Self-adaptive test method, device, equipment, medium and program product

Abstract

The embodiment of the invention provides a self-adaptive testing method and device, equipment, a medium and a program product. According to the scheme, a scale used for self-adaptive testing comprises a plurality of question groups divided into a plurality of sequentially associated measurement stages, and each question group comprises one or more questions; the method comprises the following steps: firstly, acquiring a first question answering result of a user for a first question group in a first measurement stage in a plurality of measurement stages; then, according to the first answer result, determining a first answer score of the user; determining a second question group from two or more question groups in a second measurement stage based on the first answer score according to a corresponding relationship between the answer score and the question group, the second measurement stage being a next measurement stage after the first measurement stage; and then, providing the second question group for the user. Therefore, the scale testing method with higher flexibility is provided.

Description

Technical Field

[0001] This specification relates to the field of scale measurement technology, and in particular to an adaptive testing method, apparatus, computing device, computer-readable storage medium, and computer program. Background Technology

[0002] Fixed-item scales are a traditional measurement technique that assesses test-takers using a fixed set of questions. This method typically involves a series of standardized questions, and test-takers must select the appropriate answers based on their own responses to arrive at an evaluation result. Fixed-item scales are very simple to design; each question's answer is usually assigned a score, and the scores are summed to obtain the total scale score. The advantage of fixed-item scales is their simplicity and ease of use, but the fact that all users provide identical answers results in a lack of personalization and targeted assessment.

[0003] With the development of technology, Computer Adaptive Tests (CAT) are increasingly being applied to scale measurement. CAT is a personalized measurement method based on computer algorithms, typically using Item Response Theory (IRT) as its basic model. It can automatically adjust the next questions based on the test taker's responses, achieving a certain degree of personalization and targeting. However, in existing CAT methods, the user is scored after each question to determine the next question, meaning the user has no opportunity to modify their answers, resulting in poor flexibility.

[0004] Therefore, there is a need to provide more flexible scale testing methods. Summary of the Invention

[0005] In view of this, embodiments of this specification provide an adaptive testing method. One or more embodiments of this specification also relate to an adaptive testing apparatus, a computing device, a computer-readable storage medium, and a computer program, to address the technical deficiencies existing in the prior art.

[0006] According to a first aspect of the embodiments of this specification, an adaptive testing method is provided, wherein the scale for adaptive testing includes multiple item groups divided into sequentially associated measurement phases, each item group including one or more items; the method includes:

[0007] Obtain the user's first answer result for the first set of questions in the first measurement stage of the multiple measurement stages;

[0008] Based on the first answer result, determine the user's first answer score;

[0009] Based on the correspondence between the answer scores and the question groups, and based on the first answer score, a second question group is determined from two or more question groups in the second measurement stage; the second measurement stage is the next measurement stage after the first measurement stage.

[0010] The second set of questions is provided to the user.

[0011] According to a second aspect of the embodiments of this specification, an adaptive testing apparatus is provided, wherein the scale for adaptive testing includes multiple question groups divided into sequentially associated multiple measurement phases, each question group including one or more questions; the apparatus includes:

[0012] The answer result acquisition module is used to acquire the user's first answer result for the first question group in the first measurement stage of the multiple measurement stages;

[0013] The answer score determination module is used to determine the user's first answer score based on the first answer result;

[0014] The question group routing module is used to determine a second question group from two or more question groups in the second measurement stage based on the first answer score, according to the correspondence between the answer score and the question group; the second measurement stage is the next measurement stage after the first measurement stage;

[0015] The question set providing module is used to provide the second question set to the user.

[0016] According to a third aspect of the embodiments of this specification, a computing device is provided, comprising:

[0017] Memory and processor;

[0018] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the adaptive testing method described above.

[0019] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided that stores computer-executable instructions that, when executed by a processor, implement the steps of the adaptive testing method described above.

[0020] According to a fifth aspect of the embodiments of this specification, a computer program is provided, wherein when the computer program is executed in a computer, it causes the computer to perform the steps of the adaptive testing method described above.

[0021] One embodiment of this specification can achieve at least the following beneficial effects: by using a scale that includes multiple question groups and multiple measurement stages, after a user completes the answers to the questions in the question group of the previous measurement stage, the user is routed to a question group in the next measurement stage that is suitable for the user to answer, based on the user's scores on the questions already answered. Thus, by determining the question group that the user needs to answer in the next measurement stage based on the user's scores on the question group in the previous measurement stage, the questions provided to the user are more targeted to the user, thereby making the measurement results for the user more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A schematic diagram of the structure of a scale for adaptive testing provided in the embodiments of this specification;

[0023] Figure 2 A flowchart illustrating an adaptive testing method provided in an embodiment of this specification;

[0024] Figure 3 A flowchart illustrating the processing procedure of an adaptive testing method applied to psychological testing, provided as an embodiment of this specification;

[0025] Figure 4 This is a schematic diagram of the structure of an adaptive testing device provided in the embodiments of this specification;

[0026] Figure 5 This is a structural block diagram of a computing device provided as an embodiment of this specification. Detailed Implementation

[0027] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0028] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a,” “described,” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0029] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0030] Furthermore, 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 one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0031] Rating scales are a method of measurement research that uses scales to measure subjects. A scale is a measuring standard, a series of test items (tasks or questions) designed according to the purpose of the measurement. Each task or question has a pre-defined standard score. When using a scale, the scores are calculated based on the subjects' responses. The subject's scale score indicates the subject's position on the scale, the meaning of which is explained by the researcher when the scale was developed.

[0032] The application of scale measurement techniques is wide-ranging, including but not limited to psychological state measurement scales (also known as mental health measurement scales or mental health scales), personality testing scales, intelligence scales, human resource management scales, and marital psychological locus of control scales. For example, psychological state measurement scales may include the Symptom Checklist-90 (SCL-90), the Self-Rating Anxiety Scale (SAS), and the Self-Rating Depression Scale (SDS). Similarly, personality testing scales may include the EPQ personality test, the Cattell 16 Personality Factor Questionnaire (16PF), temperament tests, sexual orientation tests, the Minnesota Multiphasic Personality Inventory (MMPI), and mood projective tests. Furthermore, intelligence scales may include the Wechsler Intelligence Scale, Raven's Progressive Matrices, and the Binet-Simon Intelligence Scale.

[0033] Among these tools, mental health scales are effective in assessing an individual's mental health status. They can help people better understand their mental state and take timely and necessary measures. By completing mental health scales, people can objectively assess their emotions, stress levels, anxiety levels, and other aspects, thereby identifying potential mental health problems and seeking professional help in a timely manner. Furthermore, mental health scales can help professionals gain a more comprehensive understanding of an individual's mental health status, providing a reference for personalized mental health interventions.

[0034] Scale measurement techniques typically include two methods: fixed-item measurement and adaptive measurement.

[0035] Fixed-item measurement is a traditional measurement technique that assesses a test subject's mental health using a fixed set of questions. It typically involves a series of standardized questions, with test subjects selecting appropriate answers based on their own responses to arrive at an assessment result. Scoring on fixed-item scales is simple; each question's answer is assigned a score, and the scores are summed to obtain the total scale score. However, fixed-item measurement suffers from a lack of personalization and specificity, and may also have some questions unsuitable for all users, as well as a large number of questions.

[0036] Taking mental health scales as an example, the use of fixed-item measurement may have the following problems. Firstly, because the items are fixed, users need to answer all questions in the scale; however, not all users are suited to answer all questions. Mental health scales often include symptoms of psychological problems, some of which may embarrass or offend users without symptoms, making them unwilling to answer. If users experience psychological resistance or interruption during the answering process, the answers may not reflect the actual situation of the test-taker, thus rendering the scale's results ineffective in providing reference and help. Secondly, the fixed-item technique is mainly based on classical measurement theory, reducing measurement error by synthesizing users' responses to multiple items. Mental health scales using fixed measurement techniques, to ensure the accuracy of the results (achieving good reliability and validity), usually need to include a large number of items. Specifically, mental health encompasses multiple aspects such as depression, anxiety, and obsessive-compulsive disorder; scales that comprehensively assess mental health status generally have a relatively large number of items. In this case, the burden and time cost of using the scale are relatively high. Too many items may lead to participant fatigue and impatience, affecting their seriousness in answering questions, and thus affecting the reliability of the scale.

[0037] With the development of technology, Computer Adaptive Test (CAT) has been gradually applied in various scale measurement fields. CAT is a personalized measurement method based on computer algorithms. It is usually based on Item Response Theory (IRT) and can automatically adjust the next questions based on the test taker's responses to achieve more accurate and effective measurement.

[0038] Item Response Theory (IRT), also known as latent trait theory or latent trait model, is a modern measurement theory proposed to address the limitations of Classical Test Theory (CTT). Its significance lies in guiding item selection and test development. IRT assumes that test takers possess a "latent trait," a statistical construct derived from the observation and analysis of test responses. In testing, latent trait generally refers to potential abilities, and the total test score is often used as an estimate of this potential. IRT posits a specific relationship between test takers' responses and performance on test items and their latent trait. Item parameters established through IRT possess the characteristic of permanence, meaning that scores across different measurement scales can be standardized.

[0039] In many educational and psychometric contexts, there is an underlying variable that is often intuitively understandable, such as descriptive terms like "intelligence," "reading ability," and "arithmetic ability." These are unobservable, latent characteristics because the variable is a concept rather than a physical dimension. While easy to describe, it cannot be directly measured like height or weight. The primary goal of educational and psychometric measurements is to determine how much of this latent characteristic a person possesses. Since most research involves variables such as academic, reading, mathematical, and arithmetic abilities, the general term "ability" is used in item response theory to refer to these latent characteristics. To measure how much of a latent characteristic a person possesses, a measurement scale is needed. The basic idea is that if a person's ability can be physically determined, this scale can be used to judge how much ability a person has, and the abilities of several people can be compared. Theoretically, the range of ability is from negative infinity to positive infinity, but in practical considerations, the range of values ​​is usually limited to, for example, between 3 and -3. IRT uses a function of the subject's ability value θ and the difficulty of the question δ to represent the probability of the subject answering the question correctly. This function defines a probability model, and the graph of the function approximates an S-shaped curve, with values ​​between 0 and 1. The higher the subject's ability value, the higher the corresponding function value. In other words, if the subject's ability value is higher, the probability of him answering the question correctly will be higher.

[0040] However, conventional CAT technology still has some drawbacks when applied to test scales. Since users are scored after each question to determine the next question, on the one hand, calculating the user's score after each question means that users have no opportunity to modify the questions they have answered, resulting in poor flexibility; on the other hand, it requires the execution of complex real-time algorithms and relies on the support of real-time computing resources.

[0041] For example, taking the application of CAT in mental health measurement as an example, during the mental health measurement process, the user's answer may be affected by a variety of factors such as emotion, cognition and environment. If the user finds that their choice is inaccurate or does not conform to the actual situation after answering the question, but cannot go back and modify the answer, this may lead to the measurement result not truly reflecting the user's psychological state, thus affecting the accuracy and reliability of the measurement result.

[0042] It is evident that, while ensuring a certain degree of personalization and targeting, it is necessary to further improve the flexibility of scale measurement technology, thereby improving the accuracy and reliability of measurement results for the tested users and reducing the dependence on real-time computing resources.

[0043] Therefore, in the embodiments of this specification, a stage-based personalized adaptive testing method is provided. The scale is designed to be staged and divided into question groups. During the user's completion of the scale, user characteristics are considered, and a more suitable question group is provided for the user at each stage based on the user's current responses. This avoids the user answering unsuitable question groups, ensuring strong personalization and targeting, and reduces the number of questions the user needs to answer while maintaining measurement accuracy and comparability. Furthermore, the question group design calculates the user's score after each question group is completed, rather than after each individual question. This allows the user to return and modify their answers within the question group, increasing flexibility and improving the accuracy and reliability of the measurement results. It also reduces reliance on real-time computing resources to some extent.

[0044] In the scale provided in this embodiment, items can be divided into multiple question groups, and these question groups can be further divided into multiple measurement phases. Each test user needs to complete all phases, specifically, one question group corresponding to each phase.

[0045] Furthermore, the question sets used by users in each measurement phase can be determined based on their performance in the previous measurement phase. For example, if a user scored low in the previous measurement phase, they can answer a set of questions with lower difficulty in the current phase.

[0046] Figure 1 This is a schematic diagram of the structure of a scale for adaptive testing provided in an embodiment of this specification.

[0047] like Figure 1 As shown, the scale comprises three measurement phases: Phase 1, Phase 2, and Phase 3. Phase 1 contains one item group a1, Phase 2 contains two item groups b1 and b2, and Phase 3 contains three item groups c1, c2, and c3.

[0048] In practical application, users need to complete Stage 1, Stage 2, and Stage 3 sequentially. Specifically, they need to complete one set of questions in Stage 1, one set of questions in Stage 2, and one set of questions in Stage 3 in sequence. For example, a user can complete set a1, set b1, and set c1 in sequence. Another example is completing set a1, set b1, and set c2 in sequence. Yet another example is completing set a1, set b2, and set c2 in sequence. And yet another example is completing set a1, set b2, and set c3 in sequence.

[0049] In the embodiments of this specification, the number of measurement stages in the scale, the number of question groups in each measurement stage, and the number of questions in each question group can be determined as needed based on information from various dimensions such as the type of the scale, the quality and quantity of questions in the question pool, and the expected user experience.

[0050] Furthermore, the first measurement phase of the scale's multiple measurement phases may include a set of questions. This is because the measurement system typically does not know a user's actual abilities before they begin answering the questions; therefore, it is possible to configure all users to answer the same set of questions.

[0051] Furthermore, each subsequent measurement phase within the scale's multiple measurement phases may contain two or more sets of items. For example, each phase may contain 2-4 sets of items.

[0052] Furthermore, there can be differences in difficulty among multiple question groups within the same measurement phase, and the difficulty of each question group can be relative to other question groups within the same measurement phase. In practical applications, if a user scored low in the previous measurement phase, they can answer the easier question groups in this phase; conversely, if a user scored high in the previous measurement phase, they can answer the more difficult question groups in this phase.

[0053] For example, in such Figure 1 In the scale, if a user scores low in Phase 1, they can answer the easier question set b1 in Phase 2; conversely, if a user scores high in Phase 1, they can answer the more difficult question set b2 in Phase 2.

[0054] Furthermore, within the same measurement phase, there can be differences in difficulty among multiple question groups associated with the same question group in the previous measurement phase. The difficulty level of each question group can be relative to other question groups in the same phase that are associated with the previous measurement phase, or it can be relative to other question groups in the same phase besides itself. In practical application, if a user scored low in the previous measurement phase, they can answer the easier question groups among the multiple question groups associated with the question groups answered in the previous measurement phase in this phase; conversely, if a user scored high in the previous measurement phase, they can answer the more difficult question groups among the multiple question groups associated with the question groups answered in the previous measurement phase in this phase.

[0055] For example, in such Figure 1 In the scale, if a user scores low on question group b1 in stage two, then in stage three they can answer question group c1, which is relatively easier than question group c1 in question group c2, which is related to question group b1; conversely, if a user scores high on question group b1 in stage two, then in stage three they can answer question group c2, which is relatively more difficult than question group c1 in question group c2, which is related to question group b1.

[0056] For example, in such Figure 1 In the scale, if a user scores low on question group b2 in stage two, then in stage three they can answer question group c2, which is relatively easier than question group c2 in question group c3, which is related to question group b2; conversely, if a user scores high on question group b2 in stage two, then in stage three they can answer question group c3, which is relatively more difficult than question group c2 in question group c3, which is related to question group b2.

[0057] In addition, as explained above, although Figure 1 The diagram shows that Phase Two contains two question sets and Phase Three contains three question sets. However, in practical applications, the number of question sets in each measurement phase is not limited to these numbers and can be set according to various parameters, actual constraints, and business needs. Similarly, although... Figure 1 The diagram shows that the next-stage question groups associated with question group a1 include question groups b1 and b2; the next-stage question groups associated with question group b1 include question groups c1 and c2; and the next-stage question groups associated with question group b2 include question groups c2 and c3. However, in practical applications, the number of next-stage question groups associated with the previous stage question groups is not limited to two; for example, it can include three or more. For instance, the next-stage question groups associated with question group b1 can include question groups d1, d2, and d3; the next-stage question groups associated with question group b2 can include question groups d3, d4, and d5; or, the next-stage question groups associated with question group b2 can include question groups d2, d3, and d4.

[0058] Furthermore, each item group in the scale can contain one or more items. In practical applications, if an item group contains multiple items, and the scoring process occurs after one item group is completed, then while answering questions within that item group, the user can go back and modify their answers to previously answered questions, offering greater flexibility and improving the accuracy of the measurement results. It is understandable that when at least one item group in a scale contains multiple items, it at least partially improves the flexibility of the user's answering process and at least partially enhances the accuracy and reliability of the measurement results.

[0059] In the embodiments of this specification, the process of implementing the system and method for adaptive testing may include three stages.

[0060] The first stage involves pre-constructing the scale according to preset scale construction rules, i.e., creating the test paper. These scale construction rules may include rules for setting the scale's structure, rules for selecting items for question groups within the scale, and rules for routing responses to the scale. Specifically, the scale's structure rules may include the number of stages in the scale, the number of question groups in each stage, and the number of items in each question group. The rules for selecting items for question groups within the scale may include constraints and specific operational procedures for selecting items from a question pool for each question group. The response routing rules are used to determine the actual questions a user will answer when using the scale.

[0061] In the second stage, given the established scale, the system provides users with question sets for each testing phase based on the scale's response routing rules. For example, based on a user's score in the current question set, the system provides the user with question sets suitable for that user in the next phase.

[0062] In the third stage, after the user completes the questions in each stage, the user's final score is calculated, and the user's measurement rating is determined based on the final score.

[0063] In practical applications, the various stages of the process can be executed on the same or different devices. For example, the first stage can be executed on a server to build a scale and send it to the user's terminal device. The user's terminal device can then interact with the user and execute the second stage to obtain the user's responses to the scale. Subsequently, the third stage can be executed on either the user's terminal device or the server to determine the user's measurement rating based on their responses, which serves as the measurement result. Of course, the executing entity for each stage is not limited to the example given here. For instance, the first stage could be executed by the user's terminal device, and the second stage could be executed by the server; the specific execution can be determined as needed in practical applications.

[0064] This specification provides an adaptive testing method, and also relates to an adaptive testing apparatus, a computing device, a computer-readable storage medium, and a computer program product, which will be described in detail in the following embodiments.

[0065] Figure 2 This is a flowchart illustrating an adaptive testing method provided in an embodiment of this specification.

[0066] From a programming perspective, the entity executing the process can be a program hosted on an application terminal or application server. It can be understood that this method can be executed by any device, equipment, platform, or cluster of devices with computing and processing capabilities.

[0067] In practical applications, the process of the method in the embodiments of this specification can be executed entirely or partially by the user terminal device. It is understood that part or all of the process can be executed by the terminal device used by the user participating in the scale measurement. For example, the user terminal device may include, but is not limited to, mobile phones, tablets, laptops, workstations, smart home terminals, in-vehicle smart terminals, etc.

[0068] In embodiments of this specification, a scale for adaptive testing may include multiple groups of items divided into multiple sequentially associated measurement phases, each group of items may include one or more items.

[0069] like Figure 2 As shown, the process may include the following steps:

[0070] Step 202: Obtain the user's first answer result for the first question group in the first measurement stage of the multiple measurement stages.

[0071] Users can be those who answer questions on the scale, and can also be referred to as test users, measurement users, tested users, measured users, subjects, etc.

[0072] The first measurement stage can refer to any one of the multiple measurement stages included in the scale. Further, the first measurement stage can refer to any one of the multiple measurement stages included in the scale, excluding the last measurement stage.

[0073] The first question set can refer to any one of the question sets included in the first measurement phase. Optionally, if the first measurement phase is the first measurement phase in the scale, then a question set from the first measurement phase can be used as the first question set. Optionally, if the first measurement phase is a subsequent measurement phase in the scale other than the first measurement phase, then the first question set can actually be determined based on the user's answers in the previous measurement phase.

[0074] Step 204: Determine the user's first answer score based on the first answer result.

[0075] In the embodiments of this specification, the first answer result corresponds to the answer result of the first question group. Therefore, the first answer result may include sub-results corresponding to each question in the first question group. In practical applications, step 204 may specifically include first determining the sub-scores corresponding to each question based on the sub-results of each question in the first question group, and then calculating the first answer score based on the sub-scores corresponding to each question.

[0076] Furthermore, in determining the sub-scores for each question based on its sub-results, this can be done using pre-set correlation information between the answer results and the answer scores. For example, the sub-results of a question may include results such as A, B, C, and D, and the answer scores may include Arabic numerals. The forms of the sub-results and the answer scores are not limited to these.

[0077] Optionally, the mapping relationship between the answer results and the answer scores can be preset. Thus, when executing step 204, the sub-scores of each question corresponding to the sub-results of each question can be queried according to the preset mapping relationship information.

[0078] Alternatively, the functional relationship between the answer result and the answer score of the question can be preset. Thus, when executing step 204, the sub-score of each question can be calculated based on the sub-result of each question according to the preset functional relationship information.

[0079] Furthermore, in the process of calculating the first answer score based on the sub-scores corresponding to each question, the calculation can be performed according to preset score statistics rules. Specifically, the feature values ​​of the sub-scores corresponding to each question in the first question group can be used as the first answer score. The feature values ​​can include sums, mean values, or values ​​calculated by substituting the sub-scores corresponding to each question in the first question group into a preset linear or nonlinear function, and are not limited to these.

[0080] Step 206: Based on the correspondence between the answer scores and the question groups, and based on the first answer score, determine a second question group from two or more question groups in the second measurement stage; the second measurement stage is the next measurement stage after the first measurement stage.

[0081] In embodiments of this specification, the response routing rules (hereinafter also referred to as routing rules) of the scale can be predetermined during the scale construction phase. The routing rules can reflect the correspondence between the user's currently obtained response scores and candidate question groups. Specifically, the response routing rules can be executed based on the user's current response scores to determine a second question group from multiple candidate question groups to be provided to the user.

[0082] In the embodiments of this specification, the second measurement phase typically includes two or more question groups associated with the first question group. In practical applications, for different users, users may be routed to one of the two or more question groups associated with the first question group according to preset routing rules. The preset routing rules may include judgment rules based on score threshold conditions corresponding to each question group.

[0083] In practical application, the step of determining a second question group from two or more question groups in the second measurement stage based on the first answer score, according to the correspondence between the answer score and the question group, can specifically include: obtaining the score threshold condition corresponding to a specified question group in two or more question groups associated with the first question group in the second measurement stage; determining whether the first answer score meets the score threshold condition corresponding to the specified question group, and obtaining a first judgment result; if the first judgment result is yes, then the specified question group is determined as the second question group.

[0084] Step 208: Provide the second set of questions to the user.

[0085] In practical applications, providing the second question set to the user can specifically mean pushing the second question set to the user so that the user can answer the questions in the second question set.

[0086] It should be understood that in the methods described in one or more embodiments of this specification, the order of some steps may be adjusted according to actual needs, or some steps may be omitted.

[0087] Figure 2The method described above uses a scale that includes multiple measurement stages and is divided into multiple question groups. After a user completes the questions in the first question group of the previous measurement stage, the user is routed to a question group in the next measurement stage that is suitable for the user's answers based on the scores of the questions answered in the previous measurement stage. Thus, on the one hand, by determining the question group that the user needs to answer in the next measurement stage based on the user's answer scores in the previous measurement stage, the questions provided to the user are more targeted, thereby making the measurement results more accurate. On the other hand, dividing the questions into question groups provides users with the convenience and flexibility to return and modify the answers to the questions within the question group during the process of answering the scale, which also makes the measurement results more accurate.

[0088] As an incomplete example, the processes executed in real-time during user interaction in the embodiments of this specification can be executed by the user terminal device; the preset information involved in the processes in the embodiments of this specification can be predetermined by the user terminal device or the server. In practical applications, the specific execution device for each step in the process can be set according to the actual situation, and this application does not limit it in this regard.

[0089] based on Figure 2 In addition to the method described herein, this specification also provides some improved or more specific implementations of the method, which will be described below.

[0090] When conventional CAT (Computer-Aided Testing) technology is applied to test scales, the selection of items is mainly based on the difficulty parameter. However, the proposer of the technical solution in this application found that in practical applications, the evaluation perspective of the items also affects the comprehensiveness and accuracy of the measurement results. Conventional CAT technology has difficulty balancing different evaluation perspectives, which may lead to some evaluation perspectives not being fully evaluated, thus affecting the comprehensiveness and accuracy of the measurement results.

[0091] Let's take a mental health scale as an example. In a mental health scale, the difficulty of the items mainly refers to the level of mental health reflected by the items. However, mental health scales actually need to assess from multiple perspectives. For example, the assessment of depressive states needs to include aspects such as emotional state, somatization symptoms, and social functioning. However, when conventional CAT (Computational Analytic Hierarchy Process) technology is applied to measure mental health scales, it usually cannot provide a comprehensive and balanced assessment of all aspects, resulting in high randomness in item selection, low accuracy of measurement results, and poor comprehensiveness.

[0092] In one or more embodiments of this specification, during the scale construction phase before the user testing process is executed, the questions in each question group can be determined according to preset question selection rules.

[0093] On the one hand, the preset question selection rules may include rules regarding difficulty constraints, and more specifically, may include difficulty constraints for question sets.

[0094] For the item group in the first measurement phase of the scale, the items in this item group can be set to have a sufficiently wide difficulty distribution. Specifically, firstly, the items in the item pool can be divided into multiple sub-item pools of different difficulty levels according to the item difficulty value; then, a predetermined number of items can be selected from each sub-item pool to determine the items in the item group of the first measurement phase. For example, 1-3 items can be extracted from each sub-item pool.

[0095] Furthermore, if the number of questions selected from different sub-question pools is consistent, for example, the same or not significantly different, it can ensure that the questions in the determined question set for the first measurement phase are not only widely distributed in terms of difficulty, but also evenly distributed in terms of difficulty.

[0096] For subsequent measurement phases within the scale, within the same phase, the overall difficulty of different question groups can be differentiated, while the difficulty of questions within a question group can be relatively concentrated. Specifically, in the embodiments of this specification, any target question group included in the scale can satisfy the following first condition: the information content function of the target question group reaches its peak at the target difficulty of the target question group; the information content function is used to represent the correlation between the question group information content of the target question group, the difficulty value of each question in the target question group, and the user's ability value. The user's ability value can be determined based on the user's answer scores for questions already answered in the scale.

[0097] In the embodiments of this specification, a method for determining the difficulty parameter of a question is provided.

[0098] Optionally, for the target question, the difficulty parameter of the target question can be determined based on the average response of the norm population to the target question. If the average response is low, it indicates that the target question is difficult; conversely, it indicates that the target question is difficult.

[0099] Alternatively, for the target question, a multi-level scoring IRT model can be used, such as the generalized partial credit model (GPCM) or the graded response model (GRM). Question parameters, including question difficulty parameters, can be obtained through methods such as weighted likelihood estimation (WLE), joint maximum likelihood estimation, marginal maximum likelihood estimation, Bayesian estimation, and expectation maximization (EM).

[0100] Taking GPCM as an example, the model formula is as follows:

[0101]

[0102] Among them, P jk (θ): The probability that the respondent receives a score k on question j;

[0103] θ: Latent level, the latent level of a certain attribute of the respondent, such as the level of depression, is the parameter to be estimated;

[0104] a j The discriminant of question j is the parameter to be estimated.

[0105] b jc : Boundary parameters of item j on score c, that is, the potential level when the probability of obtaining scores c and c-1 on item j is equal, representing the difficulty of transitioning the score from j-1 to j, and is a parameter to be estimated;

[0106] D: is a constant scaling factor, usually taken as 1.7;

[0107] m: The maximum possible score for question j.

[0108] In practical applications, the model can be fitted based on a collected dataset (e.g., the answers of N respondents on J questions) to obtain the aforementioned parameters to be estimated, θ and a. j and b jc The value of . Where, parameter θ reflects the user's ability, parameter a j The parameter b reflects the question's discriminatory power. jc This reflects the difficulty of the question.

[0109] Specifically, the general process for model parameter estimation is as follows:

[0110] 1) Collect a dataset containing the answers from N respondents on J questions, with each question having M rating levels (e.g., 4-point rating, 5-point rating, etc.).

[0111] 2) Select a specific parameter estimation method and initialize the parameters. Input the dataset and initial parameters into the estimation software, set the number of iterations and the model fitting convergence criterion, and then perform parameter estimation.

[0112] 3) Examine the fit by checking parameters such as ICC (Item Characteristic Curve), IIF (Item Information Function), AIC (Akaike Information Criterion), and BIC (Bayesian Information Criterion). ICC describes the performance of items at different capability levels. IIF quantifies the amount of information provided by items at different capability levels. AIC and BIC are two commonly used model selection and evaluation metrics, standards for assessing the relative quality of statistical models. They are primarily used for model selection, ensuring that the chosen model fits the data well without being overly complex and overfitting. In practical applications, relevant packages in R and Python can be used for calculation; specific details can be found in related technical documentation and will not be elaborated upon here.

[0113] If the results are unsatisfactory, unsuitable questions need to be deleted and the estimation repeated. This process is repeated until the model fits well, identifying the questions that conform to the model, and outputting the parameters for each question: a j b jc In practical applications, the overall difficulty of problem j can be multiples of problem j's difficulty level b. jc The mean.

[0114] On the other hand, the preset question selection rules may include rules regarding evaluation angle constraints, and more specifically, may include evaluation angle constraints for question groups.

[0115] Specifically, in the embodiments of this specification, any target question set included in the scale can satisfy the following second condition: the number of questions in the target question set corresponding to each assessment angle of the scale satisfies a preset quantity threshold condition; the quantity threshold condition includes being greater than or equal to a first quantity threshold and less than a second quantity threshold.

[0116] The first quantity threshold and the second data threshold can be predetermined. Specifically, the total number of questions for each assessment angle can be determined first based on the total number of questions in the scale and the predetermined proportion of questions for each assessment angle. Then, the sub-question range for each assessment angle in each question group can be determined based on the total number of questions for each assessment angle and the predetermined scale structure. The sub-question range can specifically include the range between the first quantity threshold and the second quantity threshold. In practical applications, the setting of the sub-question range for each assessment angle in each question group should ensure that the number of questions in each question group and the number of questions in each answer path meet the design requirements of the scale.

[0117] In practical applications, considering that more questions generally lead to more accurate measurements, but shorter scales (e.g., less than 20 questions) typically reduce the burden on respondents, the total number of questions can be determined based on classic scales and client needs. As an example, given the total number of questions and the approximate proportion of questions for different assessment angles, the range of questions for each content assessment angle can be defined as "total number of questions * proportion of questions ± a," where 'a' can be 1, 2, or 3. Another example is the range of questions for each content assessment angle defined as "total number of questions * proportion of questions - a1 to total number of questions * proportion of questions + a2," where a1 and a2 can independently be 1, 2, or 3. Then, the number of questions for each content assessment angle can be allocated to different question groups in different stages, determining the number of questions for each stage while ensuring that the number of questions for each content assessment angle in each response path of the scale meets the design requirements. In practical applications, the number of questions in each question group can be set to an appropriate value within the aforementioned range to ensure that the number of questions corresponding to each content evaluation angle in each answer path meets the design requirements.

[0118] On the other hand, the preset question selection rules may include another rule regarding difficulty constraints. More specifically, this may include difficulty constraints on the sequence of question sets, or information content constraints on the sequence of question sets. The difficulty of a question is closely related to its information content; in other words, constraining the difficulty of a question can be a way of constraining its information content. In practical applications, when constructing a scale, the feature information content corresponding to each answer path can be limited to meet a certain information content threshold. This ensures the accuracy of the measurement results obtained after users answer based on each answer path.

[0119] Specifically, the scale has multiple response paths, each of which corresponds to a sequence of question groups selected from each of the multiple measurement stages. The target question group sequence corresponding to any target response path in the multiple response paths satisfies the following third condition: the feature information content of the target question group sequence is greater than a preset information content threshold. The feature information content is determined based on the information content of all questions in the target question group sequence.

[0120] The information threshold can be a feature used to characterize the information content of the target question set sequence. It can correspond to the entire answering path, thereby avoiding a situation where the information content of each question set meets the preset conditions, but the information content of the answering path composed of multiple question sets is insufficient. By ensuring that the information content of each answering path meets the threshold, the accuracy of the measurement results can be further guaranteed.

[0121] Optionally, the feature information content of the target question group sequence is greater than or equal to a preset information content threshold. Specifically, this may include the sum of the information content of all questions included in the target question group sequence being greater than or equal to a preset first information content threshold. Alternatively, the feature information content of the target question group sequence is greater than or equal to a preset information content threshold. Specifically, this may include the average information content of all questions included in the target question group sequence being greater than or equal to a preset second information content threshold, etc.

[0122] As an example, when using GPCM to obtain question parameters, the information content of the question is calculated as follows:

[0123]

[0124] Among them, I j (θ): The amount of information in question j;

[0125] P jk (θ): The probability that the respondent receives a score k on question j;

[0126] a j The discriminant power of question j;

[0127] D: is a constant scaling factor, usually taken as 1.7;

[0128] m: The maximum possible score for question j;

[0129] k: is the rating value k;

[0130] The expected score for question j is...

[0131] Furthermore, calculate the information content of a set of questions, such as a question group or the information content of all questions in a path, which is the sum of the information content functions of all questions:

[0132]

[0133] Where n: the total number of questions in the question set.

[0134] In practical applications, the information content threshold can be determined based on industry-recognized experience. Specifically, it can be set by comprehensively considering the requirements for the standard error of the measurement results and the requirements for test reliability.

[0135] Specifically, the information content of a psychological scale is called Test Information Content (TIF), which can be calculated using the formula I(θ) above. On the one hand, Test Information Content is inversely proportional to the Standard Error of Measurement (SEM); the higher the Test Information Content, the smaller the score error obtained from the scale. An exemplary calculation formula is as follows: On the other hand, the amount of information in a test is related to its reliability. Generally, the higher the amount of information measured, the higher the reliability of the test.

[0136] For example, in a mental health scale, the information content threshold can be set to 3.3. When the TIF is greater than 3.3, the standard error is less than 0.55, which is a generally accepted level in the field of psychometrics, and the test reliability r-value is greater than 0.7, which is an acceptable level of test reliability generally accepted in the field of psychometrics. In practical applications, the TIF value can be appropriately adjusted within a certain range, provided that both the standard error and test reliability meet the preset conditions.

[0137] In one or more embodiments of this specification, optionally, the results of manual test paper assembly by the designer can be obtained, and after the test paper assembly is completed, it can be verified whether each question group and each answer path meets the aforementioned constraints.

[0138] The first verification condition is to verify whether the peak value of the test information function (TIF) curve for each question group is near the target difficulty level of the question group. By setting difficulty constraints for each question group, firstly, the question group is suitable for assessing this level of difficulty, making it suitable for users with abilities around this level to measure, thus reducing the likelihood of user resistance; secondly, it maximizes the accumulation of information, enabling an accurate assessment of user abilities.

[0139] The second verification condition, for each question group, can verify whether the number of questions under each evaluation angle in the question group meets the preset quantity threshold condition.

[0140] The third verification condition involves verifying, for each answer path, whether the feature information content of the question group sequence corresponding to the path is greater than a preset information content threshold. This feature information content can be determined based on the information content of all questions in the target question group sequence. For example, it can be verified that the information content value of all questions on the path meets a standard (e.g., TIF > 3.3). Reaching this value indicates that the evaluation error is small and has reached an industry-recognized level, thereby ensuring the accuracy of the evaluation.

[0141] In practical applications, after the document assembly is completed, at least some of the aforementioned first, second, and third verification conditions can be executed sequentially or in parallel. If at least one verification fails, a document assembly adjustment prompt message will be issued to the designer. Furthermore, after obtaining the updated document assembly results from the designer, the verification process can be executed again until all the conditions to be verified are met.

[0142] In one or more embodiments of this specification, optionally, test papers can be automatically generated according to the aforementioned question selection rules. In practical applications, the automatic test paper generation method may include heuristic test paper generation methods, linear programming methods, etc. For example, the Normalized Weighted Absolute Deviation Heuristic (NWADH) algorithm can be used.

[0143] Specifically, the automatic test paper generation process may include selecting a preset number of candidate questions from a pool of candidate questions to form a question group to be built. The process of selecting each candidate question may include: obtaining the target information content of the question group to be built; then, based on the target information content, calculating the absolute deviation value of the candidate questions in the pool, where the absolute deviation value represents the absolute value of the difference between the current information content and the target information content, wherein the current information content represents the question group information content if the candidate question is included in the question group to be built; then, calculating the cumulative weight value corresponding to the candidate question based on a preset quantity threshold condition that the number of questions for each evaluation angle of the scale should meet, the evaluation angles of the selected questions in the question group to be built, and the evaluation angle of the candidate question; and finally, based on the absolute deviation value and the cumulative weight value, determining whether to select the candidate question into the question group to be built.

[0144] Specifically, obtaining the target information content of the question group to be built may include: selecting a preset number of questions from the candidate question pool to form a pre-selected question set, wherein the preset number is determined by multiplying the preset number of questions in the question group to be built by a preset question quantity coefficient, wherein the question quantity coefficient is a value greater than or equal to 1; then, calculating the total information content of the pre-selected question set based on the information content of each question in the pre-selected question set; and then determining the target information content of the question group to be built by dividing the total information content by the question quantity coefficient of the pre-selected question set.

[0145] Furthermore, the step of selecting a preset number of questions from the candidate question pool to form a pre-selected question set may specifically include: dividing the candidate question pool into multiple candidate sub-question pools matching different difficulty levels according to the difficulty value of each question in the candidate question pool; and then selecting a preset number of questions from the candidate sub-question pools corresponding to the difficulty level of the question group to be built to form a pre-selected question set.

[0146] In practical applications, the questions in the candidate question pool can be divided into three candidate sub-pools according to difficulty, and the information content of the questions in the candidate sub-pools at the target difficulty level of the question group can be calculated. Then, the M*Z questions with the largest information content from the candidate sub-pools (where M is the preset number of questions in the question group to be built, Z is the preset question quantity coefficient, and Z is a value greater than or equal to 1) can be selected as the pre-selected question set. Then, the quotient of the sum of information content of the questions in the pre-selected question set divided by the question quantity coefficient can be calculated as the target information content of the question group to be built.

[0147] For example, the following formula can be used to calculate the target information content T(θ) of the question group to be built, based on the information content of the M*N candidate questions in the candidate question pool that have relatively large information content in terms of target difficulty:

[0148]

[0149] Where T(θ) represents the target information content of the question set;

[0150] I i (θ): The amount of information about candidate question i at the target difficulty θ.

[0151] Furthermore, the step of calculating the absolute deviation value of the candidate questions in the candidate question pool based on the target information amount may specifically include: determining the expected information amount of the candidate questions based on the target information amount, the total information amount of the selected questions in the question group to be built, and the number of candidate questions in the question group to be built; and calculating the absolute deviation value of the information amount of the candidate questions based on the information amount of the candidate questions and the expected information amount.

[0152] In practical applications, when selecting the k-th question for a question pool, we can first calculate the absolute deviation of each candidate question in the pool. Taking candidate question t as an example, we can calculate the absolute deviation d. t .

[0153] Specifically, the absolute deviation d can be calculated using the following formula. t :

[0154]

[0155] d t : Represents the absolute deviation of the number of available questions t in the question pool;

[0156] R k-1 : Indicates the question pool after k-1 questions have been selected;

[0157] M: Indicates the total number of questions in the question set;

[0158] M-(k-1): Represents the number of questions to be selected in the question set;

[0159] T: Indicates the target information content of the question set;

[0160] u t : Indicates the amount of information contained in the candidate question t in the question pool;

[0161] I: Indicates the number of questions in the question pool;

[0162] x i : This is the determining variable, used to indicate whether question i is selected; 1 for selection and 0 for non-selection.

[0163] u i : Indicates the amount of information in question i (the amount of information in question i at the target difficulty level);

[0164] This indicates the TIF of the currently selected question (i.e., the TIF of the question that has been selected from the question pool).

[0165] The above calculations show that, in practical applications, the smaller the absolute deviation of the information content of the alternative topics, the better they are from the perspective of information content.

[0166] Furthermore, after calculating the absolute deviation of the candidate items, the evaluation perspective of the candidate items also needs to be taken into consideration. In NWADH, the parameter of the evaluation perspective is described as cumulative weight.

[0167] Specifically, the step of calculating the cumulative weight value corresponding to the candidate item based on the preset quantity threshold condition that the number of items for each assessment angle of the scale should meet, the assessment angles of the selected items in the question group to be built, and the assessment angles of the candidate items may include: first, obtaining the preset quantity threshold condition that the number of items for each assessment angle of the scale should meet; then, determining the weight value corresponding to each assessment angle of the scale based on the assessment angles of the selected items in the question group to be built and the quantity threshold condition; and finally, determining the cumulative weight value of the candidate item based on the assessment angles of the candidate items and the weight values ​​corresponding to each assessment angle of the scale.

[0168] The process of determining the weight values ​​corresponding to each assessment angle of the scale based on the assessment angles of the selected items in the question group to be built and the quantity threshold condition may include: First, determining the number of selected items in the question group to be built corresponding to each assessment angle of the scale; then, determining the weight value of the assessment angle in the question group to be built whose number of selected items is less than a first quantity threshold as a first value, typically a value greater than 1, for example, 2; determining the weight value of the assessment angle in the question group to be built whose number of selected items is greater than or equal to the first quantity threshold and less than a second quantity threshold as a second value less than the first value, typically a value of 1; and determining the weight value of the assessment angle in the question group to be built whose number of selected items is greater than or equal to the second quantity threshold as a third value less than the second value, typically a value of 0.

[0169] In practical applications, assume there are D evaluation angles in the scale. When selecting the k-th item for a set of items, we can first count the number of items for each evaluation angle of the selected items, and then convert the results into a weight W for each evaluation angle. d As an example, if the number of questions with evaluation angle d in the question group (k-1) to be built does not reach the minimum number of questions, then W can be set. d =2; If the number of questions for evaluation angle d is between the minimum and maximum number of questions, then W can be set. d =1; If the number of questions at evaluation angle d is greater than or equal to the maximum number of questions, then W can be set. d =0.

[0170] In one or more embodiments of this specification, to further improve the effectiveness of the scheme for balancing evaluation angles within a question group based on the cumulative weight of evaluation angles, in practice, the weight of questions that do not include evaluation angles that have reached their maximum number of questions can be further increased. Supplementary weight values ​​are provided here. A technical solution to improve reward value.

[0171] Specifically, determining the cumulative weight value of the candidate items based on the evaluation angles of the candidate items and the weight values ​​corresponding to each evaluation angle of the scale may include: firstly, calculating the supplementary weight values ​​corresponding to each evaluation angle of the selected items in the question group to be built based on the weight values ​​corresponding to each evaluation angle of the selected items in the question group to be built; then, determining the cumulative weight value of the candidate items based on the evaluation angles of the candidate items and the weight values ​​and supplementary weight values ​​corresponding to each evaluation angle of the selected items in the question group to be built.

[0172] To make it easier to understand, the following example is provided.

[0173] In practical applications, for the candidate question t in the question pool, the steps to calculate its cumulative weight may include: first calculating the weight W of the candidate question t. d Approximate supplementary weights Then, in approximate supplementary weights Calculate the cumulative weight y based on t .

[0174] Specifically, the approximate supplementary weights can be calculated sequentially using the following formula. and cumulative weight y t :

[0175]

[0176] Where D represents the number of evaluation angles;

[0177] d: Values ​​range from 1 to D;

[0178] W d : Indicates the weight of the d-th evaluation angle;

[0179] This represents the average weight of each evaluation perspective;

[0180] W max : Represents the maximum weight W among all D evaluation angles. d In practical applications, the value can be selected based on the actual situation. For example, if among all D evaluation angles there is an angle that has not yet reached the minimum number of questions, then the maximum value for W is... d It can be 2, W max The value can be 2; if all D evaluation angles have reached the minimum number of questions but there are angles that have not yet reached the maximum number of questions, then the maximum value of W is 2. d It can be 1, W max The value can be 1;

[0181] x td : Indicates whether question t contains evaluation angle d, 1 if it contains d, 0 if it does not;

[0182] W d The supplementary weights, more specifically, represent the weights of W under the current question set to be built. d Supplementary weights.

[0183] Furthermore, determining whether to select the candidate question into the question group to be built based on the absolute deviation value and the cumulative weight value may specifically include: calculating the standardized absolute deviation value corresponding to the absolute deviation value; calculating the standardized cumulative weight value corresponding to the cumulative weight value; and then selecting the candidate question with the largest sum of standardized absolute deviation value and standardized cumulative weight value from the candidate question pool into the question group to be built.

[0184] Specifically, on the one hand, the standardized absolute deviation can be calculated based on the absolute deviation. Taking the candidate question t as an example, the standardized absolute deviation can be calculated based on the absolute deviation d using the following formula. t Calculate the standardized absolute deviation e t :

[0185]

[0186] Among them, R k-1 : Indicates the question pool after k-1 questions have been selected;

[0187] d t : Represents the absolute deviation of the number of available questions t in the question pool;

[0188] e t : Represents the standardized absolute deviation of the candidate questions t in the question pool.

[0189] On the other hand, standardized cumulative weights can be calculated based on cumulative weights. Taking the candidate question t as an example, the standardized cumulative weights can be calculated based on the cumulative weights y using the following formula. t Calculate the standardized cumulative weights

[0190]

[0191] Among them, R k-1 : Indicates the question pool after k-1 questions have been selected;

[0192] y t : Represents the cumulative weight of the candidate question t in the question pool;

[0193] This represents the standardized cumulative weight of the candidate question t in the question pool.

[0194] Furthermore, it can be calculated t∈R k-1 and will The largest question was selected for inclusion in the question group to be built.

[0195] In practical applications, when selecting candidate questions from a question pool, the standardized absolute deviation e can be calculated for each candidate question in the question pool. t and standardized cumulative weight values And in each of the alternative questions The largest candidate question was selected for inclusion in the question group to be built.

[0196] Specifically, when the absolute deviation d of the alternative question t t The smaller (standardized absolute deviation e) t ), cumulative weight y t The larger (standardized cumulative weight) The larger the value, the better.

[0197] In one or more embodiments of this specification, when the measurement phase of a scale for adaptive testing includes multiple question groups, the difficulty levels of these question groups are typically different. In practical applications, the difficulty level of each question group can be determined based on the difficulty values ​​of the questions within that question group. Specifically, during the scale construction (i.e., test paper assembly) phase, the difficulty level of each question group can be determined after the scale structure and the corresponding questions for each question group within the scale have been determined. For example, the difficulty level can be expressed as low, medium, or high, or it can be represented by letters or other identifiers.

[0198] Furthermore, in the second measurement phase, a division score can be set between every two question groups with adjacent difficulty levels in the two or more question groups associated with the first question group. That is, the routing rule can be a routing rule based on the division score. In practical applications, the user can be routed to the appropriate question group in the current phase based on the division score between every two question groups with adjacent difficulty levels in the current phase.

[0199] Suppose there are N question groups associated with the first question group. These N question groups can be sorted according to their difficulty level (which may include ascending or descending order) to obtain a question group difficulty sequence. Then, a division score can be set between every two question groups with adjacent difficulty levels. That is, a division score can be set between every two adjacent question groups in the question group difficulty sequence, thus obtaining N-1 division scores.

[0200] In a specific embodiment, determining a second question group from two or more question groups in the second measurement stage based on the first answer score, according to the correspondence between the answer score and the question group, may specifically include: obtaining the division scores corresponding to one or more question group pairs involved in the two or more question groups associated with the first question group in the second measurement stage, wherein the question group pair consists of two question groups with adjacent difficulty levels in the two or more question groups in the second measurement stage; then, for a target question group pair in the one or more question group pairs, if the first answer score meets a first score threshold condition, then the question group with a relatively lower difficulty level in the target question group pair is determined as the second question group, wherein the first score threshold condition includes at least that the first answer score is less than the target division score of the target question group pair; if the first answer score meets a second score threshold condition, then the question group with a relatively higher difficulty level in the target question group pair is determined as the second question group, wherein the second score threshold condition includes at least that the first answer score is greater than or equal to the target division score.

[0201] The number of question pairs is one less than the number of question groups associated with the first question group. Continuing with the example above, if there are N question groups associated with the first question group, then there are N-1 question pairs, each corresponding to a score, resulting in N-1 score distributions.

[0202] As described above, when there are one or more question pairs, it is necessary to determine which specific question pair within each pair to route a user to, based on the corresponding score for each question pair. In practical applications, the question pairs associated with the first question pair can include two or more, which will be explained below for different scenarios.

[0203] Optionally, the second measurement stage may include two question groups associated with the first question group, wherein the two question groups are question groups with adjacent difficulty levels, forming a question group pair. A pre-set division score corresponding to the question group pair can be provided, or in other words, a division score can be set between the two question groups in the question group pair. As an example, assuming the two question groups include question group b1 and question group b2, where question group b1 has a relatively low difficulty level and question group b2 has a relatively high difficulty level, a division score S can be set between question group b1 and question group b2. b1-b2 .

[0204] Therefore, if the first answer score is less than the division score, the user can be routed to the question group with a relatively lower difficulty level; if the first answer score is greater than or equal to the division score, the user can be routed to the question group with a relatively higher difficulty level. Continuing with the example above, let's assume the first answer score is S. aIf S a <S b1-b2 Then the user can be routed to question group b1; if S a ≥S b1-b2 Then the user can be routed to question group b2.

[0205] Optionally, the second measurement stage may include three question groups associated with the first question group. These three question groups are question groups with progressively adjacent difficulty levels, including two question group pairs. A pre-set division score can be provided for each question group pair. Specifically, a first division score can be provided for the first question group pair with a relatively low difficulty level, and a second division score can be provided for the second question group pair with a relatively high difficulty level, with the first division score being less than the second division score. As an example, suppose the three question groups include question groups c1, c2, and c3 with difficulty levels ranging from low to high. Question groups c1 and c2, with adjacent difficulty levels, form the first question group pair with a relatively low difficulty level. A division score S can be set between question groups c1 and c2. c1-c2 Question groups c2 and c3, which are close in difficulty level, form a second question group with a relatively higher difficulty level. A division score S can be set between question groups c2 and c3. c2-c3 And S c1-c2 <S c2-c3 .

[0206] Therefore, if the first answer score is less than the first threshold score, the user can be routed to the question group with the relatively lower difficulty level among the three question groups; if the first answer score is greater than or equal to the first threshold score and less than the second threshold score, the user can be routed to the question group with the medium difficulty level among the three question groups; if the first answer score is greater than or equal to the second threshold score, the user can be routed to the question group with the relatively higher difficulty level among the three question groups. Continuing with the example above, let's assume the first answer score is S. a If S a <S c1-c2 Then the user can be routed to question group c1; if S c1-c2 ≤S a <S c2-c3 Then the user can be routed to question group c2; if S a ≥S c2-c3 Then the user can be routed to question group c3.

[0207] Optionally, the second measurement stage may include more question groups associated with the first question group. Similar to the case described above, which includes two or three question groups associated with the first question group, the method provided in the embodiments of this specification can be referred to, and the user can be routed to a suitable second question group based on the score corresponding to each question group in the multiple question groups associated with the first question group and the user's first answer score in the first question group.

[0208] Furthermore, before obtaining the respective segmentation scores of one or more question pairs involved in the two or more question pairs associated with the first question pair in the second measurement phase, the method may further include: determining the user ability threshold corresponding to each question pair in the second measurement phase based on the preset population proportion corresponding to each question pair in the second measurement phase, wherein the user ability threshold is used to reflect the ability differences of users assigned to different question groups in the corresponding question pair; then, determining the segmentation score corresponding to the question pair based on the user ability threshold corresponding to the question pair. In practical applications, the segmentation score can be used as the basis for routing users to different question groups in the question pair.

[0209] The preset population proportion can be determined during the scale construction phase, or more specifically, during the design of the scale's structure. For example... Figure 1 As shown, for example, if Phase Two contains two question sets, the preset population ratio for both question sets can be set to 50%. Similarly, if Phase Three contains three question sets, the preset population ratio for all three question sets can be set to 33%.

[0210] In practical applications, when determining the user ability threshold corresponding to each question group in the second measurement stage based on the preset population ratio corresponding to each question group in the second measurement stage, considering that the ability distribution value is basically normal, the preset population ratio can be converted into the corresponding ability value θ through a normal cumulative function.

[0211] Specifically, the cutoff value for the pre-defined proportion of the population corresponding to each item group can be set as p. Then, the p value can be substituted into the Inverse Cumulative Distribution Function (ICDF) to calculate the ability value θ. The ICDF is a standard function that can be obtained using Python's stats package, MATLAB, or by looking up tables. In practical applications, the proportion of the population among multiple item groups in each measurement phase can be evenly distributed or unevenly distributed, depending on the scale's assessment objectives. For example, if the measurement phase includes two item groups, the pre-defined proportion of the population for each item group can be set to 50%, then the population cutoff value p = 0.5; if the measurement phase includes three item groups, the cutoff values ​​between each pair of item groups can be set as p1 = 0.33 and p2 = 0.67, respectively. For example, if the scale wants to measure the high-scoring group in more detail, then for the measurement stage which includes two question groups, it can be grouped as low score -60% and high score -40%; for the measurement stage which includes three question groups, it can be grouped as low score -50%, medium score -25%, and high score -25%.

[0212] by Figure 1 For example, in the first to second stage, the population is divided into two groups (the preset population ratio can be set to 50% each, p = 50%). The ability value corresponding to this division operation can be the average ability value θ = 0 (conforming to a standard normal distribution). In the second to third stage, the population is divided into three groups (e.g., 33% each). Users from question group b1 in the second stage can be divided into question group c1 and question group c2, dividing the population into the top 33% and the middle 33%. The ability value corresponding to this division operation can be θ = -0.41 (conforming to a standard normal distribution). Similarly, users from question group b2 can be divided into question group c2 and question group c3, dividing the population into the middle 33% and the bottom 33%. The ability value corresponding to this division operation can be θ = 0.41 (conforming to a standard normal distribution).

[0213] Furthermore, when determining the corresponding score for the question set based on the user's ability threshold (θ) for the question set, specifically, the ability value can be converted into the score for the routing rule using a true score formula. In other words, the expected score of the total score on the question set (a total of J questions) is calculated when the ability value is θ.

[0214] Specifically, the ability value θ can be converted into a division score τ(θ) using the following formula:

[0215] τ(θ)=∑ j ∑ k kP jk (θ);

[0216] Where j: is the question j in the question set;

[0217] k: The score k obtained for question j;

[0218] P jk (θ): represents the probability that question j will get score k when the ability is θ;

[0219] τ(θ): represents the score in the question set when the ability is θ; specifically, the expected value of the total score in the question set (J questions in total).

[0220] In one or more embodiments of this specification, in Figure 2 Based on this method, users can sequentially complete the responses to a set of questions corresponding to each stage of the scale, thereby obtaining the user's actual response path and the user's scores for the multiple actual response sets included in the actual response path.

[0221] Specifically, in one or more executions such as Figure 2 Following the illustrated process, the adaptive testing method further includes: calculating the user's overall score based on the user's answer scores for multiple actual answer question groups included in the actual answer path; the multiple actual answer question groups include one question group from each of the multiple measurement stages of the scale.

[0222] Combine Figure 2 In the described embodiment, when the scale includes a first measurement stage and a second measurement stage, and the user answers the first set of questions in the first measurement stage and the second set of questions in the second measurement stage, the actual answering path may include the first set of questions and the second set of questions.

[0223] In the embodiments of this specification, the actual response path consists of multiple sequentially associated actual response question groups. Further, the actual response path is a sequence of actual response question groups composed of multiple sequentially associated actual response question groups. The actual response question groups included in the sequence can sequentially correspond one-to-one with multiple sequentially associated measurement stages of the scale.

[0224] In practical applications, each actual answer question group in the actual answer question group sequence can be selected from the question groups included in the current measurement stage according to preset routing rules during the user's answering process. For example, for the first measurement stage among the multiple measurement stages, since it only contains one question group, this included question group can be determined as the question group provided to the user. Furthermore, after the user answers, the provided question group becomes a question group in the user's actual answer question group sequence. Similarly, for any subsequent measurement stage among the multiple measurement stages, since it can contain multiple question groups, the question groups among the multiple question groups included in that stage that conform to preset routing rules can be determined as the question groups provided to the user. Furthermore, after the user answers, the provided question group becomes a question group in the user's actual answer question group sequence.

[0225] As an example, assuming the scale contains M stages, the response path can contain M sets of questions, and the questions in the M sets of questions correspond one-to-one with the M stages.

[0226] In one or more embodiments of this specification, the adaptive testing method further includes, after calculating the user's overall score based on the user's answer scores for multiple actual question groups included in the actual answer path, determining the user's measurement rating based on the overall score.

[0227] Specifically, the preset mapping relationship between measurement ratings and overall scores can be pre-defined. In practical applications, the preset mapping relationship between measurement ratings and overall scores can be related to the type of the scale and rating rules, and can be set according to actual business needs.

[0228] In practical applications, after a user completes all the questions in each stage, the questions answered by the user in each stage can be summarized, and the user's final score can be determined based on the answers to all questions. For example, when applying the scheme of the embodiments in this specification to a mental health scale, an IRT model suitable for multi-level rating (e.g., GPCM) can be used to estimate the user's final score of mental health status. Then, the final score can be converted into a level (e.g., healthy, normal fluctuation, slightly uncomfortable, at risk, etc.) to describe the user's mental state.

[0229] The various technical features in the above embodiments can be combined arbitrarily, as long as there is no conflict or contradiction between the combinations of features. However, due to space limitations, they have not been described one by one. Therefore, the arbitrary combination of various technical features in the above embodiments is also within the scope of this specification.

[0230] The following is in conjunction with the appendix Figure 3Taking the application of the adaptive testing method provided in this specification in the field of mental health measurement as an example, the adaptive testing method will be described in general.

[0231] in, Figure 3 The present specification illustrates a flowchart of the processing procedure for an adaptive testing method applied to psychological testing, according to an embodiment of this specification, which specifically includes the following steps.

[0232] Step 302: Determine the structure of the phased assessment scale.

[0233] Specifically, the number of testing phases of the scale, the number of item groups in each phase, and the number of items in each item group can be determined.

[0234] Optionally, the number of test phases can be determined based on the characteristics of the scale. For example, a psychological test scale may contain 2-3 test phases.

[0235] Optionally, the first test phase of the scale may typically contain one set of items, and each subsequent test phase of the scale may typically contain 2-4 sets of items.

[0236] Optionally, each item group in the scale may contain one or more items.

[0237] Step 304: Determine the constraints on the selection of questions for the question set. More specifically, this can include constraints on difficulty and evaluation perspective, etc.

[0238] Regarding difficulty constraints: For the question sets in the first testing phase, the settings can be configured to include a sufficiently wide range of question difficulty levels. Specifically, this can be achieved by stratifying the question pool according to difficulty. For the question sets in subsequent measurement phases, the settings can be configured to ensure that the overall difficulty varies between different question sets within the same testing phase, while the difficulty of questions within each question set is relatively concentrated. Specifically, the target difficulty of the question sets can be determined first, and the information content of the question sets at the target difficulty level can be maximized during test paper compilation.

[0239] Regarding the constraints on assessment angles: the minimum and maximum number of assessment angles in each question group can be determined, and the assessment angles of the questions in each question group after the test paper is compiled can meet the constraints of these minimum and maximum numbers.

[0240] Step 306: Compile the test paper according to the question selection constraints.

[0241] Alternatively, the papers can be assembled manually, or they can be assembled automatically by executing a preset program.

[0242] Furthermore, after the test paper is completed, the test paper results can be verified by passing at least some of the following conditions: the target TIF of the question group (corresponding to the difficulty constraint), the evaluation angle of the question group (corresponding to the assessment angle constraint), and the answer path TIF (corresponding to another difficulty constraint, also known as the information content constraint). If the verification is successful, the test paper is determined to be successfully generated.

[0243] Step 308: Determine the answer routing rules, and associate the question groups of different test stages according to the answer routing rules.

[0244] Specifically, the answer routing rules can include multiple segmentation scores. These segmentation scores can be determined based on the proportion of participants advancing to different question groups in the next stage. More specifically, firstly, the proportion of participants can be converted into ability values ​​using a normal distribution cumulative function; then, the ability values ​​can be converted into segmentation scores.

[0245] Step 310: During user testing, advance the personalized testing process according to the response routing rules.

[0246] Specifically, once a user completes a set of questions, the system can push the next set of questions to them based on their score in that set.

[0247] In practical applications, when a user's score is lower than the cutoff score, the user can be given a set of questions for the next stage that are easier; when the score is higher than the cutoff score, the user can be given a set of questions for the next stage that are more difficult.

[0248] Step 312: After the user completes all test phases, the user's final score can be evaluated based on the answers to all questions.

[0249] Specifically, for mental health scales, IRT models (such as GPCM) applicable to multi-level rating can be used to estimate the final score of a user's mental health status.

[0250] Based on one or more embodiments of this specification, at least some or all of the following beneficial effects can be achieved:

[0251] First, the embodiments of this application divide the test into multiple stages. After evaluating the user in each stage, a more suitable set of questions is selected for the next stage based on the evaluation results of the current stage. This design ensures that the questions answered by the user are more relevant to the user, reducing user resistance to answering questions and improving the user experience.

[0252] Secondly, the question set design is based on the target information content, which can ensure that the question set answered by the user can quickly accumulate information. Without sacrificing the evaluation accuracy, the measurement results can be obtained with fewer questions, thus improving the measurement efficiency.

[0253] Furthermore, the question design considers both difficulty level and evaluation perspective. By balancing the evaluation perspectives of the questions, it is easier to avoid omissions and obtain more comprehensive and accurate measurement results.

[0254] Finally, the question set design allows the scoring process to occur after a question set is completed. Users can modify their answers within the question set without affecting the evaluation process. This makes the testing process more flexible and helps improve the accuracy and reliability of the measurement results.

[0255] Corresponding to the above method embodiments, this specification also provides embodiments of adaptive testing devices. Figure 4 A schematic diagram of an adaptive testing apparatus provided in one embodiment of this specification is shown.

[0256] like Figure 4 As shown, in this adaptive testing apparatus, the scale used for adaptive testing comprises multiple item groups divided into sequentially associated measurement phases, each item group including one or more items; the apparatus includes:

[0257] The answer result acquisition module 402 is used to acquire the user's first answer result for the first question group in the first measurement stage of the multiple measurement stages;

[0258] Answer score determination module 404 is used to determine the user's first answer score based on the first answer result;

[0259] The question group routing module 406 is used to determine a second question group from two or more question groups in the second measurement stage based on the first answer score, according to the correspondence between the answer score and the question group; the second measurement stage is the next measurement stage after the first measurement stage;

[0260] The question set providing module 408 is used to provide the second question set to the user.

[0261] based on Figure 4 The embodiments of this specification also provide some specific implementation schemes of the method, which are described below.

[0262] Optionally, any target question set included in the scale can satisfy the following first condition: the information content function of the target question set reaches its peak at the target difficulty of the target question set; the information content function is used to represent the correlation between the question set information content of the target question set and the difficulty value of each question in the target question set and the user's ability value.

[0263] Optionally, any target question group included in the scale may satisfy the following second condition: the number of questions in the target question group corresponding to each assessment angle of the scale satisfies a preset quantity threshold condition; the quantity threshold condition includes being greater than or equal to a first quantity threshold and less than a second quantity threshold.

[0264] Optionally, the scale has multiple response paths, each of which corresponds to a sequence of question groups selected from each of the multiple measurement stages; the target question group sequence corresponding to any target response path in the multiple response paths can satisfy the following third condition: the feature information content of the target question group sequence is greater than a preset information content threshold; the feature information content is determined based on the information content of all questions in the target question group sequence.

[0265] Optionally, the adaptive testing device can also be used to: acquire the target information content of the question group to be built; calculate the absolute deviation value of the candidate questions in the candidate question pool based on the target information content; the absolute deviation value represents the absolute value of the difference between the current information content and the target information content, wherein the current information content is used to represent the question group information content when the candidate questions are placed in the question group to be built; calculate the cumulative weight value corresponding to the candidate questions according to the preset quantity threshold conditions that the number of questions in each evaluation angle of the scale should meet, the evaluation angle of the selected questions in the question group to be built, and the evaluation angle of the candidate questions; and determine whether to select the candidate questions into the question group to be built based on the absolute deviation value and the cumulative weight value.

[0266] Optionally, obtaining the target information content of the question group to be built may specifically include: selecting a preset number of questions from the candidate question pool to form a pre-selected question set; the preset number is determined by multiplying the preset number of questions in the question group to be built by a preset question quantity coefficient; the question quantity coefficient is a value greater than or equal to 1; calculating the total information content of the pre-selected question set based on the information content of each question in the pre-selected question set; and determining the quotient of the total information content and the question quantity coefficient of the pre-selected question set as the target information content of the question group to be built.

[0267] Optionally, the step of selecting a preset number of questions from the candidate question pool to form a pre-selected question set may specifically include: dividing the candidate question pool into multiple candidate sub-question pools matching different difficulty levels according to the difficulty value of each question in the candidate question pool; and selecting a preset number of questions from the candidate sub-question pools corresponding to the difficulty level of the question group to be built to form a pre-selected question set.

[0268] Optionally, the step of calculating the absolute deviation value of the candidate questions in the candidate question pool based on the target information amount specifically includes: determining the expected information amount of the candidate questions based on the target information amount, the total information amount of the selected questions in the question group to be built, and the number of candidate questions in the question group to be built; and calculating the absolute deviation value of the information amount of the candidate questions based on the information amount of the candidate questions and the expected information amount.

[0269] Optionally, the step of calculating the cumulative weight value corresponding to the candidate item based on the preset quantity threshold condition that the number of items for each assessment angle of the scale should meet, the assessment angles of the selected items in the question group to be built, and the assessment angles of the candidate items may specifically include: obtaining the preset quantity threshold condition that the number of items for each assessment angle of the scale should meet; determining the weight value corresponding to each assessment angle of the scale based on the assessment angles of the selected items in the question group to be built and the quantity threshold condition; and determining the cumulative weight value of the candidate item based on the assessment angles of the candidate items and the weight values ​​corresponding to each assessment angle of the scale.

[0270] Optionally, determining the weight value corresponding to each assessment angle of the scale based on the assessment angles of the selected items in the question group to be built and the quantity threshold condition may specifically include: determining the number of selected items in the question group to be built corresponding to each assessment angle of the scale; determining the weight value of the assessment angles in the question group to be built whose corresponding number of selected items is less than a first quantity threshold as a first value; determining the weight value of the assessment angles in the question group to be built whose corresponding number of selected items is greater than or equal to the first quantity threshold and less than a second quantity threshold as a second value less than the first value; and determining the weight value of the assessment angles in the question group to be built whose corresponding number of selected items is greater than or equal to the second quantity threshold as a third value less than the second value.

[0271] Optionally, determining the cumulative weight value of the candidate items based on the evaluation angles of the candidate items and the weight values ​​corresponding to each evaluation angle of the scale may specifically include: calculating the supplementary weight values ​​corresponding to each evaluation angle of the selected items in the question group to be built based on the weight values ​​corresponding to each evaluation angle of the selected items in the question group to be built; and determining the cumulative weight value of the candidate items based on the evaluation angles of the candidate items and the weight values ​​and supplementary weight values ​​corresponding to each evaluation angle of the selected items in the question group to be built.

[0272] Optionally, determining whether to select the candidate questions into the question group to be built based on the absolute deviation value and the cumulative weight value may specifically include: calculating the standardized absolute deviation value corresponding to the absolute deviation value; calculating the standardized cumulative weight value corresponding to the cumulative weight value; and selecting the candidate question with the largest sum of standardized absolute deviation value and standardized cumulative weight value among all candidate questions in the candidate question pool into the question group to be built.

[0273] Optionally, determining a second question group from two or more question groups in the second measurement stage based on the first answer score, according to the correspondence between the answer score and the question group, may specifically include: obtaining the score threshold condition corresponding to a specified question group among the two or more question groups associated with the first question group in the second measurement stage; determining whether the first answer score satisfies the score threshold condition corresponding to the specified question group, and obtaining a first judgment result; if the first judgment result is yes, then determining the specified question group as the second question group.

[0274] Optionally, determining a second question group from two or more question groups in the second measurement stage based on the first answer score, according to the correspondence between the answer score and the question group, may specifically include: obtaining the division scores corresponding to one or more question group pairs involved in the two or more question groups associated with the first question group in the second measurement stage; the question group pair consists of two question groups with adjacent difficulty levels in the two or more question groups in the second measurement stage; for a target question group pair in the one or more question group pairs, if the first answer score meets a first score threshold condition, then the question group with a relatively lower difficulty level in the target question group pair is determined as the second question group; the first score threshold condition includes at least that the first answer score is less than the target division score of the target question group pair; if the first answer score meets a second score threshold condition, then the question group with a relatively higher difficulty level in the target question group pair is determined as the second question group; the second score threshold condition includes at least that the first answer score is greater than or equal to the target division score.

[0275] Optionally, before obtaining the respective division scores of one or more question pairs involved in two or more question groups associated with the first question group in the second measurement stage, the method may further include: determining the user ability threshold corresponding to each question pair in the second measurement stage based on the preset population ratio corresponding to each question group in the second measurement stage; the user ability threshold is used to reflect the ability differences of users assigned to different question groups in the corresponding question pair; and determining the division score corresponding to the question pair based on the user ability threshold corresponding to the question pair.

[0276] Optionally, the adaptive testing device is further configured to: calculate the user's overall score based on the user's answer scores for multiple actual answer question groups included in the actual answer path; the multiple actual answer question groups include one question group from each of the multiple measurement stages of the scale.

[0277] It is understood that the modules mentioned above refer to computer programs or program segments used to perform one or more specific functions. Furthermore, the distinction between these modules does not imply that the actual program code must also be separate.

[0278] The above is a schematic scheme of an adaptive testing device according to this embodiment. It should be noted that the technical solution of this adaptive testing device and the technical solution of the adaptive testing method described above belong to the same concept. For details not described in detail in the technical solution of the adaptive testing device, please refer to the description of the technical solution of the adaptive testing method described above.

[0279] Figure 5 A structural block diagram of a computing device 500 according to one embodiment of this specification is shown. The components of the computing device 500 include, but are not limited to, a memory 510 and a processor 520. The processor 520 is connected to the memory 510 via a bus 530, and a database 550 is used to store data.

[0280] The computing device 500 also includes an access device 540, which enables the computing device 500 to communicate via one or more networks 560. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 540 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a Near Field Communication (NFC) interface.

[0281] In one embodiment of this specification, the above-described components of the computing device 500 and Figure 5 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 5 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0282] Computing device 500 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). Computing device 500 can also be a mobile or stationary server.

[0283] The processor 520 is used to execute the following computer-executable instructions, which, when executed by the processor, implement the steps of the above-described adaptive testing method.

[0284] The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device and the technical solution of the adaptive testing method described above belong to the same concept. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the adaptive testing method described above.

[0285] An embodiment of this specification also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the adaptive testing method described above.

[0286] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium belongs to the same concept as the technical solution of the adaptive testing method described above. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the adaptive testing method described above.

[0287] An embodiment of this specification also provides a computer program, wherein when the computer program is executed in a computer, it causes the computer to perform the steps of the above-described adaptive testing method.

[0288] The above is an illustrative example of a computer program according to this embodiment. It should be noted that the technical solution of this computer program belongs to the same concept as the technical solution of the aforementioned adaptive testing method. Details not described in detail in the computer program's technical solution can be found in the description of the technical solution of the aforementioned adaptive testing method.

[0289] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0290] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added or removed according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0291] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, because according to the embodiments in this specification, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.

[0292] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0293] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. An adaptive testing method, characterized in that: The scale for adaptive testing includes a plurality of question groups divided into a plurality of measurement stages associated in sequence, each question group includes one or more questions; the method includes: Obtaining a first answer result of a user for a first question group in a first measurement stage in the multiple measurement stages; Determining a first answer score of the user according to the first answer result; According to the correspondence between the answer scores and the question groups, based on the first answer scores, a second question group is determined from two or more question groups in the second measurement phase; the second measurement phase is the next measurement phase after the first measurement phase; The second question set is provided to the user.

2. The method according to claim 1, characterized in that Any target question group included in the scale satisfies the following first condition: the information amount function of the target question group reaches a peak value at the target difficulty of the target question group; the information amount function is used to represent the correlation between the question group information amount of the target question group and the difficulty value of each question in the target question group and the ability value of the user.

3. The method according to claim 2, characterized in that Any target question group included in the scale satisfies the following second condition: the number of questions in the target question group corresponding to each evaluation angle of the scale satisfies a preset quantity threshold condition; the quantity threshold condition includes being greater than or equal to a first quantity threshold and less than a second quantity threshold.

4. The method according to claim 3, characterized in that The scale has a plurality of answer paths, each of the plurality of answer paths corresponds to a question group sequence consisting of a question group selected from each measurement stage of the plurality of measurement stages; The target question group sequence corresponding to any target answer path among the multiple answer paths satisfies the following third condition: the characteristic information amount of the target question group sequence is greater than a preset information amount threshold; the characteristic information amount is determined based on the information amount of all questions in the target question group sequence.

5. The method according to claim 1, characterized in that The method further comprises: Obtain the target amount of information for the question group to be created; Based on the target information amount, the absolute deviation value of the candidate questions in the candidate question pool is calculated; the absolute deviation value represents the absolute value of the difference between the current information amount and the target information amount, wherein the current information amount is used to represent the information amount of the question group when the candidate questions are placed in the question group to be created; Calculate the cumulative weight values ​​corresponding to the candidate questions according to the preset quantity threshold conditions that the number of questions at each evaluation angle of the scale should meet, the evaluation angles of the selected questions in the to-be-created question group, and the evaluation angles of the candidate questions; Based on the absolute deviation value and the cumulative weight value, it is determined whether to select the candidate question into the group of questions to be created.

6. The method according to claim 5, characterized in that The step of obtaining the target information amount of the question group to be created specifically includes: A preset number of questions are selected from the candidate question pool to form a pre-selected question set; the preset number is determined by multiplying the preset number of questions in the question group to be constructed by a preset question quantity coefficient; the question quantity coefficient is a value greater than or equal to 1; Calculating the total amount of information in the pre-selected topic set according to the amount of information of each topic in the pre-selected topic set; The quotient of the total amount of information and the question amount coefficient of the pre-selected question set is determined as the target information amount of the question group to be constructed.

7. The method according to claim 6, characterized in that The step of selecting a preset number of questions from the candidate question pool to form a pre-selected question set specifically includes: According to the difficulty value of each question in the candidate question pool, the candidate question pool is divided into a plurality of candidate sub-question pools matching different difficulty levels; A preset number of questions are selected from a pool of candidate sub-questions corresponding to the difficulty level of the question group to be constructed to form a pre-selected question set.

8. The method according to claim 5, characterized in that The step of calculating the absolute deviation values ​​of the candidate questions in the candidate question pool based on the target information amount specifically includes: Determining the expected amount of information of the candidate questions based on the target amount of information, the total amount of information of the selected questions in the group of questions to be created, and the number of the to-be-selected questions in the group of questions to be created; Based on the information amount of the candidate questions and the expected information amount, an absolute deviation value of the information amount of the candidate questions is calculated.

9. The method according to claim 5, characterized in that The calculating of the cumulative weight values ​​corresponding to the candidate questions according to the preset quantity threshold conditions that the quantity of questions at each evaluation angle of the scale should meet, the evaluation angles of the selected questions in the to-be-created question group, and the evaluation angles of the candidate questions specifically includes: Obtaining a preset quantity threshold condition that the number of questions at each evaluation angle of the scale should meet; Determine the weight value corresponding to each evaluation angle of the scale according to the evaluation angles of the selected questions in the to-be-created question group and the quantity threshold condition; The cumulative weight values ​​of the candidate questions are determined according to the evaluation angles of the candidate questions and the weight values ​​corresponding to the evaluation angles of the scale.

10. The method according to claim 9, characterized in that Determining the weight values ​​corresponding to the evaluation angles of the scale according to the evaluation angles of the selected questions in the to-be-created question group and the quantity threshold condition specifically includes: Determine the number of selected topics included in the to-be-created topic group corresponding to each evaluation angle of the scale; Determine the weight value of the evaluation angle corresponding to the number of selected questions in the to-be-created question group being less than the first quantity threshold as a first value; Determine the weight value of the evaluation angle whose number of selected questions in the to-be-created question group is greater than or equal to the first number threshold and less than the second number threshold as a second value less than the first value; The weight value of the evaluation angle corresponding to the number of selected questions in the to-be-created question group is greater than or equal to the second number threshold is determined as a third value that is smaller than the second value.

11. The method according to claim 9, characterized in that Determining the cumulative weight value of the candidate questions according to the evaluation angles of the candidate questions and the weight values ​​corresponding to the evaluation angles of the scale specifically includes: Based on the weight values ​​corresponding to the evaluation angles of the selected questions in the to-be-created question group, calculate the supplementary weight values ​​corresponding to the evaluation angles of the selected questions in the to-be-created question group; The cumulative weight values ​​of the candidate questions are determined according to the evaluation angles of the candidate questions and the weight values ​​and supplementary weight values ​​corresponding to the evaluation angles of the selected questions in the to-be-created question group.

12. The method according to claim 5, characterized in that The determining, based on the absolute deviation value and the cumulative weight value, whether to select the candidate question into the group of questions to be created specifically includes: Calculating a standardized absolute deviation value corresponding to the absolute deviation value; Calculating a normalized cumulative weight value corresponding to the cumulative weight value; The candidate question having the largest sum of the standardized absolute deviation value and the standardized cumulative weight value among the candidate questions in the candidate question pool is selected into the question group to be constructed.

13. The method according to claim 1, characterized in that The determining, based on the correspondence between the answer scores and the question groups and the first answer scores, a second question group from two or more question groups in the second measurement phase specifically includes: Obtaining a score threshold condition corresponding to a specified question group among two or more question groups associated with the first question group in the second measurement phase; Determine whether the first answer score meets the score threshold condition corresponding to the specified question group, and obtain a first determination result; If the first judgment result is yes, the designated question group is determined as the second question group.

14. The method according to claim 1, wherein: The determining, based on the correspondence between the answer scores and the question groups and the first answer scores, a second question group from two or more question groups in the second measurement phase specifically includes: Obtaining the division scores corresponding to one or more question group pairs involved in two or more question groups associated with the first question group in the second measurement phase; the question group pair is composed of two question groups with adjacent difficulty levels among the two or more question groups in the second measurement phase; For a target question pair among the one or more question pairs, If the first answering score satisfies a first score threshold condition, determining the question group with a relatively low difficulty level in the target question group pair as the second question group; the first score threshold condition at least includes that the first answering score is less than the target division score for the target question group pair; If the first answering score meets the second score threshold condition, the question group with a relatively high difficulty level in the target question group pair is determined as the second question group; the second score threshold condition at least includes that the first answering score is greater than or equal to the target division score.

15. The method according to claim 14, characterized in that Before obtaining the division scores corresponding to one or more question group pairs involved in two or more question groups associated with the first question group in the second measurement phase, the method further includes: Determine the user ability threshold corresponding to each question group pair in the second measurement phase according to the preset population ratio corresponding to each question group in the second measurement phase; the user ability threshold is used to reflect the ability difference of users divided into different question groups in the corresponding question group pair; The division scores corresponding to the question set pairs are determined according to the user ability thresholds corresponding to the question set pairs.

16. The method according to claim 1, wherein: The method further comprises: The user's comprehensive score is calculated based on the user's answer scores for multiple actual answer question groups included in the actual answer path; the multiple actual answer question groups include a question group for each of the multiple measurement stages of the scale.

17. An adaptive testing device, characterized in that: The scale for adaptive testing includes a plurality of question groups divided into a plurality of measurement stages associated in sequence, each question group includes one or more questions; the device includes: An answer result acquisition module is used to obtain a first answer result of a user for a first question group in a first measurement stage in the multiple measurement stages; A question answering score determination module, used to determine the first question answering score of the user according to the first question answering result; a question group routing module, configured to determine a second question group from two or more question groups in a second measurement phase according to the correspondence between the answer scores and the question groups and based on the first answer scores; the second measurement phase is the next measurement phase after the first measurement phase; A question set providing module is used to provide the second question set to the user.

18. A computing device, characterized in that include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the adaptive testing method described in any one of claims 1 to 16 are implemented.

19. A computer-readable storage medium, characterized in that: It stores computer executable instructions, which, when executed by a processor, implement the steps of the adaptive testing method described in any one of claims 1 to 16.

20. A computer program, characterized in that When the computer program is executed in a computer, the computer is caused to execute the steps of the adaptive testing method according to any one of claims 1 to 16.

Images