Chinese hierarchical reading ability evaluation system, text evaluation system and method

Abstract

The embodiment of the invention provides a Chinese hierarchical reading ability evaluation system and method and a text evaluation system and method. The system specifically comprises a standardized measuring scale construction module used for generating a reading ability evaluation question set based on a reference text corpus, each reference text being associated with at least one group of question units; for each group of question units, acquiring evaluation data of a plurality of users, respectively calibrating reading ability values of the users and text difficulty values of the reference texts associated with each group of question units in combination with the project reaction theoretical model, and calibrating the reading ability values and the text difficulty values to the same standardized measuring scale; and the dynamic evaluation module is used for selecting question units from the reading ability evaluation question set to form an ability evaluation question group according to feature data or evaluation requirements of a target user, and outputting a reading ability evaluation result based on answer data of the target user to the ability evaluation question group in combination with the standardized measuring scale. According to the system, joint evaluation of the text difficulty and the reading level is realized through a standardized measuring scale.

Description

Technical Field

[0001] This application relates to electronic data processing technology, specifically to a Chinese graded reading ability assessment system, a text assessment system, and a method. Background Technology

[0002] Texts are an important vehicle and resource for reading and learning, and text selection is the starting point and foundation of reading learning. my country's graded reading system began at the beginning of this century, and its development history is relatively short. Existing Chinese graded reading systems are mostly based on the idea of ​​a grade system. By selecting quantifiable text features that influence reading difficulty, mathematical formulas are used to express the functional relationship between text features and text difficulty, ultimately establishing a text readability model that can give the difficulty level (grade level, etc.) corresponding to any text, thus achieving a rough correspondence between texts and student grade levels.

[0003] However, besides its sequential nature, reading development also exhibits significant individual differences. Even within the same developmental stage, individuals demonstrate marked variations in reading ability due to differences in cognitive development, oral communication skills, and reading experience. Therefore, the current grading system cannot achieve precise, personalized matching, limiting its effectiveness in selecting reading materials and promoting reading development for students. Summary of the Invention

[0004] To address the aforementioned issues, the purpose of this application is to provide a Chinese graded reading ability assessment system, a text assessment system, and a method.

[0005] In a first aspect, embodiments of this application provide a Chinese graded reading ability assessment system, including:

[0006] The standardized scale construction module is used to generate a reading ability assessment question set based on a benchmark text corpus. Each benchmark text is associated with at least one set of question units. For each set of question units, assessment data from multiple users are collected. Combined with the item response theory model, the reading ability value of the users and the text difficulty value of the benchmark text associated with each set of question units are respectively calibrated to the same standardized scale.

[0007] The dynamic assessment module is used to select question units from the reading ability assessment question set to form an ability assessment question group based on the target user's characteristic data or assessment needs, and output the reading ability assessment result based on the target user's answer data to the ability assessment question group and the standardized scale.

[0008] Secondly, embodiments of this application also provide a Chinese graded reading text assessment system, including:

[0009] The standardized scale construction module is used to generate a reading ability assessment question set based on a benchmark text corpus. Each benchmark text is associated with at least one set of question units. For each set of question units, assessment data from multiple users are collected. Combined with the item response theory model, the reading ability value of the users and the text difficulty value of the benchmark text associated with each set of question units are respectively calibrated to the same standardized scale.

[0010] The text difficulty labeling module is used to train a text readability model using a benchmark text corpus and labeled text difficulty values; and to output the text difficulty value of the target text using the trained text readability model.

[0011] Thirdly, embodiments of this application also provide a method for assessing Chinese graded reading ability, including:

[0012] A reading comprehension assessment question set is generated based on a benchmark text corpus, with each benchmark text associated with at least one set of question units;

[0013] For each set of questions, assessment data from multiple users was collected. Combined with the Item Response Theory model, the reading ability values ​​of the users and the text difficulty values ​​of the benchmark text associated with each set of questions were respectively calibrated to the same standardized scale.

[0014] Based on the characteristic data or assessment needs of the target users, select question units from the reading ability assessment question set to form an ability assessment question group;

[0015] Based on the target user's responses to the assessment questions, and combined with the standardized scale, the reading ability assessment results are output.

[0016] Fourthly, embodiments of this application also provide a method for evaluating Chinese graded reading texts, including:

[0017] A reading comprehension assessment question set is generated based on a benchmark text corpus, with each benchmark text associated with at least one set of question units;

[0018] For each set of questions, assessment data from multiple users was collected. Combined with the Item Response Theory model, the reading ability values ​​of the users and the text difficulty values ​​of the benchmark text associated with each set of questions were respectively calibrated to the same standardized scale.

[0019] A text readability model is trained using a benchmark text corpus and labeled text difficulty values.

[0020] Using the trained text readability model, output the text difficulty value of the target text.

[0021] Fifthly, embodiments of this application also provide an electronic device, including a processor and a memory, the memory storing a computer program, the processor executing the computer program stored in the memory to implement the system in the first or second aspect described above.

[0022] Sixthly, embodiments of this application also provide a computer program, including computer instructions, which, when executed by a processor, implement the system described in the first or second aspect above.

[0023] In a seventh aspect, embodiments of this application also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the system described in the first or second aspect above.

[0024] This application has the following beneficial effects:

[0025] In the solution provided in the first aspect of the embodiments of this application, this application constructs a standardized scale based on item response theory to achieve a joint assessment of text difficulty and reading level, and establishes a correspondence between student reading ability, text difficulty and predicted reading comprehension rate. This system is more in line with the essence of graded reading and is more conducive to the development of children's reading ability. Attached Figure Description

[0026] The accompanying drawings, which are provided to further illustrate this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application.

[0027] Figure 1 This illustration shows a structural diagram of a Chinese graded reading ability assessment system provided in an embodiment of this application;

[0028] Figure 2 A flowchart illustrating the creation of a standardized measuring scale provided in an embodiment of this application is shown;

[0029] Figure 3 The flowchart illustrates a method for generating cloze test questions provided in an embodiment of this application.

[0030] Figure 4 This application provides a flowchart illustrating a method for assigning questions to a user, as illustrated in an embodiment of this application.

[0031] Figure 5 This document illustrates a flowchart of a standardized measuring scale based on the Rasch model, provided in an embodiment of this application.

[0032] Figure 6 This application provides a flowchart for assessing a user's reading ability according to an embodiment of the present application.

[0033] Figure 7 This illustration shows a structural diagram of a Chinese graded reading text assessment system provided in an embodiment of this application;

[0034] Figure 8 This illustration shows a structural diagram of a Chinese graded reading joint assessment system provided in an embodiment of this application;

[0035] Figure 9 A flowchart of a Chinese graded reading ability assessment method provided in an embodiment of this application is shown;

[0036] Figure 10 A flowchart of a Chinese graded reading text evaluation method provided in an embodiment of this application is shown. Detailed Implementation

[0037] To make the technical solutions and advantages of the embodiments of this application clearer, the exemplary embodiments of this application will be described in further detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not an exhaustive list of all embodiments. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.

[0038] In the description of this application, it should be understood that the terms "vertical", "horizontal", "inner", "outer", "upper", "lower", "front", "rear", "left", "right", "center", "longitudinal", "transverse", "length", "width", "thickness", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0039] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0040] In this application, unless otherwise expressly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0041] In response to the problems raised in the background art, this application provides a Chinese graded reading ability assessment system. This system constructs a standardized scale based on item response theory to achieve a joint assessment of text difficulty and reading level, which is more in line with the essence of graded reading.

[0042] Figure 1 A schematic diagram of the structure of a Chinese graded reading ability assessment system provided in an embodiment of this application is shown. Figure 1 As shown, the system includes:

[0043] The standardized scale construction module 101 is used to generate a reading ability assessment question set based on a benchmark text corpus. Each benchmark text is associated with at least one set of question units. For each set of question units, assessment data from multiple users are collected. Combined with the item response theory model, the reading ability value of the users and the text difficulty value of the benchmark text associated with each set of question units are respectively calibrated to the same standardized scale.

[0044] The dynamic assessment module 102 is used to dynamically match question units from the reading ability assessment question set to form an ability assessment question group based on the target user's characteristic data or assessment needs, and output the reading ability assessment results through a standardized scale based on the target user's answer data to the ability assessment question group.

[0045] The two functional modules will be explained in detail below.

[0046] First, regarding the standardized measuring scale construction module 101:

[0047] This application designs and constructs a standardized scale that simultaneously measures text difficulty and reading ability. Figure 2 A flowchart illustrating the creation of a standardized measuring scale provided in an embodiment of this application is shown. Figure 2 As shown, the process includes:

[0048] Step 201: Obtain the baseline text corpus.

[0049] In one embodiment, texts that meet specified requirements can be selected from existing texts as benchmark texts. Specifically, the specified requirements may include at least one of the following: (1) the total number of words in the benchmark texts is within a preferred range (e.g., between 80 and 2000 words); (2) the values, themes, and language of the benchmark texts meet the reading needs of users of the Chinese graded reading ability assessment system; (3) the benchmark texts are not in the teaching materials used by users of the Chinese graded reading ability assessment system; and (4) the reading difficulty of the benchmark texts is suitable for users of the Chinese graded reading ability assessment system. This application exemplarily selects 240 benchmark texts, and the following examples will continue to use this number, but it should be understood that this number should not be regarded as a limitation on the scope of protection of this application.

[0050] Step 202: Generate a set of reading comprehension assessment questions based on a benchmark text corpus.

[0051] In some embodiments, the reading ability assessment question set in step 202 can be used to assess the user's reading ability and reflect the differences in reading ability among multiple users. Specifically, it can take various forms, such as reading comprehension questions, cloze test questions, etc.

[0052] The following section uses a cloze test as an example to illustrate the specific implementation of step 202 in the reading comprehension assessment question set. Figure 3 This document presents a flowchart illustrating a method for generating cloze test questions. (For example...) Figure 3 As shown, the method includes:

[0053] Step 2021: Obtain any benchmark text from the benchmark text corpus.

[0054] In theory, all benchmark texts in the benchmark text corpus can be used to generate reading comprehension question units. However, in practical applications, considering factors such as the size of the question set and processing complexity, only a portion of the benchmark texts can be used to generate reading comprehension question units each time. For any given benchmark text, it can be processed according to... Figure 3 The method shown generates a corresponding set of cloze test question units.

[0055] Step 2022: Perform word segmentation and part-of-speech tagging on the above benchmark text, and select N non-repeating words as test words, where N is a positive integer and the N test words are not adjacent to each other.

[0056] There are multiple ways to select test words, which will be explained below with reference to an example.

[0057] In one embodiment, the baseline text can be divided into an option area and a fixed text area. Test words are selected only within the option area, while the fixed text area assists test takers in understanding the entire text or its context. Specifically, the fixed text area may include the first and last sentences of the baseline text, meaning test word selection can begin from the second sentence. When selecting test words, they can be chosen at fixed word intervals, such as every 7 or 15 words. Alternatively, non-adjacent words can be selected, provided the ratio of test words to the total number of words or characters meets preset requirements. Selection can also be based on various factors such as text information content and word frequency. For example, a two-factor within-subjects experimental design of 2 (word selection ratio: 7:1, 15:1) × 2 (option difficulty: easy, difficult) was used to explore the impact of these two factors and their interaction on the validity of the cloze test. The results showed that a high word selection ratio effectively mitigated the context fragmentation caused by overly dense options, while highly difficult distractors significantly improved the validity of the test for deep semantic understanding. In the experiment, the validity of the cloze test was optimal when the ratio of selected words was 15:1 (i.e., one word was selected as the test word every 15 words) and the difficulty of the options was "difficult" (i.e., both distractors had the same part of speech as the test word and were consistent with the sentence context, but not with the context of the text).

[0058] In one embodiment, a set of unselectable words that are not considered as test words is pre-defined. Before determining the test word, selected words can be used as candidate test words. It is then determined whether each candidate word is unselectable; if so, it is discarded (no longer considered as a candidate test word). Specifically, unselectable words can be set using word sets, discrimination rules, etc. For example, unselectable words can include proper nouns such as names of people and places, which can be set using a set of proper nouns; unselectable words can also include time-related terms (such as 1972, today) and quantity-specific terms (such as one hundred, one thousand), which can be set using discrimination rules (such as judging based on context), or directly identified by a trained model. This avoids including proper nouns, time-related terms, and quantity-specific terms—which are often difficult to judge in context—as options.

[0059] In one embodiment, test words selected from the same original text are not repeated. Specifically, if a candidate test word appears for the second time in the same original text, that candidate test word is discarded. This avoids duplicate options.

[0060] In one embodiment, test words can be selected using at least one strategy. If, when using a certain test word selection strategy, the number of test words ultimately selected in the original text does not reach the basic number (e.g., 5), then it can be determined that the original text is not suitable for generating cloze test questions. Alternatively, the number of test words can be supplemented to the basic number using other test word selection strategies.

[0061] Step 2023: For each test word, generate at least two corresponding distractors, combine the test word and the corresponding distractors into a question unit, and take all the question units as a set of question units (or a question unit group) associated with the above benchmark text.

[0062] In some embodiments, any distractor item satisfies the following conditions: the distractor item and the test word have the same part of speech, and within a preset statistical range, the co-occurrence frequency of the words preceding and following the distractor item is higher than that of the test word. Specifically, the co-occurrence frequency of the words preceding the test word and the distractor item appears more frequently in sequence than the words preceding the test word and the test word itself, and the co-occurrence frequency of the words following the test word and the distractor item appears more frequently in sequence than the words following the test word and the test word itself. The word preceding the test word refers to the word before the test word in the original text, and the word following the test word refers to the word after the test word in the original text. This aims to ensure that the automatically generated distractor items conform to sentence grammar as much as possible.

[0063] In some embodiments, the distractor must also meet the following condition: the distractor and the corresponding test word must have the same word length. For example, if the test word is a two-character word, then the distractor should also have a word length of 2.

[0064] In some embodiments, the interference item must also meet the following condition: the interference item is a non-low-frequency word. The non-low-frequency word here can also be statistically determined based on a preset statistical range. Taking a children's book corpus as an example, the non-low-frequency word can be a word with a cumulative frequency of over 95% in the children's book corpus. Whether the interference item is a non-low-frequency word can also be determined using other corpora outside the preset statistical range.

[0065] In some embodiments, distractors must also meet the following condition: they do not conform to the textual context. This ensures that, in a cloze test consisting of a test word and distractors, only the test word is the correct option when considering the textual context.

[0066] This step involves statistical analysis of the frequency of word combinations. When conducting the analysis, existing corpora or specially generated corpora can be used, i.e., the analysis is conducted within a preset statistical range.

[0067] To facilitate the understanding of the above process of generating interference items, the following takes the text "After being passed down for thousands of years, the ancient waterwheel was finally replaced by a diesel-powered or electric-powered water pump", where the word "passed down" is the test word, as an example for illustration:

[0068] The word segmentation result of this text is "ancient / of / waterwheel / in / passed down / thousands of years / after / finally being / diesel-powered / or / electric-powered / of / water pump / replaced". Among them, the test word "passed down" is a verb, the word before the test word is "in", and the word after the test word is "thousands of years".

[0069] First, within the preset statistical scope (such as a specific corpus), select words with the same词性 (both are verbs) as the test word "passed down" as candidate interference items. Then, within the preset statistical scope, screen out candidate interference items where the combination of the word before the test word and the candidate interference item in sequence (i.e., "‘in’ + interference item") appears more frequently than the combination of the word before the test word and the test word in sequence (i.e., "‘in’ + ‘passed down’"), and where the combination of the word after the test word and the candidate interference item in sequence (i.e., "interference item + ‘thousands of years’") appears more frequently than the combination of the word after the test word and the test word in sequence (i.e., "‘passed down’ + ‘thousands of years’"). Finally, determine the final interference items from the screened candidate interference items.

[0070] Suppose the candidate interference items are as shown in the following table:

[0071]

[0072] Among them, the frequency of the combination of "in" and "waiting, searching, sleeping, passing through, experiencing, sleeping in eternal rest, sleeping, dreaming, continuing" is greater than that of "‘in’ + passed down", and the frequency of the combination of "passing through, experiencing, sleeping in eternal rest, waiting" and "thousands of years" is greater than that of "‘passed down’ + ‘thousands of years’". Thus, taking the intersection of the two sets of words in the table, the candidate interference items that meet the frequency requirements can be obtained, namely "passing through, experiencing, sleeping in eternal rest, waiting", and further determine the final interference items from them.

[0073] Step 203, collect evaluation data of multiple users for each group of question units.

[0074] In one embodiment, since the number of reading ability question units is large and the difficulty range is wide, the method of vertical and horizontal equating is adopted to conduct test paper administration for the question units. Each group of question units is assigned to multiple users for answering, and each user answers multiple groups of question units.

[0075] In one example, the process of assigning question units to users can be seen in Figure 4 the shown process. As Figure 4 shown, this process includes:

[0076] Step 2031: Pre-calibrate the applicable grade level for each benchmark text.

[0077] The applicable grade level is also the applicable grade level of the question unit associated with the benchmark text. The applicable grade level of the benchmark text can be determined in various ways. For example, when the benchmark text is selected from a textbook, the grade level of the textbook corresponding to the benchmark text can be directly identified as the applicable grade level of the benchmark text; another example is that when there is already a relatively authoritative grade level for the benchmark text, that grade level can be directly adopted; yet another example is that the applicable grade level can be automatically identified based on the textual characteristics of the benchmark text, such as the frequency of Chinese characters; and so on.

[0078] Step 2032: Assign question units to users of the corresponding grade level according to the applicable grade level, and generate evaluation data based on the answer data.

[0079] For example, when assigning question units to users of corresponding grade levels, the following two types of questions can be assigned to the same user: 1) question units of the same grade as the user; 2) question units of one grade higher than the user's grade, and / or question units of one grade lower than the user's grade. The number of question units of type 1) is greater than the number of question units of type 2). For example, 35 test texts applicable to grade 1 can be divided into 10 question sets, each with 3-4 cloze passages. Each test taker needs to complete 3 question sets. Some test takers complete all 3 question sets of the same grade, while others complete 3 question sets with one passage from a grade lower or higher. This achieves anchoring of question units.

[0080] This application provides a specific example: students from grades 1 to 6 were recruited as subjects, ensuring that each baseline text contained at least 100 data points, meaning that each question unit associated with each baseline text was answered by at least 100 students. All students were native speakers of Chinese and had no obvious intellectual or language developmental delays or significant physical disabilities. Ultimately, 4359 students from grades 1 to 6 of a typical primary school (2304 boys and 2055 girls) were selected to participate in this study, with a mean age of 114.17 ± 20.08 months.

[0081] Step 204: Using the Project Response Theory model, the user's reading ability value and the text difficulty value of the benchmark text associated with each question unit are respectively calibrated to the same standardized scale.

[0082] Item Response Theory (IRT) is a collective term for a series of psychostatistical models. Common IRT models include the Logistic model, the Rasch model, Lord's normal oval curve model, and more than twenty others.

[0083] Among the many models in Project Response Theory, the Rasch model is more suitable for the purpose of jointly assessing text difficulty and reading ability due to its simplicity and relative objectivity. This application provides an example of using the Rasch model to establish a standardized scale. Figure 5 This diagram illustrates a flowchart for establishing a standardized measurement scale based on the Rasch model. Figure 5 As shown, the process includes:

[0084] Step 2041: Perform hypothesis testing on the assessment data using the Rasch model. If the hypothesis is satisfied, analyze the assessment data using the Rasch model to obtain the reading ability value of each student and the question difficulty value of each question unit.

[0085] In one example, hypothesis testing for the Rasch model includes at least a oneness test and / or a local independence test. If the evaluation data meets the assumptions after hypothesis testing, then the evaluation data is considered suitable for analysis using the Rasch model.

[0086] Regarding the unidimensionality test, it is generally considered that if the eigenvalue of the first factor in the exploratory factor analysis is more than 3 to 5 times greater than the eigenvalue of the second factor, there is sufficient reason to consider it a unidimensional test. This application exemplarily conducted three exploratory factor analyses on the assessment data collected in step 203 according to the user's grade level. The results showed that in the lower (grades 1-2), middle (grades 3-4), and upper (grades 5-6) grades, the ratios of the eigenvalues ​​of the first and second factors were 21.19, 26.71, and 23.85, respectively, all greater than 5. Therefore, the assessment data satisfies the unidimensionality hypothesis.

[0087] Regarding the test of local independence, it is generally considered that if the absolute value of the Q3 statistic is less than 0.2, the local independence hypothesis is satisfied. In the example provided in this application, the mean absolute value of the Q3 statistic for the 240 benchmark texts ranges from 0.032 to 0.167, which is less than 0.2. Therefore, the evaluation data satisfies the local independence hypothesis.

[0088] The parameters of the Rasch model are typically used as indicators to measure an individual's ability relative to the difficulty of the questions. In this model, question difficulty is usually defined as the trait level that gives an individual a 0.5 probability of answering correctly when their ability level is equal to the question difficulty. That is, according to the existing formula of the Rasch model, when a user's reading ability value is equal to the text difficulty value of a benchmark text, the user's reading comprehension accuracy for that benchmark text is approximately 50%. In other words, theoretically, a student with a reading ability value of 0.5 has a 50% probability of accurately understanding a benchmark text with a text difficulty value of 0.5, a higher probability of accurately understanding a benchmark text with a text difficulty value of 0.4, and a lower probability of accurately understanding a benchmark text with a text difficulty value of 0.6. In one embodiment, a reading comprehension accuracy rate other than 50% can be set according to actual needs, and the existing formula of the Rasch model can be adjusted accordingly based on the reset reading comprehension accuracy rate. The existing formula of the Rasch model is as follows:

[0089]

[0090] Where P represents the user's reading comprehension accuracy of the benchmark text, θ represents the user's reading ability value, and δ represents the text difficulty value.

[0091] Step 2042: Calculate the text difficulty value of each benchmark text based on the question difficulty value of the question unit associated with it.

[0092] For example, the average difficulty value of all question units associated with each benchmark text can be calculated as the text difficulty value of that benchmark text. For instance, if a set of cloze test questions generated corresponding to a benchmark text contains 15 question units, the average difficulty value of these 15 question units can be calculated as the text difficulty value of that benchmark text. Those skilled in the art should understand that different methods for calculating text difficulty values ​​can also be selected based on the type and number of question units, assessment requirements, etc.

[0093] This completes the work on... Figure 2 Description of the process shown.

[0094] pass Figure 2 The process shown constructs a refined, equidistant standardized scale based on item response theory. Compared with the existing "same scale" that corresponds text difficulty to student grade (which is essentially a pseudo-fit based on the average reading level of the group), this application can more accurately reflect the correspondence between students' reading ability, text difficulty and predicted reading comprehension rate, which is more in line with the essence of graded reading and more conducive to the development of children's reading ability.

[0095] The dynamic evaluation module 102 will be described in detail below.

[0096] The standardized reading scale construction module 101 completes the construction of the standardized reading scale, achieving a correspondence between users' reading abilities and text difficulty. During the standardized reading scale construction process, reading ability values ​​of some users have already been collected through question units, and through the dynamic assessment module 102, other users can also obtain specific reading ability values. Figure 6 This illustrates a flowchart for assessing a user's reading ability, such as... Figure 6 As shown, the process includes:

[0097] Step 601: Select question units from the reading ability assessment question set to form an ability assessment question group based on the target user's characteristic data or assessment needs.

[0098] In one embodiment, parameters are pre-calibrated for each group of question units in the reading ability assessment question set. This allows for rapid matching of question units with corresponding calibration parameters based on assessment needs and target user characteristic data during the selection process, thereby forming a higher-matching ability assessment question set with the assessment needs and target users. For example, the calibration parameters for question units may include:

[0099] ① Number of questions: This refers to the specific number of question units contained in a set of question units associated with the base text.

[0100] ② Reliability: Based on the evaluation data collected by the standardized scale construction module 101, the internal consistency coefficient of the text is calculated as the test reliability of the benchmark text.

[0101] ③ Discrimination: Based on the evaluation data collected by the standardized scale construction module 101, the total correlation calculation method is adopted, that is, the correlation between the user's raw test score on a benchmark text and the user's reading ability value is calculated as the discrimination of the text.

[0102] ④ Text type: The baseline texts are categorized into three types: literary texts, informational texts, and mixed texts.

[0103] ⑤ Text difficulty (grade): The average of the baseline text difficulty predicted by the traditional graded reading system.

[0104] ⑥ Text "reading ability" value: that is, the text difficulty value obtained in the standardized scale construction module 101.

[0105] ⑦ Text features: Text features extracted from the baseline text at the character, word, sentence, and paragraph levels.

[0106] In one embodiment, a user only needs to complete a subset of the question units in the reading ability assessment question set to obtain their reading ability score through a standardized scale. For the same target user, at least one set of question units can be selected to form an assessment question set. This question set can be automatically generated by electronic devices or by the user selecting the question units themselves.

[0107] Using students from different grades in elementary school as examples, we will now explain some principles and methods for compiling test papers:

[0108] ① Regarding the test materials and number of questions, considering that the test time is usually one class period, and that first-grade students have fewer characters and slower reading speeds, the ability assessment question sets selected for first-grade students should ideally include related question units from 6 benchmark texts. At the same time, to ensure test reliability, the number of question units in all ability assessment question sets for first-grade students should ideally be greater than 50. Fifth-grade students read longer texts and have more question units; therefore, the ability assessment question sets selected for fifth-grade students should ideally include related question units from 6 benchmark texts, while the ability assessment question sets selected for second, third, and fourth-grade students should ideally include related question units from 7-8 benchmark texts.

[0109] ② In terms of reliability and discrimination, try to select benchmark texts with high reliability (reliability > 0.5) and high discrimination (discrimination > 0.5).

[0110] ③ In terms of text type, the benchmark texts selected for first and second grade students are all literary texts. Starting from the third grade, informational texts are added to the test. From the fifth grade onwards, the test is conducted with half literary texts and half informational texts.

[0111] ④ In terms of text difficulty and reading ability, select benchmark texts that are close to the corresponding grade level but have a certain range of difficulty.

[0112] In one embodiment, the test paper compilation process can be dynamic. Specifically, a subset of question units can be provided to the target user initially, and subsequent question units can be provided based on the target user's actual performance. In one example, if the total number of question units in each assessment set is fixed at N (N is an integer greater than or equal to 2), then A (A is less than N) question units can be selected and provided to the target user initially. Based on the target user's performance on these A question units, the subsequent NA question units can be dynamically selected. For example, if the target user's total score on these A question units is lower than the average score, the difficulty of the subsequently selected question units can be reduced. In another example, if the total number of question units in each assessment set is not fixed, then A question units can be provided to the target user initially. Based on the target user's performance, at least one subsequent question unit can be dynamically selected until the target user's performance meets preset requirements. For example, if the user's total score on these A question units is much higher than the average score, the difficulty of the subsequently selected question units can be increased until the total score approaches the average score.

[0113] Step 602: Based on the target users' response data, combined with standardized metrics, output reading ability assessment results.

[0114] Specifically, the target user's reading ability score can be obtained by combining the difficulty values ​​of the question units within the ability assessment question set, or the text difficulty value of the benchmark text associated with the ability assessment question set, and the target user's correct answer rate for each question unit with a standardized scale. Furthermore, the final reading ability assessment result, in addition to the reading ability score and its related explanation, may also include reading intervention strategies for the target user and recommended reading texts. For example, for a target user whose reading ability score is lower than that of their peers, it is recommended to strengthen their reading intervention and recommend texts suitable for improving their reading ability. Or, based on the target user's reading ability score, texts with a difficulty level matching their reading ability score can be recommended.

[0115] This completes the work on... Figure 6 The process shown is described in the diagram, as well as the description of the dynamic evaluation module 102.

[0116] Based on the Chinese graded reading ability assessment system provided in this application, a detailed and equidistant standardized scale was constructed based on item response theory to achieve a joint assessment of text difficulty and reading level. It establishes a correspondence between students' reading ability, text difficulty and predicted reading comprehension rate. This system is more in line with the essence of graded reading and is more conducive to the development of children's reading ability.

[0117] This application also provides validation of a Chinese leveled reading ability assessment system. Specifically, it includes:

[0118] 1. Selecting participants

[0119] A total of 739 students from grades one to six of a typical primary school were selected as subjects, including 381 boys and 358 girls. The average age of the subjects was 110.28 ± 17.58 months. All subjects' native language was Chinese, and none of them had obvious intellectual or language developmental delays or obvious physical disabilities such as visual or hearing impairments.

[0120] 2. Experimental Design

[0121] A two-factor mixed experimental design was used, consisting of 3 (difficulty levels: easy [independent level], medium [teaching level], difficult [frustration level]) × 5 (grades: 1, 2, 3, 4, 5). The dependent variables were reading performance on the validation test, subjective difficulty rating, and text-based interest rating. "Easy" texts were defined as those with a prediction accuracy of 80% or higher; "medium" texts as those with a prediction accuracy between 70% and 80%; and "difficult" texts as those with a prediction accuracy below 70%.

[0122] The students were required to complete a total of 7 texts, ranging from easy (2 texts), medium (3 texts), to difficult (2 texts).

[0123] 3. Experimental Procedure

[0124] First, the reading ability scores of the test students were obtained using the Chinese graded reading ability assessment system provided in this application. Then, based on their reading ability scores, texts at easy, medium, and difficult levels were individually matched to them to explore whether there were differences in the test students' reading performance on these texts, their subjective perception of difficulty, and their interest based on the texts. Simultaneously, the Chinese graded reading ability assessment system provided in this application was compared with traditional graded reading systems to explore the potential differences between these two text reading recommendation methods.

[0125] To control for the position effect, the order in which texts of different difficulties appeared in the test booklets was randomized across all test booklets. Students read and completed cloze tests on seven articles sequentially, and after each article, they were given a questionnaire to assess their difficulty and interest in that text.

[0126] 4. Verification and Judgment

[0127] The results showed that, both overall and by grade level, the students' reading comprehension test scores decreased as the text difficulty increased, their subjective perception of difficulty increased, and their interest in the text decreased. This indicates that the Chinese graded reading ability assessment system provided in this application effectively achieves personalized matching between readers and texts.

[0128] Furthermore, a comparative analysis of the Chinese leveled reading ability assessment system provided in this application and traditional leveled reading systems reveals significant differences in the difficulty of suitable reading texts among test students within the same grade level. Moreover, when test students are given texts predicted as suitable by the Chinese leveled reading ability assessment system, regardless of the grade level predicted by the system, their reading comprehension accuracy meets the requirements. This demonstrates that the Chinese leveled reading ability assessment system provided in this application has clear advantages over traditional leveled reading systems.

[0129] This application also provides a Chinese graded reading text assessment system, such as... Figure 7 As shown, the system includes:

[0130] The standardized scale construction module 101 is used to generate a reading ability assessment question set based on a benchmark text corpus. Each benchmark text is associated with at least one set of question units. For each set of question units, assessment data from multiple users are collected. Combined with the item response theory model, the reading ability value of the users and the text difficulty value of the benchmark text associated with each set of question units are respectively calibrated to the same standardized scale.

[0131] The text difficulty labeling module 103 is used to train a text readability model using a benchmark text corpus and labeled text difficulty values; and to output the text difficulty value of the target text using the trained text readability model.

[0132] The standardized measuring scale construction module 101 has been described in detail above. The following section will focus on the text difficulty calibration module 103.

[0133] Based on the standardized scale construction module 101, this application constructs a standardized scale that can achieve the correspondence between students' reading ability and text difficulty. However, currently only the benchmark text corpus has the corresponding text difficulty value. In order to obtain text difficulty values ​​for other texts, a text readability model can be established.

[0134] The training process of a text readability model is illustrated below. First, a training set for the text readability model is generated based on benchmark texts in a benchmark text corpus and their pre-labeled text difficulty values. Then, key text features are extracted from the benchmark texts in the training set. These extracted text features can encompass multiple text features at the character, word, sentence, and text levels. A machine learning model algorithm is then used to construct the text readability model, with the key text features of each benchmark text serving as the model output. The model is then trained using the training set.

[0135] In one embodiment, key text features can be determined as follows: based on a systematic review of existing text complexity theory models (Mesmer et al., 2012) and readable literature, text features at the character, word, sentence, and paragraph levels are extracted for each text, and different algorithms are used to select features to find key text features.

[0136] In one embodiment, multiple machine learning algorithms, such as Support Vector Machines, Random Forests, and Lasso Regression, can be used to construct readability models. Each model's construction and validation employs 5-fold cross-validation to avoid overfitting. For example, a benchmark text corpus can be randomly divided into five parts: four as training text sets and one as test text sets. A readability model is constructed by fitting the relationship between key text features and text difficulty values. Then, the constructed readability model is used to predict the text difficulty values ​​of the texts in the test text set to obtain two evaluation metrics for model performance (R², R², R², R²). 2 R 2 The larger the value, the better the model fit; the smaller the mean standard deviation, the better the model fit (the absolute deviation between the text difficulty value already labeled in the benchmark text and the text difficulty value predicted by the readability model).

[0137] For each readability model, the above procedure is repeated four times until each of the five texts is used as the test text set. The average of these five runs on each evaluation metric is then calculated to represent the performance of the readability model. Finally, the performance of each readability model is compared, and the model with the best performance is selected as the final readability model.

[0138] This completes the work on... Figure 7 The description shown is of a Chinese graded reading text assessment system.

[0139] To facilitate understanding of the Chinese graded reading ability assessment system and the Chinese graded reading text assessment system provided in this application, a specific example is given below.

[0140] This example provides a Chinese graded reading joint assessment system that integrates a Chinese graded reading ability assessment system and a Chinese graded reading text assessment system. For example... Figure 8 As shown, the system includes:

[0141] The standardized scale construction module 101 is used to generate a reading ability assessment question set based on a benchmark text corpus. Each benchmark text is associated with at least one set of question units. For each set of question units, assessment data from multiple users are collected. Combined with the item response theory model, the reading ability value of the users and the text difficulty value of the benchmark text associated with each set of question units are respectively calibrated to the same standardized scale.

[0142] The dynamic assessment module 102 is used to dynamically match question units from the reading ability assessment question set to form an ability assessment question group based on the target user's characteristic data or assessment needs, and output the reading ability assessment results through a standardized scale based on the target user's answer data to the ability assessment question group.

[0143] The text difficulty labeling module 103 is used to train a text readability model using a benchmark text corpus and labeled text difficulty values; and to output the text difficulty value of the target text using the trained text readability model.

[0144] The reading suitability assessment module 104 is used to determine whether the target text is suitable for the target user to read based on the target user's reading ability value and the text difficulty value of the target text.

[0145] Application Scenario 1: Reading Ability Assessment

[0146] The Chinese graded reading joint assessment system provides a user interface where users can input personal data, select the assessment purpose, and obtain the assessment report generated by the system after answering the questions provided by the system.

[0147] Specifically, users can register and input their personal data through the system interface. For example, a student could input "age 12 years old" and "grade 6," and select "routine assessment" as their assessment request. Based on the input age, grade, and assessment request, the system automatically extracts three initial sets of questions from a set of reading comprehension assessment questions. These questions are in cloze test format. After the user completes the three initial sets of questions, the system analyzes the answer data in real time and finds that the accuracy rate for the last two sets of questions is only 30%, and the time spent answering them has significantly increased. Based on this, the system dynamically adjusts its question-generating strategy, automatically reducing the difficulty of subsequent questions and adding three sets of questions with lower text difficulty. Finally, the system integrates all the answer data, calculates the user's reading comprehension score using a standardized scale, and generates a structured assessment report. This report includes an evaluation of the reading comprehension score relative to the user's grade level, as well as recommended texts based on the user's reading comprehension score.

[0148] Application Scenario 2: Text Difficulty Assessment

[0149] The Chinese graded reading joint assessment system provides a user interface where users can input text content, select the assessment purpose, and obtain the assessment report generated by the system.

[0150] Specifically, users can upload text to be evaluated, such as a chapter from a newly compiled textbook, and select "Difficulty Assessment" as the evaluation purpose. The system first extracts the text features of the text to be evaluated. Then, using a pre-trained text readability model, it outputs a text difficulty value for the text to be evaluated based on the extracted text features, such as 0.5. The system can further generate a difficulty assessment report based on this text difficulty value. The difficulty assessment report may include a recommended category for the text, suggested reading materials, etc., such as recommending it to be categorized as "Recommended Reading Material Library for Fifth Grade". When a user with the same reading ability value of 0.5 logs into the system, this text may appear first in the personalized recommendation list.

[0151] Application Scenario 3: Personalized Reading Adaptability Assessment

[0152] After logging into the system, users input their target reading text (e.g., uploading Chapter 3 of the extracurricular reading material *The Insect World*) on the reading suitability interface. Simultaneously, the system automatically retrieves the user's recent reading ability assessment result (e.g., a reading ability score of 0.5). The text difficulty assessment module analyzes the uploaded target reading text in real time, extracts its text features, and inputs them into a pre-trained text readability model to obtain an output text difficulty value. The user's reading ability value is matched with the text difficulty value; if the difference between the two does not reach a threshold, the target reading text is determined to be suitable for the user.

[0153] This completes the work on... Figure 8 The description shown is of the Chinese graded reading joint assessment system.

[0154] In some embodiments, the Chinese graded reading ability assessment system and the Chinese graded reading text assessment system can obtain the answer data, reading ability scores, and text difficulty scores of different users, respectively. Based on this, the baseline text corpus and the reading ability assessment question set can be updated regularly, and the standardized scale can be updated regularly by combining newly collected user answer data and reading ability scores. Thus, as education develops, the standardized scale can keep pace with the gradual improvement of children's reading ability.

[0155] Based on the same concept as the Chinese leveled reading ability assessment system provided in this application, this application also provides a Chinese leveled reading ability assessment method, such as... Figure 9 As shown, the method includes:

[0156] Step 901: Generate a reading comprehension assessment question set based on a benchmark text corpus, with each benchmark text associated with at least one set of question units;

[0157] Step 902: Collect assessment data from multiple users for each set of questions, and combine the item response theory model to calibrate the reading ability value of the users and the text difficulty value of the benchmark text associated with each set of questions, and calibrate the two to the same standardized scale.

[0158] Step 903: Based on the characteristic data or assessment needs of the target users, select question units from the reading ability assessment question set to form an ability assessment question group;

[0159] Step 904: Based on the target users' responses to the competency assessment questions, output the reading competency assessment results using a standardized scale.

[0160] It should be noted that the above method does not need to be executed in the order of steps 901 to 904 every time. After executing steps 901 and 902 once, the results of these two steps (such as a standardized measuring scale) can be used directly without repeating these two steps.

[0161] For details on the specific implementation methods of each step above, please refer to the description of the Chinese graded reading ability assessment system above, which will not be repeated here.

[0162] Based on the same concept as the Chinese graded reading text evaluation system provided in this application, this application also provides a Chinese graded reading text evaluation method, such as... Figure 10 As shown, the method includes:

[0163] Step 1001: Generate a reading comprehension assessment question set based on a benchmark text corpus, with each benchmark text associated with at least one set of question units;

[0164] Step 1002: Collect assessment data from multiple users for each set of questions, and combine the item response theory model to calibrate the reading ability value of the users and the text difficulty value of the benchmark text associated with each set of questions, and calibrate the two to the same standardized scale.

[0165] Step 1003: Train a text readability model using a benchmark text corpus and labeled text difficulty values;

[0166] Step 1004: Using the trained text readability model, output the text difficulty value of the target text.

[0167] It should be noted that the above method does not need to be executed in the order of steps 1001 to 1004 every time. After executing steps 1001 to 1003 once, the results of these three steps (the trained text readability model) can be used directly without repeating these two steps.

[0168] For details on the specific implementation methods of each step above, please refer to the description of the Chinese graded reading text assessment system above, which will not be repeated here.

[0169] Furthermore, embodiments of the present invention also provide an electronic device, which includes at least a processor and a memory. The memory stores a computer program, and the processor is used to execute the computer program stored in the memory to implement the Chinese graded reading text assessment method and / or Chinese graded reading ability assessment method provided in the embodiments of this application, or to carry the Chinese graded reading text assessment system and / or Chinese graded reading ability assessment system provided in the embodiments of this application.

[0170] It should be understood that in the embodiments of the present invention, the memory may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. Non-volatile memory includes: read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The memory may also be a memory remotely configured relative to the processor, and the remotely configured memory can be connected to a server via a network.

[0171] This application also provides a computer program including one or more computer instructions. When the computer instructions are loaded and executed on a processor, all or part of the program generates the Chinese graded reading ability assessment method and / or the Chinese graded reading text assessment method according to the embodiments of this application, or carries the Chinese graded reading text assessment system and / or the Chinese graded reading ability assessment system provided by the embodiments of this application.

[0172] This application also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the Chinese graded reading ability assessment method and / or the Chinese graded reading text assessment method described in this application, or carries the Chinese graded reading text assessment system and / or the Chinese graded reading ability assessment system provided in this application.

[0173] The solutions in this application embodiment can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0174] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0175] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0176] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0177] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0178] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0179] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A Chinese graded reading ability assessment system, characterized in that, include: The standardized scale construction module is used to generate a reading ability assessment question set based on a benchmark text corpus, with each benchmark text associated with at least one set of question units; For each set of questions, assessment data from multiple users was collected. Combined with the Item Response Theory model, the reading ability values ​​of the users and the text difficulty values ​​of the benchmark text associated with each set of questions were respectively calibrated to the same standardized scale. The dynamic assessment module is used to select question units from the reading ability assessment question set to form an ability assessment question group based on the target user's characteristic data or assessment needs, and output the reading ability assessment result based on the target user's answer data to the ability assessment question group and the standardized scale.

2. The system according to claim 1, characterized in that, The reference text must meet at least the following requirements: The total number of words in the reference text is within the preferred range; The values, themes, and language of the benchmark text meet the reading needs of users targeted by the Chinese graded reading ability assessment system; The benchmark text is not included in the textbooks used by users of the Chinese graded reading ability assessment system. The reading difficulty of the benchmark text is suitable for users of the Chinese graded reading ability assessment system.

3. The system according to claim 1, characterized in that, The system also includes: The text difficulty labeling module is used to train a text readability model using a benchmark text corpus and labeled text difficulty values, and then outputs the text difficulty value of the target text using the trained text readability model.

4. The system according to claim 3, characterized in that, The system also includes: The reading suitability assessment module is used to determine whether the target text is suitable for the target user to read based on the target user's reading ability value and the text difficulty value of the target text.

5. The system according to claim 1, characterized in that, The reading comprehension assessment question set generated based on the benchmark text corpus includes: Obtain any benchmark text from the benchmark text corpus; The baseline text is segmented and tagged with parts of speech. N unique words are selected as test words, where N is a positive integer. The N test words are not adjacent to each other and are not proper nouns, time expressions, or quantity qualifiers. For each test word, at least two corresponding distractors are generated. The test word and its corresponding distractors are combined into a question unit. All the question units are combined into a set of question units associated with the benchmark text.

6. The system according to claim 5, characterized in that, Any interference term satisfies the following condition: The interference item and the test word have the same part of speech, and within the preset statistical range, the frequency of the word before the test word and the interference item appearing in sequence is greater than the frequency of the word before the test word and the test word appearing in sequence, and the frequency of the word after the test word and the interference item appearing in sequence is greater than the frequency of the word after the test word and the test word appearing in sequence; the word before the test word refers to the word before the test word, and the word after the test word refers to the word after the test word.

7. The system according to claim 6, characterized in that, Each interference term also satisfies the following condition: The interference terms and test words have the same word length; The frequency of the interference item in the specified corpus is greater than a preset value; The interference item does not conform to the context of the benchmark text.

8. The system according to claim 1, characterized in that, The collection of evaluation data from multiple users for each set of questions includes: Pre-determine the applicable grade level for each benchmark text; Question units are assigned to users of the corresponding grade level based on their grade level, and evaluation data is generated based on the answer data.

9. The system according to claim 1, characterized in that, The process of separately calibrating the user's reading ability value and the text difficulty value of the benchmark text associated with each question unit, and then calibrating them to the same standardized scale, includes: The Rasch model hypothesis was tested on the assessment data. If the hypothesis was satisfied, the Rasch model was used to analyze the assessment data to obtain the reading ability value of each student and the question difficulty value of each question unit. The text difficulty value of each benchmark text is calculated based on the question difficulty value of the question unit associated with it.

10. The system according to claim 1, characterized in that, The step of selecting question units from the reading ability assessment question set to form an ability assessment question group includes: Parameters were pre-calibrated for each group of question units in the reading comprehension assessment question set; Based on the assessment needs and / or the characteristic data of the target users, match question units with corresponding calibration parameters to form a capability assessment question set.

11. A Chinese graded reading text assessment system, characterized in that, include: The standardized scale construction module is used to generate a reading ability assessment question set based on a benchmark text corpus, with each benchmark text associated with at least one set of question units; For each set of questions, assessment data from multiple users was collected. Combined with the Item Response Theory model, the reading ability values ​​of the users and the text difficulty values ​​of the benchmark text associated with each set of questions were respectively calibrated to the same standardized scale. The text difficulty labeling module is used to train a text readability model using a benchmark text corpus and labeled text difficulty values; and to output the text difficulty value of the target text using the trained text readability model.

12. The system according to claim 11, characterized in that, The text readability model, after training, outputs a text difficulty value for the target text, including: Extract key text features from the target text, input the key text features into the text readability model, and obtain the text difficulty value of the target text; the key text features cover text features at four levels: character, word, sentence, and text.

13. A method for assessing Chinese graded reading ability, characterized in that, include: A reading comprehension assessment question set is generated based on a benchmark text corpus, with each benchmark text associated with at least one set of question units; For each set of questions, assessment data from multiple users was collected. Combined with the Item Response Theory model, the reading ability values ​​of the users and the text difficulty values ​​of the benchmark text associated with each set of questions were respectively calibrated to the same standardized scale. Based on the characteristic data or assessment needs of the target users, select question units from the reading ability assessment question set to form an ability assessment question group; Based on the target user's responses to the assessment questions, and combined with the standardized scale, the reading ability assessment results are output.

14. A method for assessing Chinese graded reading texts, characterized in that, include: A reading comprehension assessment question set is generated based on a benchmark text corpus, with each benchmark text associated with at least one set of question units; For each set of questions, assessment data from multiple users was collected. Combined with the Item Response Theory model, the reading ability values ​​of the users and the text difficulty values ​​of the benchmark text associated with each set of questions were respectively calibrated to the same standardized scale. A text readability model is trained using a benchmark text corpus and labeled text difficulty values. Using the trained text readability model, output the text difficulty value of the target text.

15. An electronic device comprising a processor and a memory, the memory storing a computer program, characterized in that, The processor executes the computer program stored in the memory to implement the Chinese graded reading ability assessment system as described in any one of claims 1 to 10, or to implement the Chinese graded reading text assessment system as described in any one of claims 11 to 12.

16. A computer program comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the Chinese graded reading ability assessment system as described in any one of claims 1 to 10, or the Chinese graded reading text assessment system as described in any one of claims 11 to 12.

17. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the Chinese graded reading ability assessment system as described in any one of claims 1 to 10, or the Chinese graded reading text assessment system as described in any one of claims 11 to 12.

Images