidate solution words that match the problem type label of the unit to be optimized is retrieved from the problem solution mapping knowledge base. For example, if the problem type is resource scarcity, solution words related to resource integration are retrieved; if it is technical obstacle, solution words related to technical breakthroughs are retrieved.

[0072] In the candidate solution terminology set, the candidate solution term with the highest solution fit value is selected. The original solution feature terms in the unit to be optimized are replaced with this candidate term with a high fit value, generating an updated narrative unit. After the replacement is completed, the updated narrative unit is re-tested for logical loop closure to determine whether it meets the conditions for activating the logical loop closure marker, thereby increasing the proportion of effective narrative in the personal statement and enhancing the presentation of the applicant's abilities in the document.

[0073] In step S6, text confidence verification is performed before outputting the implicit competitiveness index:

[0074] Before outputting the implicit competitiveness index, text confidence verification is required to ensure that the information presented in the personal statement and recommendation letters corroborates each other, thus improving the reliability of the evaluation results. During verification, a cross-validation channel is first established between the personal statement and recommendation letter texts. This involves analyzing the consistency of information by identifying the same abilities or experiences mentioned in both texts. Specifically, when a recommendation letter mentions an applicant's ability, such as "project coordination skills" or "technical problem-solving skills," and this ability serves as a core action in the event chain of the personal statement, the logical closed-loop marker corresponding to this event chain must be accurately located. This marker is linked to the complete analytical details of the "challenge-action-result" process in the event chain, including the specific methods of action decision-making, the key steps in problem-solving, and the final achieved results.

[0075] Next, these parsing details are extracted and semantically consistent with the description of the ability in the recommendation letter. The core of the detection is to determine whether the two types of texts are consistent in logic, scope, and degree regarding the same ability, avoiding information conflicts or ambiguities. Based on the detection results, the confidence weight of the implicit competitiveness index is adjusted: if the consistency is high, the confidence of the index is increased; if there are significant differences, the weight is decreased, thus reflecting the fluctuation of information credibility.

[0076] The semantic consistency detection process requires multi-dimensional comparison to ensure the accuracy of the assessment. During the detection, key elements are first extracted from the competency description in the recommendation letter, including action verbs and the recipient, thus forming competency triplets. These triplets fully encompass three core elements: the specific behavior (e.g., "leading," "optimizing," "coordinating"), the target of the behavior (e.g., "cross-departmental projects," "algorithm models," "team disagreements"), and the effects of the behavior (e.g., "efficiency improvement," "conflict resolution," "results implementation").

[0077] Subsequently, in the action decision-making module corresponding to the event chain in the personal statement, content matching the triplet of that ability is searched. Two core requirements must be met during the matching process: First, the tenses of the action verbs must be consistent. For example, if the recommendation letter uses "led" (past tense), the corresponding expression in the personal statement should also use the past tense to ensure timeline consistency. Second, the scope of the subject matter must overlap; that is, the action subject mentioned in the personal statement must belong to the same field or type as the subject described in the recommendation letter, avoiding scope discrepancies such as "the recommendation letter describes a 'technical project' while the personal statement corresponds to 'marketing activities'."

[0078] If the two types of texts differ numerically in their descriptions of effects—for example, a recommendation letter mentions a "30% cost reduction," while a personal statement states a "20% cost reduction"—then the value in the recommendation letter is used as the baseline to calibrate the results data in the personal statement, ensuring consistency in the quantitative description of the same outcome. Ultimately, the semantic consistency score is determined by two parts: first, the number of successful matches of the capability triples, i.e., the number of consistent element combinations in both types of texts; and second, the magnitude of the numerical calibration, i.e., the degree of difference between the personal statement data and the baseline value in the recommendation letter. The product of these two factors constitutes the final semantic consistency score, directly reflecting the degree of agreement between the two types of text information and providing a clear basis for subsequently adjusting the confidence weights of the implicit competitiveness index.

[0079] Please see Figure 2 As shown, the present invention also includes an intelligent quantitative assessment system for personal competitiveness in study abroad applications, used to implement the above-described intelligent quantitative assessment method for personal competitiveness in study abroad applications, comprising:

[0080] The data acquisition module is used to receive unstructured text datasets submitted by users, which include personal statement texts and recommendation letters from recommenders.

[0081] The database construction module is used to crawl samples of publicly written recommendation letters from recommenders in the past and extract all evaluative phrases to build a database of historical evaluations of recommenders;

[0082] The evaluation optimization module is used to identify evaluation phrases in the current recommendation letter, compare the frequency of occurrence of the evaluation phrases in the recommender's historical evaluation database, and generate an evaluation intensity value that is inversely proportional to the frequency of occurrence.

[0083] The text analysis module is used to segment the personal statement text into several narrative units, each containing a sequentially arranged challenge description module, action decision module, and outcome statement module;

[0084] The logic analysis module is used to detect the density of causal association words between the challenge description module and the result statement module. When the problem mentioned in the challenge description module is solved by quantitative indicators in the result statement module, the logic closed loop marker is activated.

[0085] The competitiveness output module is used to count the number of activated logical closed-loop markers in all narrative units, and output the implicit competitiveness index based on the evaluation intensity value and the distribution of the number of logical closed-loop markers.

[0086] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for intelligent quantitative assessment of personal competitiveness in study abroad applications, characterized in that, Includes the following steps: S1. Receive an unstructured text dataset submitted by a user, wherein the unstructured text dataset includes personal statement text and recommendation letter text from recommenders; S2. Capture samples of publicly written recommendation letters from the recommender's past work, extract all evaluative phrases, and construct a database of the recommender's historical evaluations. S3. Identify the evaluation phrases in the current recommendation letter, compare the frequency of the evaluation phrases with the frequency of occurrence in the recommender's historical evaluation database, and generate an evaluation intensity value that is inversely proportional to the frequency of occurrence. S4. Divide the personal statement text into several narrative units, each containing a sequentially arranged challenge description module, action decision module, and outcome statement module; S5. Detect the density of causal association words between the challenge description module and the result statement module. When the problem mentioned in the challenge description module is solved by quantitative indicators in the result statement module, activate the logical closed loop marker. S6. Count the number of activated logical closed-loop markers in all narrative units, and output the implicit competitiveness index based on the evaluation intensity value and the distribution of the number of logical closed-loop markers.

2. The intelligent quantitative assessment method for personal competitiveness in study abroad applications according to claim 1, characterized in that, In step S2, the process of constructing the recommender's historical evaluation database is as follows: A grading system for evaluation phrases is established, which divides evaluation phrases into three levels according to their effectiveness: baseline phrases, enhanced phrases, and top-level phrases. The frequency percentage of enhanced phrases appearing in the recommender's historical samples is calculated. When the frequency percentage is lower than a preset threshold, the recommender's enhanced phrase is automatically upgraded to a top-level phrase. When a recommender adds new recommendation letter samples, the distribution ratio of phrases at each level is recalculated and the evaluation intensity value calculation rules are updated synchronously.

3. The intelligent quantitative assessment method for personal competitiveness in study abroad applications according to claim 2, characterized in that, In step S3, the process of generating the evaluation intensity value is as follows: The effectiveness level of the current evaluation phrase is retrieved from the recommender's historical evaluation database. If the current evaluation phrase is an enhanced phrase and its historical usage frequency is below the threshold, the evaluation strength value is calculated using the effectiveness coefficient of the top-level phrase. When the distribution of levels changes due to the addition of new recommendation letter samples, the evaluation strength value is recalculated using the updated effectiveness coefficient.

4. The intelligent quantitative assessment method for personal competitiveness in study abroad applications according to claim 1, characterized in that, In S4, the process of dividing the personal statement text into several narrative units is as follows: Context boundary recognition is performed on personal statement texts to detect the set of transitional conjunctions and conclusion indicators in paragraphs; based on the location tags of transitional conjunctions and conclusion indicators, the text is segmented into continuous text segments, and each text segment is used as a candidate narrative unit; Scan each candidate to see if it contains the set of noun keywords corresponding to the challenge description module, the set of verb keywords corresponding to the action decision module, and the set of quantitative keywords corresponding to the result statement module; When a narrative unit candidate is missing the keyword set of any module, a predefined template framework is inserted at the module position. The template framework includes the standard question sentence of the challenge description module, the typical solution structure of the action decision module, and the quantitative indicator placeholders of the result statement module. The narrative unit candidate that completes the module is marked as a qualified narrative unit and output to step S5 for logical loop detection.

5. The intelligent quantitative assessment method for personal competitiveness in study abroad applications according to claim 1, characterized in that, In S5, the specific process of activating the logic closed-loop marker is as follows: In the challenge description module, problem type tags are identified. Problem type tags include three basic categories: resource scarcity, technical obstacles, and team conflict. In the action decision module, solution feature words are extracted and solution adaptation values ​​are generated based on the matching degree between solution feature words and problem type tags. In the result statement module, the occurrence status of numerical descriptive words is detected when verifying quantitative indicators. When there is quantifiable improvement rate growth rate or percentage data, the result credibility marker is activated. When there are both a scheme adaptation value higher than the preset threshold and a result credibility marker in the same narrative unit, a logical closed loop marker is generated.

6. The intelligent quantitative assessment method for personal competitiveness in study abroad applications according to claim 5, characterized in that, The process for generating the adaptation value of the scheme is as follows: A problem-solution mapping knowledge base is established, which stores the terms for fundraising and manpower expansion solutions corresponding to resource-scarce problems, and the terms for prototype iterative algorithm optimization solutions corresponding to technical obstacle problems. When the solution feature words in the action decision module match the preset solution words in the problem-solution mapping knowledge base, a basic fit score is assigned. When identifying the innovative description of the solution, an innovative identifier word set is detected. The innovative identifier word set includes pioneering reconstruction of cross-domain application vocabulary. An innovation bonus coefficient is added according to the level of the innovative identifier word in the innovation level table. The product of the basic fit score and the innovation bonus coefficient constitutes the solution fit value.

7. The intelligent quantitative assessment method for personal competitiveness in study abroad applications according to claim 6, characterized in that, In S5, the proportion of units with activated logical closed-loop markers in all narrative units is counted as the effective event chain proportion value. Identify narrative units with inactive logical closed-loop markers as narrative units to be optimized; locate the original solution feature words of the action decision module in the narrative units to be optimized; From the problem solution mapping knowledge base, retrieve a set of candidate solution words that match the problem type tag of the narrative unit to be optimized, and select the candidate solution word with the highest solution fit value in the candidate solution word set; replace the original solution feature word with the candidate solution word with the highest solution fit value to generate the updated narrative unit; The updated narrative units were re-tested for logical closure.

8. The intelligent quantitative assessment method for personal competitiveness in study abroad applications according to claim 1, characterized in that, In step S6, text confidence verification is performed before outputting the implicit competitiveness index: A cross-validation channel is established between the personal statement text and the recommendation letter text. When a certain ability mentioned in the recommendation letter is used as the core action basis in the event chain of the personal statement, the logical closed loop marker corresponding to the event chain is located. The parsing details of the logical closed loop marker are extracted and semantic consistency is checked with the ability description in the recommendation letter. The confidence weight of the implicit competitiveness index is adjusted according to the semantic consistency detection results.

9. The intelligent quantitative assessment method for personal competitiveness in study abroad applications according to claim 8, characterized in that, The semantic consistency detection process is as follows: Action verbs and recipients are extracted from the description of abilities in the recommendation letter to form ability triples. These ability triples include the action recipient and the effect element. When searching for matching ability triples in the action decision module of the event chain corresponding to the personal statement, the action verbs must be in the same tense and the recipient categories must overlap. When there are numerical differences in the effect descriptions, the recommendation letter value is used as the benchmark to calibrate the personal statement result data. The product of the number of successful matching of ability triples and the numerical calibration magnitude determines the semantic consistency score.

10. A smart quantitative assessment system for personal competitiveness in study abroad applications, used to implement the smart quantitative assessment method for personal competitiveness in study abroad applications as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to receive unstructured text datasets submitted by users, which include personal statement texts and recommendation letters from recommenders. The database construction module is used to crawl samples of publicly written recommendation letters from recommenders in the past and extract all evaluative phrases to build a database of historical evaluations of recommenders; The evaluation optimization module is used to identify evaluation phrases in the current recommendation letter, compare the frequency of occurrence of the evaluation phrases in the recommender's historical evaluation database, and generate an evaluation intensity value that is inversely proportional to the frequency of occurrence. The text analysis module is used to segment the personal statement text into several narrative units, each containing a sequentially arranged challenge description module, action decision module, and outcome statement module; The logic analysis module is used to detect the density of causal association words between the challenge description module and the result statement module. When the problem mentioned in the challenge description module is solved by quantitative indicators in the result statement module, the logic closed loop marker is activated. The competitiveness output module is used to count the number of activated logical closed-loop markers in all narrative units, and output the implicit competitiveness index based on the evaluation intensity value and the distribution of the number of logical closed-loop markers.

Images