Knowledge guiding method for AI point reading and related device thereof

Abstract

This application belongs to the field of intelligent learning assistance technology, specifically providing a knowledge guidance method and related equipment for AI-powered point-and-read reading. The method mainly includes: acquiring the target concept, its contextual text, and historically related concepts that the user is pointing to in the current reading material, as well as historical guidance records for these related concepts, including the number of successful guidance attempts and the total number of guidance attempts; calculating the semantic relevance between historically related concepts and the contextual text; calculating a historical utility index based on the historical guidance records; and inputting the semantic relevance and historical utility index into a preset evaluation model. This application achieves precise guidance based on the user's individual cognitive state and the immediate learning scenario, enabling the explanation content to adaptively connect with the user's existing knowledge and match the exploration focus of the current reading material, effectively improving the efficiency of the user's knowledge network formation and enhancing the information absorption capacity of in-depth reading.

Description

Technical Field

[0001] This application belongs to the field of intelligent learning assistance technology, and specifically relates to knowledge guidance methods and related equipment for AI-based point-and-reading. Background Technology

[0002] With the integration of artificial intelligence and educational technology, interactive learning applications such as intelligent point-and-read are becoming increasingly popular. These systems can identify the knowledge concepts selected by the user and automatically provide corresponding explanations. This technology provides learners with instant information support and improves the convenience of knowledge acquisition.

[0003] In the process of continuous reading and learning, users' need to understand the same concept will dynamically evolve as the reading progresses. When users encounter materials of different depths and focuses at different times, their knowledge background and search intentions have undergone substantial changes. They may shift from initial conceptual cognition to later mechanistic exploration or correlation analysis. Knowledge explanations are difficult to effectively support users to deepen and expand on the basis of existing cognition, affecting users' construction of knowledge systems. Summary of the Invention

[0004] This application provides a knowledge guidance method and related equipment for AI-powered point-and-read, which effectively solves the problem in existing technologies that fail to perceive the user's continuous learning process and dynamic cognitive intent by treating each point-and-read interaction as an isolated session, resulting in a mismatch between the guidance content and the user's actual needs. It effectively improves the efficiency of the user's knowledge network formation and enhances the information absorption ability of in-depth reading.

[0005] To achieve the above objectives, this application adopts the following technical solution:

[0006] Firstly, this application provides a knowledge guidance method for AI-powered point-and-read reading, comprising: acquiring the target concept, contextual text, and historical related concepts that the user is pointing to in the current reading material, as well as historical guidance records for the historical related concepts, the historical guidance records including the number of successful guidance sessions and the total number of guidance sessions; calculating the semantic relevance between the historical related concepts and the contextual text, calculating a historical utility index based on the historical guidance records, inputting the semantic relevance and historical utility index into a preset evaluation model to obtain a related evaluation value for each historical related concept; determining the knowledge scenario type for the current query based on the contextual text, and adjusting the related evaluation value according to the knowledge scenario type; calculating the knowledge mastery score for each historical related concept based on the historical guidance records, and calculating a comprehensive evaluation value for each historical related concept based on the knowledge mastery score and the adjusted related evaluation value; determining the final historical related concept from the historical related concepts based on the comprehensive evaluation value, and generating a personalized related explanation based on the final historical related concept, the target concept, and the contextual text.

[0007] Furthermore, obtaining historical related concepts of the target concept includes: in a preset cognitive association network, querying the node represented by the target concept as the target node, where nodes in the cognitive association network represent concepts, edges represent historical associations between concepts, and the weight of the edges represents the degree of association between concepts; extracting the neighboring nodes of the target node whose degree of association is greater than a preset first threshold to form historical related concepts.

[0008] Further, calculating the semantic relevance between historically related concepts and contextual text includes: obtaining the definition text of the historically related concepts; inputting the definition text and contextual text into a pre-trained semantic encoding model to obtain a first semantic vector and a second semantic vector; calculating the cosine similarity between the first semantic vector and the second semantic vector; comparing the cosine similarity with a preset second threshold; if the cosine similarity is lower than the second threshold, setting the semantic relevance to zero; otherwise, using the cosine similarity as the semantic relevance.

[0009] Furthermore, the knowledge scenario type for this query is determined based on the context text, including: extracting text features from the context text, such as text length, interrogative sentence structure markers, and specific keywords; inputting the text features into a pre-trained scenario classification model to obtain a probability distribution of knowledge scenario types; and extracting the knowledge scenario type with the highest probability value from the knowledge scenario type probability distribution as the knowledge scenario type for this query.

[0010] Furthermore, the association evaluation value is adjusted according to the knowledge scenario type, including: obtaining the historical number of successful guidances and the total number of guidances for historically associated concepts under the knowledge scenario type; determining whether the total number of guidances is lower than a preset third threshold: if so, the adjustment amount of the association evaluation value is zero; otherwise, a dynamic adjustment coefficient is calculated according to the knowledge scenario type, and the association evaluation value is adjusted according to the dynamic adjustment coefficient to obtain the adjusted association evaluation value.

[0011] Furthermore, the dynamic adjustment coefficient is calculated based on the knowledge scenario type, including: querying the preset evaluation value adjustment coefficient table according to the knowledge scenario type to obtain the basic adjustment coefficient; using the ratio of the historical successful guidance count to the total guidance count under the knowledge scenario type as the scenario adaptation factor; and multiplying the basic adjustment coefficient by the scenario adaptation factor to obtain the dynamic adjustment coefficient.

[0012] Furthermore, the knowledge mastery score for each historically related concept is calculated based on the historical guidance records, including: the historical guidance records also include the user's historical query frequency, recent query frequency change rate, and average accuracy of related tests for each historically related concept; the total number of historical queries, recent query frequency change rate, and average accuracy are standardized and the arithmetic mean is calculated to obtain the knowledge mastery score for each historically related concept.

[0013] Furthermore, a comprehensive evaluation value for each historically related concept is calculated based on the knowledge mastery score and the adjusted association evaluation value, including: a weighted average of the knowledge mastery score and the adjusted association evaluation value to obtain a comprehensive evaluation value for each historically related concept.

[0014] Secondly, this application provides a knowledge guidance device for AI-powered point-and-read functionality, comprising:

[0015] Data acquisition module: It is used to acquire the target concept that the user clicks on in the current reading material, the context text of the target concept and the historical related concepts, as well as the historical guidance records of the historical related concepts. The historical guidance records include the number of successful guidances and the total number of guidances.

[0016] The association calculation module is used to calculate the semantic association degree between historical associated concepts and contextual text, calculate the historical utility index based on historical guidance records, and input the semantic association degree and historical utility index into a preset evaluation model to obtain the association evaluation value of each historical associated concept.

[0017] Scene Adjustment Module: It is used to determine the knowledge scene type of this query based on the context text, and adjust the associated evaluation value according to the knowledge scene type.

[0018] The comprehensive assessment module is used to calculate the knowledge mastery score for each historically related concept based on the historical guidance records, and to calculate the comprehensive assessment value for each historically related concept based on the knowledge mastery score and the adjusted related assessment value.

[0019] Explanation generation module: It is used to determine the final historical association concept from the historical association concepts based on the comprehensive evaluation value, and generate personalized association explanations based on the final historical association concept, the target concept and the context text.

[0020] Thirdly, this application provides a readable storage medium storing computer program instructions, which are read and executed by a processor to perform steps of a knowledge guidance method for AI-based point-reading.

[0021] The beneficial effects of this application are:

[0022] This application obtains target concepts, contextual and historical related concepts and guidance records, integrates semantic relevance and historical utility indicators for evaluation, adjusts and associates knowledge mastery scores based on scenario type to calculate a comprehensive evaluation value, and selects concepts to generate personalized explanations. This effectively solves the problem in existing technologies that treat each interactive session as an isolated conversation, failing to perceive the user's continuous learning process and dynamic cognitive intent, resulting in a mismatch between guidance content and the user's actual needs. It achieves precise guidance based on the user's individual cognitive state and real-time learning scenario, enabling the explanation content to adaptively connect with the user's existing knowledge and match the exploration focus of the current reading material, effectively improving the efficiency of the user's knowledge network formation and strengthening the information absorption ability of in-depth reading.

[0023] Other features and advantages of this application will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures pointed out in the description and the accompanying drawings. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0025] Figure 1 A flowchart illustrating the knowledge guidance method for AI-based point-and-reading in this application is shown.

[0026] Figure 2 This application illustrates a flowchart of the process for determining the knowledge scenario type for this query.

[0027] Figure 3 A flowchart illustrating the process of adjusting the correlation assessment value in this application is shown;

[0028] Figure 4 A schematic diagram of the calculation process for the dynamic adjustment coefficients in this application is shown. Detailed Implementation

[0029] To address the problems raised in the background technology, this application obtains target concepts, contextual and historical related concepts and guidance records, integrates semantic relevance and historical utility indicators for evaluation, adjusts and associates knowledge mastery scores based on scenario type to calculate a comprehensive evaluation value, selects concepts to generate personalized explanations, effectively improves the efficiency of users' knowledge network formation, and enhances the information absorption ability of in-depth reading.

[0030] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0031] In some embodiments, such as Figure 1 As shown, this application provides a knowledge guidance method for AI-powered point-and-read functionality, including:

[0032] S1. Obtain the target concept that the user clicks on in the current reading material, the contextual text of the target concept, the historical related concepts, and the historical guidance records of the historical related concepts.

[0033] The target concept represents the core knowledge entity corresponding to the text selected by the user in digital reading materials through touch, click, or voice.

[0034] For example, in an article about biology, a user selects the word "photosynthesis," which is the target concept.

[0035] Contextual text represents a certain length of continuous text extracted from the paragraph or sentence containing the target concept, providing the specific context in which the target concept appears.

[0036] For example, when the target concept is "photosynthesis", the contextual text might contain a sentence that says, "Photosynthesis is the process by which plants use light energy to convert carbon dioxide and water into organic matter and oxygen."

[0037] Historically related concepts represent other knowledge concepts that are associated with the target concept. For example, "chloroplast" is a related concept to "photosynthesis".

[0038] Historical boot records represent a collection of data on the system's boot process for users. Historical boot records include at least the number of successful boots and the total number of boots.

[0039] For example, regarding the concept pair of "photosynthesis" and "chloroplast," there have been three guiding records in the system's history:

[0040] Record 1: Photosynthesis occurs in chloroplasts.

[0041] Record 2: Chloroplasts are organelles that perform photosynthesis.

[0042] Record 3: The thylakoid membrane in chloroplasts is the site of the light reaction.

[0043] For records 1 and 3, the user did not ask any follow-up questions and answered the subsequent quiz questions directly related to the guided content correctly, indicating successful feedback. For record 2, the user asked "Why?", indicating failed feedback. There are a total of 3 guided records, so the total number of guided sessions is 3. Among these 3 guided records, 2 had successful feedback, so the number of successful guided sessions is 2.

[0044] S2. Calculate the semantic relevance between historical related concepts and contextual text, calculate the historical utility index based on historical guidance records, input the semantic relevance and historical utility index into the preset evaluation model, and obtain the relevance evaluation value of each historical related concept.

[0045] Semantic relevance measures the semantic relevance between historical relevance concepts and the current context text, reflecting the immediate contextual relevance.

[0046] Historical utility metrics measure the success rate of using the historical association concept to interpret the target concept in historical interactions, reflecting long-term guiding effectiveness. See the formula below: ,in, Represents historical utility indicators. This represents the number of times the guidance was successful. This represents the total number of times the message has been guided.

[0047] The evaluation model is used to quantitatively integrate semantic relevance and historical utility metrics. For example, the evaluation model could be: ;in, The association evaluation value represents the historically associated concept. This value comprehensively reflects the immediate semantic relevance and historical guidance effectiveness between the historically associated concept and the target concept. Represents semantic relevance. , These represent the weights of semantic relevance and historical utility metrics, respectively.

[0048] S3. Determine the knowledge scenario type for this query based on the context text. The knowledge scenario type represents the learning stage or intent of the user's current query. Adjust the association evaluation value according to the knowledge scenario type so that the association evaluation value can better reflect the applicability of the historical association concept in a specific scenario.

[0049] Specifically, the following four knowledge scenario types can be preset, each corresponding to a typical knowledge guidance direction:

[0050] Concept introduction: This refers to the scenario where a user first encounters or needs to understand the definition, attributes, or core characteristics of a basic concept. Its textual features often include explanatory and descriptive language.

[0051] Comparative analysis: This refers to scenarios where users need to distinguish the differences and connections between two or more similar or easily confused concepts. Its textual features often include comparative and contrastive vocabulary.

[0052] Application of principles: This refers to scenarios where users, based on their understanding of concepts, further explore their operational mechanisms, causal relationships, or specific application scenarios. Its textual characteristics often include logical reasoning, conditional statements, or illustrative examples.

[0053] Knowledge review: This refers to scenarios where users review, summarize, or establish higher-level knowledge connections to previously learned concepts. Its textual features often include general, summarizing vocabulary or references to prior knowledge.

[0054] S4. Calculate the knowledge mastery score for each historically related concept based on the historical guidance record. The knowledge mastery score reflects the user's understanding and mastery of the historically related concept.

[0055] The comprehensive evaluation value of each historical related concept is calculated based on the knowledge mastery score and the adjusted association evaluation value. The comprehensive evaluation value combines the strength of the association between the historical related concept and the target concept after scenario adaptation, as well as the user's personal cognitive state, and serves as the basis for the final decision.

[0056] S5. Determine the final historical association concept from the historical association concepts based on the comprehensive evaluation value, and generate a personalized association interpretation based on the final historical association concept, the target concept, and the contextual text.

[0057] In some embodiments, obtaining historical related concepts of the target concept includes:

[0058] S1.1. In the pre-defined cognitive association network, the node represented by the query and the target concept is used as the target node. The nodes in the cognitive association network represent concepts, the edges represent the historical associations between concepts, and the weight of the edges represents the degree of association between concepts.

[0059] For example, if the system has previously used "chloroplast" to explain "photosynthesis," then there is an edge between the nodes "chloroplast" and "photosynthesis." The correlation degree can be specifically defined as the success rate of explaining the target node using nodes that have a historical association with the target node. For example, if the success rate of "chloroplast" explaining "photosynthesis" is approximately 0.67, then the correlation degree is 0.67.

[0060] S1.2. Extract the neighboring nodes of the target node whose correlation is greater than the preset first threshold to form a historical correlation concept.

[0061] The first threshold is used to filter out weak associations and can be set empirically. For example, the first threshold can be 0.1, which means that only strong associations with an association degree greater than 0.1 are considered. For example, if the weights of "chloroplast", "carbon dioxide", and "mitochondria" are 0.67, 0.45, and 0.08 respectively, it means that the adjacent concepts of "chloroplast" and "carbon dioxide" are extracted as historical association concepts.

[0062] In some embodiments, calculating the semantic relevance between historically associated concepts and contextual text includes:

[0063] S2.1. Obtain the definition text of the historical associated concept, and input the definition text and context text into the pre-trained semantic encoding model to obtain the first semantic vector and the second semantic vector.

[0064] Specifically, the standardized definition text of the historically related concept can be obtained from the preset concept definition knowledge base, which is used to store the mapping relationship between the concept and its standard definition.

[0065] Semantic encoding models, such as BERT, are deep learning models trained on large-scale corpora. They are used to map text strings of arbitrary length into fixed-dimensional vectors in a high-dimensional space. The semantic encoding model outputs a first semantic vector. Second semantic vector .

[0066] S2.2. Calculate the cosine similarity between the first semantic vector and the second semantic vector. The range of cosine similarity is [-1, 1], and the closer the value is to 1, the more semantically similar the meaning.

[0067] S2.3. Compare the cosine similarity with the preset second threshold. If the cosine similarity is lower than the second threshold, set the semantic association degree to zero; otherwise, use the cosine similarity as the semantic association degree.

[0068] The second threshold is used to determine whether semantic relevance is significant. It can be determined experimentally based on the actual application scenario. For example, it can be set to 0.3, if the cosine similarity... If the cosine similarity is less than 0.3, the semantic relevance is set to 0, indicating that the historical related concept has too low a semantic relevance to the current context and is insufficient to serve as a valid guiding candidate. Otherwise, the cosine similarity is directly set to 0. The value assigns semantic relevance. .

[0069] For example, obtain the semantic vectors of the definition text of "chloroplast" and the context text, and calculate the cosine similarity. If the value is 0.75, which is greater than the second threshold of 0.3, then the semantic relevance is... The value is 0.75, and the calculated historical utility index is approximately 0.67. These values ​​are then substituted into the evaluation model. , and If we take 0.6 and 0.4 respectively, the calculated correlation evaluation value is 0.718.

[0070] In some embodiments, such as Figure 2 As shown, the knowledge scenario type for this query is determined based on the context text, including:

[0071] S3.1. Extract text features from the context text, including text length, interrogative sentence markers, and specific keywords.

[0072] Text length represents the number of characters contained in the context text. For example, the context text "Photosynthesis is the process by which green plants use water and carbon dioxide to produce organic matter and release oxygen under sunlight" has 35 characters.

[0073] Determine whether the text contains interrogative words or question marks. If it does, the interrogative sentence flag is 1; otherwise, it is 0. For example, if the context text above does not contain interrogative sentences, then the interrogative sentence flag is 0.

[0074] Specific keywords represent predefined high-frequency words related to the four knowledge scenario types mentioned above. For example, if the keyword "is" is detected in the text, which is related to concept introduction, the keyword feature for concept introduction is recorded as 1. If other scenario keywords such as "distinction," "for example," and "before" are not detected, the corresponding feature is recorded as 0.

[0075] S3.2. Input the text features into the pre-trained scene classification model to obtain the probability distribution of knowledge scene types.

[0076] The scene classification model is a supervised machine learning classifier, such as a support vector machine, which has been trained on a large amount of text data labeled with four categories: "concept introduction," "comparison and analysis," "principle application," and "knowledge review." After receiving feature vectors, the model outputs a probability distribution vector P, where each component of the probability distribution vector P represents the predicted probability that the context text belongs to each knowledge scene type.

[0077] S3.3. Extract the knowledge scenario type with the highest probability value from the knowledge scenario type probability distribution as the knowledge scenario type for this query.

[0078] For example, the final output of the scene classification model is P=[0.85,0.05,0.08,0.02], where 0.85, 0.05, 0.08, and 0.02 represent the probabilities of the knowledge scene type being concept introduction, comparative analysis, principle application, and knowledge review, respectively. 0.85 is the maximum value, so concept introduction is determined to be the knowledge scene type of this query.

[0079] In some embodiments, such as Figure 3 As shown, the correlation evaluation value is adjusted according to the knowledge scenario type, including:

[0080] S3.4. Obtain the historical number of successful guidances and the total number of guidances for historically related concepts under the knowledge scenario type.

[0081] S3.5. Determine whether the total number of guidance times is lower than the preset third threshold: if so, the adjustment amount of the associated evaluation value is zero; otherwise, calculate the dynamic adjustment coefficient according to the knowledge scenario type, and then adjust the associated evaluation value according to the dynamic adjustment coefficient. The product of the dynamic adjustment coefficient and the associated evaluation value is used as the adjusted associated evaluation value.

[0082] The third threshold represents the minimum number of valid samples. For example, the third threshold can be set to 5, which means that at least 5 historical guidance records are required for the statistical results to be considered reliable.

[0083] In some embodiments, such as Figure 4 As shown, the dynamic adjustment coefficient is calculated based on the knowledge scenario type, including:

[0084] S3.5.1. Query the preset evaluation value adjustment coefficient table according to the knowledge scenario type to obtain the basic adjustment coefficient.

[0085] The base adjustment coefficient is used to characterize the system's preferences or strategies for different knowledge scenario types: If the knowledge scenario type is concept introduction, the user's goal at this stage is mainly to establish broad associations. The user needs to understand the new concept from multiple perspectives. The base adjustment coefficient can be set to 1.2 to improve the association evaluation value, increase the number of historically related concepts selected, and avoid missing potentially effective explanatory angles. If the knowledge scenario type is comparative analysis, the user's goal at this stage is mainly to accurately distinguish concepts. The base adjustment coefficient can be set to 0.9 to reduce the association evaluation value, decrease the number of historically related concepts selected, and prevent concept confusion. If the knowledge scenario type is principle application, the user's goal at this stage is to objectively understand the mechanism. The base adjustment coefficient can be set to 1.0. If the knowledge scenario type is knowledge review, the user's goal at this stage is to activate and consolidate memory. The base adjustment coefficient can be set to 1.1 to improve the association evaluation value, increase the number of historically related concepts selected, and thus awaken the user's long-term memory through multiple paths.

[0086] The ratio of the number of successful historical guidance sessions to the total number of guidance sessions under a specific knowledge scenario type is used as the scenario adaptation factor. The scenario adaptation factor reflects the historical performance of the historically related concept under a specific knowledge scenario type.

[0087] S3.5.2. Multiply the basic adjustment coefficient by the scene adaptation factor to obtain the dynamic adjustment coefficient. Dynamic adjustment coefficient This combines the inherent attributes of the knowledge scenario type with the historical performance of the related historical concept within that knowledge scenario type. When this occurs, it indicates that the system has adopted an encouraging or reinforcing strategy for the current knowledge scenario type, aiming to improve the relevance evaluation value. This indicates that the system has adopted a suppressive or cautious strategy towards the current knowledge scenario type, aiming to reduce the correlation evaluation value.

[0088] For example, if the knowledge scenario type is concept introduction, the basic adjustment coefficient is 1.2. If the historical successful guidance count and total guidance count for this knowledge scenario type are 7 and 10 respectively, then the calculated scenario adaptation factor is 0.7, and the dynamic adjustment coefficient is... The adjusted correlation coefficient is approximately 0.84, and the adjusted correlation coefficient is approximately 0.603.

[0089] In some embodiments, a knowledge mastery score for each historically associated concept is calculated based on historical guidance records, including:

[0090] S4.1. The history guidance record also includes the user's historical query frequency for each historical related concept, the recent query frequency change rate, and the average accuracy of related tests.

[0091] Historical query frequency represents the total number of times a user has actively queried this historically related concept over all past periods. For example, it can be calculated by counting the total number of query records for this concept in the query event log table.

[0092] The recent query frequency change rate represents the recent trend of user attention to the concept. It can be obtained by calculating the ratio of the number of queries in the most recent preset time window to the number of queries in an earlier time window of the same length. If the recent query frequency change rate is greater than 1, it means that the recent attention has increased. If the recent query frequency change rate is less than 1, it means that the attention has decreased.

[0093] The average accuracy rate of related tests represents the user's average correct answer rate in subsequent tests or exercises involving the historically related concept, reflecting the user's level of mastery of the historically related concept.

[0094] S4.2. Standardize the total number of historical queries, the recent query frequency change rate, and the average accuracy rate. For example, use the min-max normalization method to unify them to the [0,1] interval, and then calculate the arithmetic mean to obtain the knowledge mastery score for each historical related concept. The knowledge mastery score comprehensively reflects the user's query popularity, recent changes in attention, and test performance for the historical related concept. The higher the score, the better the user may have mastered the concept or the more interested they are in it.

[0095] For example, if the historical query frequency is 15 times, the recent query frequency change rate is 1.25, and the average accuracy rate of related tests is 0.8, then the calculated knowledge mastery score is 0.725.

[0096] In some embodiments, a comprehensive evaluation value for each historically related concept is calculated based on the knowledge mastery score and the adjusted association evaluation value, including:

[0097] S4.3. A weighted average of the knowledge mastery score and the adjusted association assessment score is used to obtain the comprehensive assessment score for each historical association concept. This comprehensive assessment score reflects the combined priority of the adjusted association assessment score based on the current knowledge scenario type and the user's knowledge mastery score. Refer to the formula: ;in, Represents the overall evaluation value. This represents the adjusted correlation assessment value. The score represents the knowledge mastery score. The weights representing the adjusted associated assessment values, The weighting of the knowledge mastery score is used to reflect the priority of the current query context and relevance. 0.7 is acceptable. 0.3 is acceptable. , It can be adjusted according to the actual application scenario.

[0098] For example, substituting the correlation assessment value of 0.603 and the knowledge mastery score of 0.725 into... The calculated comprehensive evaluation value is approximately 0.6396.

[0099] In some embodiments, determining the final historical association concept from historical association concepts based on a comprehensive evaluation value includes:

[0100] S5.1. Randomly select historical association concepts whose comprehensive evaluation value is greater than the preset fourth threshold as the final historical association concepts.

[0101] The fourth threshold represents the passing standard for the comprehensive evaluation value. The fourth threshold can be adjusted according to the type of knowledge scenario, avoiding the rigidity of the guidance mode caused by always selecting a single historical related concept.

[0102] If there are no historical association concepts with a comprehensive evaluation value greater than the fourth threshold, personalized association explanation data will not be generated. This avoids generating potentially inaccurate or invalid explanations when data support is insufficient or the association is weak, thus enhancing the reliability of the system.

[0103] For example, if the fourth threshold is 0.5, and the comprehensive evaluation values ​​of the historically associated concepts chloroplast and carbon dioxide are 0.6396 and 0.52 respectively, both greater than 0.5, then chloroplast or carbon dioxide is randomly selected as the final historically associated concept.

[0104] In some embodiments, generating personalized association interpretations based on the final historical association concept, the target concept, and the contextual text includes:

[0105] S5.2. From the preset template library, retrieve the basic explanation template that matches the knowledge scenario type determined in this query. The basic explanation template contains a text frame with a fixed structure and placeholders. From the historical guidance records, select the guiding statements between the final historical related concepts and the target concepts as content templates.

[0106] S5.3. Use the content template as the core explanation content and fill it into the placeholder of the basic explanation template. Then, through string replacement operation, form the initial explanation text.

[0107] S5.4. Analyze the stylistic features of the context text, such as formality and terminology density, and call a lightweight language style transfer model to adaptively adjust the wording of the initial explanatory text to generate the final explanatory text.

[0108] S5.5. The target concept, the final historical related concept, and the final explanatory text are encapsulated in a structured manner and output as a personalized related explanation.

[0109] For example, if the final associated concept is chloroplast, the basic explanation template could be: To understand the [target concept], you can refer to the [final historical associated concept]. Specifically, the [associated explanation] content template is: The site of photosynthesis is the chloroplast. An example of the generated final explanation text would be: To understand photosynthesis, you can refer to the chloroplast. Specifically, the site of photosynthesis is the chloroplast.

[0110] In some embodiments, this application provides a knowledge guidance device for AI-powered point-and-read functionality, comprising:

[0111] Data acquisition module: It is used to acquire the target concept that the user clicks on in the current reading material, the context text of the target concept and the historical related concepts, as well as the historical guidance records of the historical related concepts. The historical guidance records include the number of successful guidances and the total number of guidances.

[0112] The association calculation module is used to calculate the semantic association degree between historical associated concepts and contextual text, calculate the historical utility index based on historical guidance records, and input the semantic association degree and historical utility index into a preset evaluation model to obtain the association evaluation value of each historical associated concept.

[0113] Scene Adjustment Module: It is used to determine the knowledge scene type of this query based on the context text, and adjust the associated evaluation value according to the knowledge scene type.

[0114] The comprehensive assessment module is used to calculate the knowledge mastery score for each historically related concept based on the historical guidance records, and to calculate the comprehensive assessment value for each historically related concept based on the knowledge mastery score and the adjusted related assessment value.

[0115] Explanation generation module: It is used to determine the final historical association concept from the historical association concepts based on the comprehensive evaluation value, and generate personalized association explanations based on the final historical association concept, the target concept and the context text.

[0116] In some embodiments, this application provides a readable storage medium storing computer program instructions, which are read and executed by a processor to perform steps of a knowledge guidance method for AI-based point-and-read.

[0117] It should be noted that, in this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0118] Any references to memory, storage, database, or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory.

[0119] Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A knowledge guidance method for AI-powered point-and-read reading, characterized in that, include: The system obtains the target concept that the user clicks on in the current reading material, the context text of the target concept, and historical related concepts, as well as the historical guidance records of the historical related concepts. The historical guidance records include the number of successful guidance sessions and the total number of guidance sessions. Calculate the semantic relevance between the historical associated concepts and the context text, calculate the historical utility index based on the historical guidance records, and input the semantic relevance and historical utility index into a preset evaluation model to obtain the association evaluation value for each historical associated concept. The knowledge scenario type for this query is determined based on the context text, and the association evaluation value is adjusted accordingly. The knowledge mastery score for each historically related concept is calculated based on the historical guidance record, and the comprehensive evaluation value for each historically related concept is calculated based on the knowledge mastery score and the adjusted association evaluation value. Based on the comprehensive evaluation value, the final historical association concept is determined from the historical association concepts, and a personalized association interpretation is generated based on the final historical association concept, the target concept, and the contextual text.

2. The method according to claim 1, characterized in that, Obtaining the historical related concepts of the target concept, including: In the pre-defined cognitive association network, the node represented by the target concept is queried as the target node. The nodes in the cognitive association network represent concepts, the edges represent the historical associations between concepts, and the weight of the edges represents the degree of association between concepts. Extract the neighboring nodes of the target node whose correlation is greater than a preset first threshold to form a historical correlation concept.

3. The method according to claim 1, characterized in that, Calculating the semantic relevance between the historical related concepts and the contextual text includes: Obtain the definition text of the historical associated concept, and input the definition text and the context text into a pre-trained semantic encoding model to obtain the first semantic vector and the second semantic vector; Calculate the cosine similarity between the first semantic vector and the second semantic vector; The cosine similarity is compared with a preset second threshold. If the cosine similarity is lower than the second threshold, the semantic association degree is set to zero; otherwise, the cosine similarity is used as the semantic association degree.

4. The method according to claim 1, characterized in that, The knowledge scenario type for this query is determined based on the context text, including: Extract text features from the context text, including text length, interrogative sentence structure markers, and specific keywords; The text features are input into a pre-trained scene classification model to obtain the probability distribution of knowledge scene types; Extract the knowledge scenario type with the highest probability value from the probability distribution of knowledge scenario types as the knowledge scenario type for this query.

5. The method according to claim 4, characterized in that, The correlation evaluation value is adjusted according to the knowledge scenario type, including: Obtain the historical successful guidance count and total guidance count of the historical associated concept under the knowledge scenario type; Determine whether the total number of guidance times is lower than a preset third threshold: if so, the adjustment amount of the association evaluation value is zero; otherwise, calculate the dynamic adjustment coefficient according to the knowledge scenario type, and adjust the association evaluation value according to the dynamic adjustment coefficient to obtain the adjusted association evaluation value.

6. The method according to claim 5, characterized in that, Calculate dynamic adjustment coefficients based on the knowledge scenario type, including: Based on the knowledge scenario type, query the preset evaluation value adjustment coefficient table to obtain the basic adjustment coefficient; use the ratio of the historical successful guidance count to the total guidance count under the knowledge scenario type as the scenario adaptation factor; The dynamic adjustment coefficient is obtained by multiplying the basic adjustment coefficient by the scene adaptation factor.

7. The method according to claim 1, characterized in that, The knowledge mastery score for each historically related concept is calculated based on the aforementioned historical guidance records, including: The historical guidance record also includes the user's historical query frequency for each historical related concept, the recent query frequency change rate, and the average accuracy of related tests. The total number of historical queries, the rate of change in recent query frequency, and the average accuracy are standardized and their arithmetic mean is calculated to obtain the knowledge mastery score for each historical related concept.

8. The method according to claim 1, characterized in that, Based on the knowledge mastery score and the adjusted correlation evaluation value, calculate the comprehensive evaluation value for each historical correlation concept, including: The comprehensive evaluation value for each historical related concept is obtained by weighting the knowledge mastery score and the adjusted correlation evaluation value.

9. A knowledge guidance device for AI-powered point-and-read learning, characterized in that, include: Data acquisition module: It is used to acquire the target concept that the user clicks on in the current reading material, the context text of the target concept and the historical related concepts, as well as the historical guidance record of the historical related concepts. The historical guidance record includes the number of successful guidances and the total number of guidances. The association calculation module is used to calculate the semantic association degree between the historical associated concepts and the context text, calculate the historical utility index based on the historical guidance record, and input the semantic association degree and the historical utility index into a preset evaluation model to obtain the association evaluation value of each historical associated concept. Scene adjustment module: It is used to determine the knowledge scene type of this query based on the context text, and adjust the association evaluation value according to the knowledge scene type; Comprehensive evaluation module: It is used to calculate the knowledge mastery score of each historical related concept based on the historical guidance record, and to calculate the comprehensive evaluation value of each historical related concept based on the knowledge mastery score and the adjusted related evaluation value; Explanation generation module: It is used to determine the final historical association concept from the historical association concepts based on the comprehensive evaluation value, and generate personalized association explanations based on the final historical association concept, the target concept and the context text.

10. A readable storage medium, characterized in that, The readable storage medium stores computer program instructions, which are read and executed by a processor to perform the steps of the knowledge guidance method for AI-oriented point-reading as described in any one of claims 1-8.

Images