Electronic document reading management method and reading management device

Abstract

The present invention relates to the technical field of electronic document management and discloses a method for managing electronic document reading, comprising the following steps: S1, a document information entropy calculation step, which analyzes the text content of an electronic document based on an information entropy calculation method to determine the complexity of different content; S2, a user behavior data collection step, which collects user behavior data on scrolling, pausing, returning, highlighting, and note-taking while reading the electronic document. A device for managing electronic document reading is also provided, comprising: a document information entropy calculation module, which calculates the complexity of the electronic document text and sets a reading weight; and a user behavior monitoring module, which collects user reading behavior data. The present invention achieves more intelligent, efficient, and personalized reading management through reinforcement learning to optimize recommendation ranking, temperature parameter control to balance exploration and utilization, a time decay mechanism to dynamically adjust interest preferences, and interactive recommendations to enhance user control.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic document management, and in particular to an electronic document reading management method and a reading management device. Background Art

[0002] With the widespread adoption of electronic documents, the way people access and read information has dramatically changed. Whether for study, work, or everyday browsing, users encounter a vast amount of electronic documents daily. However, faced with this massive flow of information, quickly finding content that meets personal needs has become a major pain point for users. An efficient reading management system must not only provide accurate content recommendations but also dynamically adapt to changing user interests, improving reading efficiency and experience.

[0003] Existing electronic document reading management systems already offer a certain degree of content recommendation and personalized presentation. Some systems utilize recommendation algorithms based on historical clickthrough data, which can, to a certain extent, match user interests and improve the relevance of recommended content. Furthermore, some reading platforms support simple interactive adjustments, such as allowing users to manually mark content they are not interested in or using filtering functions to adjust the scope of recommendations. These methods have improved the reading experience to a certain extent, making it easier for users to manage and access information.

[0004] However, existing recommendation methods still have some shortcomings. First, the recommendation mechanisms of many systems are relatively fixed and fail to fully consider the dynamic changes in user interests, resulting in the recommended content being easily solidified and lacking freshness. In addition, some systems tend to reinforce existing interests during the recommendation process and lack reasonable exploration strategies, making it difficult for users to access new high-value information. On the other hand, the sorting method of recommended content is often based on static weights and fails to adjust in combination with real-time user feedback, resulting in insufficient recommendation accuracy. Furthermore, existing recommendation systems mostly adopt fixed display modes and lack flexible interaction methods, which makes it impossible for users to actively adjust recommended content, affecting the overall reading experience. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a reading management method and a reading management device for electronic documents, which solve the problems in the existing technology such as solidified recommended content, lack of dynamic adaptability, difficulty in balancing exploration and accuracy, and single interactive mode.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method for managing the reading of electronic documents, comprising the following steps:

[0007] S1, document information entropy calculation step, analyzing the text content of the electronic document based on the information entropy calculation method to determine the complexity of different contents;

[0008] S2, user behavior data collection step, collecting user behavior data on scrolling, staying, returning, highlighting, and taking notes when reading electronic documents;

[0009] S3, information gain calculation step, based on the information entropy calculation result and user behavior data, calculating the user's information gain for different contents of the electronic document, which is used to evaluate the user's reading depth;

[0010] S4, Bayesian reasoning progress calculation step, based on user behavior data and information gain calculation results, using Bayesian reasoning method to update the user's reading status and dynamically calculate the user's reading progress;

[0011] S5. Personalized optimization step: Based on the reading progress calculation results, adjust the personalized reading recommendation strategy and optimize the order of subsequent content recommendations to match the user's reading habits and understanding level.

[0012] Preferably, the document information entropy calculation step includes:

[0013] Parse the text content of electronic documents and calculate the information entropy value of each text paragraph based on the statistical model of the text content;

[0014] Based on the information entropy value, the complexity of each paragraph of text is determined, and different reading weights are assigned to content of different complexities; an entropy-based algorithm is used to mark important parts of the document for optimization in subsequent user behavior analysis.

[0015] Preferably, the information gain calculation step includes:

[0016] Based on the complexity analysis results in the document information entropy calculation step, combined with the dwell time, scrolling behavior, and highlighting in the user behavior data collection step, the user's information gain for each content paragraph in the document is calculated;

[0017] Based on the information gain value, the user's depth of understanding of different content is evaluated, and the user's reading progress is dynamically adjusted based on the result; for content with high information gain, more detailed follow-up content is recommended, while for content with low information gain, relatively brief or summary content can be recommended.

[0018] Preferably, the information gain calculation step includes:

[0019] Calculate the initial information entropy of the document;

[0020] Calculate the change in text information entropy after the user reads;

[0021] Use information gain calculation method to measure user reading depth;

[0022] According to the information gain results, adjust the contribution weight of the page to the reading progress.

[0023] Preferably, the Bayesian reasoning progress calculation step includes:

[0024] Set the user's reading prior probability;

[0025] Calculate the relationship between the user's reading behavior pattern on a certain page and the degree of reading completion;

[0026] Use Bayesian inference to calculate the probability that the user actually reads the page;

[0027] The reading progress is dynamically adjusted based on the reading probability of different pages, and the reading progress calculation results are used to dynamically adjust subsequent content recommendations.

[0028] Preferably, the personalized optimization step includes:

[0029] Adjust reading time allocation strategy based on user reading behavior patterns;

[0030] Predict the best learning path for the user's unread parts and adjust the recommendation order. The recommended reading path is dynamically adjusted based on the user's reading behavior;

[0031] Generate personalized reading reports for users and provide review and reinforcement learning suggestions.

[0032] The present invention also provides an electronic document reading management device, comprising:

[0033] Document information entropy calculation module, used to calculate the complexity of electronic document text and set reading weight;

[0034] User behavior monitoring module, used to collect user reading behavior data, including scrolling, staying, returning, highlighting and note-taking behavior data;

[0035] The information gain calculation module is used to calculate the depth of understanding of document content based on user interaction behavior and adjust the reading progress weight; the Bayesian reasoning progress calculation module is used to dynamically update the reading progress based on user behavior data;

[0036] The personalized optimization module is used to adjust the reading recommendation strategy based on the user's historical reading behavior.

[0037] Preferably, the document information entropy calculation module includes:

[0038] A text parsing unit for analyzing the text structure of electronic documents;

[0039] The word frequency statistics unit is used to count the occurrence probability of words in the document;

[0040] Information entropy calculation unit, used to calculate the complexity of the text and set the reading time weight.

[0041] Preferably, the user behavior monitoring module includes:

[0042] Time monitoring unit, used to record the user's stay time on the page;

[0043] Scroll monitoring unit, used to monitor the user's scrolling behavior and page turning speed;

[0044] The interaction monitoring unit is used to record the user's highlighting, annotation and note-taking behavior data.

[0045] Preferably, the Bayesian reasoning progress calculation module includes:

[0046] Behavior probability calculation unit, used to calculate the relationship between user reading behavior and reading completion;

[0047] Bayesian reasoning unit, used to calculate the user's true reading probability using the Bayesian method;

[0048] The progress dynamic adjustment unit is used to dynamically calculate the reading progress based on the reading probability and optimize the learning path.

[0049] The present invention provides a method and device for managing the reading of electronic documents, which have the following beneficial effects:

[0050] 1. This invention uses reinforcement learning to optimize recommendation ranking and introduces soft maximization to control the probability distribution of recommended content. Compared to traditional fixed ranking methods, this strategy can adapt to changes in user interests in real time, making recommendation results more personalized and avoiding the problem of content uniformity.

[0051] 2. By introducing a temperature parameter, this invention intelligently adjusts the balance between exploration and exploitation. The system is neither limited to existing preferences nor blindly recommends low-relevance content. Compared to traditional recommendations based solely on historical behavior, this method expands user interests while maintaining the accuracy of recommended content.

[0052] 3. This invention incorporates a time decay factor, giving recent user behavior a greater influence on recommendation ranking. This prevents the system from relying on outdated data for extended periods of time and allows it to dynamically respond to recent changes in user preferences. Compared to existing recommendation systems, which are prone to "fixing" user interests, this solution ensures the timeliness and flexibility of recommended content.

[0053] 4. This invention utilizes interactive recommendation strategies, such as sliding, folding, and user-defined adjustments, to provide a more flexible presentation of recommended content. Compared to traditional fixed card-style presentations, this invention can adapt to different user habits, improve the readability of recommended information, reduce information overload, and enhance the user's reading experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 is a flow chart of the method of the present invention;

[0055] Figure 2 It is a structural schematic diagram of the present invention. DETAILED DESCRIPTION

[0056] The following is a description of the present invention Figure 1-2 The technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0057] Please see the attached Figure 1-2 , an embodiment of the present invention provides a method for managing the reading of an electronic document, comprising the following steps:

[0058] S1, document information entropy calculation step, analyzing the text content of the electronic document based on the information entropy calculation method to determine the complexity of different contents;

[0059] The primary purpose of calculating document entropy is to quantify the complexity of each section within an electronic document, providing fundamental data for subsequent reading behavior analysis, information gain calculation, and personalized optimization. By calculating document entropy, we can measure the difficulty of different content and adjust reading weights accordingly, enabling reading management systems to rationally allocate users' reading time and optimize subsequent recommendation strategies.

[0060] Generally, the text content of electronic documents contains multiple layers of information, each with varying degrees of complexity, leading to varying cognitive loads when reading. Therefore, as an option, we can quantify text complexity based on information entropy theory and combine this with user behavioral data to more accurately assess user reading depth and comprehension.

[0061] Specifically, this step completes the complexity analysis of the electronic document through four sub-processes: text parsing, word frequency statistics, information entropy calculation, and reading weight assignment. In some embodiments, natural language processing (NLP) techniques such as topic modeling (LDA), word vector representation (Word2Vec), or semantic similarity analysis (BERT) can be combined to improve the accuracy of text complexity assessment.

[0062] In this embodiment, the electronic document is first structured and parsed to extract the text content, and then basic cleaning and preprocessing are performed. Generally, text preprocessing includes:

[0063] Remove stop words (such as words with no actual semantic contribution like "de", "shi", "zai", etc.);

[0064] Perform word segmentation to split the text into independent words or phrases;

[0065] Perform词性 tagging to identify different categories of words such as verbs, nouns, adjectives, etc.;

[0066] Perform syntactic analysis to identify the subject-predicate-object structure of sentences for better understanding of the text content hierarchy;

[0067] In a possible implementation, the text parsing unit can use the TF-IDF (Term Frequency-Inverse Document Frequency) method to measure the importance of words, or use deep learning semantic vector representations such as Word2Vec or BERT to extract richer semantic information.

[0068] In some embodiments, the complexity of the document content can be preliminarily analyzed through word frequency statistics. Generally, complex texts tend to contain more low-frequency words and long sentences, while simpler texts are usually composed of high-frequency words and short sentences. Therefore, the n-gram statistical method can be used to calculate the word frequency distribution in the text and perform modeling analysis on it. \

[0069] [[ID=l8]]As an option, the word frequency histogram of each paragraph can be calculated, and further the entropy value of the text can be calculated. In certain implementations, the system can also analyze the sentence length distribution of the text and the complexity of the subject-predicate-object structure to further evaluate the reading difficulty of the text.

[0070] In this embodiment, the calculation of information entropy is based on the following mathematical formula:

[0071]

[0072] Where: H(X) represents the information entropy value of the document or paragraph, reflecting the text complexity; P(x i ) represents the occurrence probability of the i-th word in the text; N is the total number of different words in the text; x i is an individual word in the document.

[0073] The core idea of information entropy calculation is: when the word frequency distribution of a certain text is relatively uniform, the information entropy is higher, meaning that the information contained in the text is more dispersed and the complexity is higher; conversely, if certain words occupy a large proportion (such as a large number of repeated high-frequency words), the information entropy is lower, indicating that the complexity of the text is lower.

[0074] In a possible implementation, the information entropy can be calculated according to the following steps:

[0075] Count the number of occurrences of each word in the document and calculate its relative frequency P(xi ), use the above entropy formula to calculate the total information entropy of the entire document.

[0076] The document is divided into multiple paragraphs, and the information entropy of each paragraph is calculated to obtain the complexity distribution of different contents.

[0077] In this embodiment, based on the calculation results of information entropy, the system assigns different reading weights to each text paragraph. Generally speaking, paragraphs with higher information entropy contain more new information, and users may need longer to read them. Therefore, the system can use the following formula to calculate the reading time weight:

[0078]

[0079] Where: W(T) is the reading time weight of the current paragraph; H(X) is the information entropy of the current paragraph; M is the total number of paragraphs in the document; H(X) j ) is the information entropy of the j-th paragraph; It represents the sum of the information entropy values ​​of all paragraphs in the entire document. It is the normalized denominator part and is used to ensure that the sum of the weights of each paragraph is equal to 1.

[0080] The purpose of this formula is to ensure that content with higher information entropy occupies a larger proportion of reading time, while content with lower information entropy occupies a relatively smaller proportion. As an option, this weight can be dynamically adjusted during the reading process to further optimize based on the user's reading behavior.

[0081] In some embodiments, syntactic tree parsing can be further combined to analyze the structural complexity of the text. For example, the dependency tree depth of a sentence can be calculated to measure the grammatical complexity of the sentence. Furthermore, TF-IDF and cosine similarity can be combined to analyze the degree of information redundancy between different paragraphs, thereby optimizing the reading time allocation strategy.

[0082] As another possible implementation method, a machine learning model (such as a random forest or deep learning model) can be introduced. By training the features of a large amount of text, the system can more accurately predict the complexity of the text and optimize the accuracy of information entropy calculation.

[0083] The document entropy calculation process quantitatively assesses the complexity of electronic documents through four key sub-processes: text parsing, word frequency statistics, entropy calculation, and reading weight assignment. Entropy calculation is based on a probability distribution formula and combined with a reading time weighting method to ensure the system can rationally allocate users' reading time and improve the intelligence of reading management. Further optimization solutions can incorporate natural language processing techniques such as syntactic structure analysis, topic modeling, and TF-IDF to improve calculation accuracy and applicability.

[0084] S2, user behavior data collection step, collecting user behavior data on scrolling, staying, returning, highlighting, and taking notes when reading electronic documents;

[0085] The user behavior monitoring step is a key step in the system's acquisition of user reading interaction data. The main goal of this step is to collect user reading behavior information in real time and use this data to construct reading patterns to optimize subsequent information gain calculations and personalized recommendation strategies.

[0086] Generally speaking, users' behavioral characteristics when reading electronic documents are diverse, mainly including: page dwell time, scrolling operations, interactive behaviors (clicking, annotating, taking notes, etc.), rereading and skipping, etc. These behaviors not only reflect the user's interest and preferences in the text content, but also provide indirect information on the depth of their understanding of the content. Therefore, in this step, the system uses multi-dimensional monitoring methods (such as mouse trajectory, touch interaction, keyboard input, etc.) to establish a user reading behavior dataset to achieve accurate user behavior modeling.

[0087] Specifically, the core of this step includes four sub-processes: time monitoring, scrolling monitoring, interaction monitoring, and reading interest inference. Each sub-process is responsible for capturing different dimensions of user reading behavior data. In some embodiments, combined with machine learning methods (such as long short-term memory networks (LSTMs) and hidden Markov models (HMMs), the system can predict user reading patterns and adjust personalized recommendation strategies in real time to optimize the user experience.

[0088] In this embodiment, the time monitoring unit is used to record the length of time a user spends on each page or paragraph and analyze the user's reading pattern based on this data. Generally speaking, a longer stay time usually indicates that the user is interested in the content or that the content is difficult to understand, while a short browsing time may mean that the user is not interested in the content or is already familiar with it.

[0089] As an option, the following formula can be used to calculate the reading time weight of a unit paragraph:

[0090]

[0091] Where: T w is the reading time weight of the current paragraph; T p Indicates the time the user stays in the current paragraph; T t The total reading time of the user in the entire document.

[0092] The function of this formula is to measure the time allocation of users to different paragraphs, thereby inferring the focus of users. In some embodiments, a sliding window averaging method can be combined to smooth the time data to reduce the impact of outliers on the results.

[0093] Furthermore, time monitoring can be combined with analysis of reading cadence changes, further identifying fluctuations in user interest by calculating the rate of change in time between paragraphs. For example, if the dwell time on a particular paragraph is significantly higher than the preceding and following paragraphs, it may contain key content, and the system can adjust subsequent recommendation strategies accordingly.

[0094] In this embodiment, the scrolling monitoring unit is used to capture the user's scrolling behavior characteristics, including scrolling speed, direction, pause location, etc. Scrolling behavior can reflect the user's browsing pattern of the document content: fast scrolling usually indicates skipping or browsing, while multiple pauses or scrolling back may indicate that a certain part of the content is more complex or the user is paying close attention to it.

[0095] In general, the scrolling speed can be calculated by the following formula:

[0096]

[0097] Where: S r is the scrolling speed (unit: pixels / second or lines / second); D s Indicates the scrolling distance (pixels or lines); T s The time (in seconds) to scroll.

[0098] In some embodiments, the system can calculate scrolling acceleration, that is, the rate of change of scrolling speed over time, to analyze the user's reading rhythm. For example, areas with large changes in scrolling speed may be where the user is quickly looking for information, while lower scrolling acceleration generally means that the user is reading carefully.

[0099] In addition, to detect whether the user has replayed a certain content, the scrolling direction can be recorded and the replay frequency can be calculated. The more times the replay operation is performed, the more complex the content is or the more likely the user is interested in the content.

[0100] In this embodiment, the interaction monitoring unit is used to record the user's active interaction during the reading process, including: marking, highlighting, taking notes, clicking, etc. Generally speaking, when a user interacts with a certain content, it usually means that the content is of high value to the user.

[0101] In one possible implementation, an interaction behavior weighting model can be defined to calculate weighted values ​​for different types of interaction behaviors:

[0102] I w =w1×N h +w2×N n +w3×N c ;

[0103] Where: I w is the weighted value of the interaction behavior; N hNumber of times highlighted for the user; N n Number of times a note is added for the user; N c is the number of times a user clicks on a certain part of the content; w1, w2, and w3 are weight coefficients for different interactive behaviors, which can be adjusted based on experimental data.

[0104] Generally speaking, highlights and notes are weighted higher, while ordinary clicks are weighted lower, because the former usually means that the user is thinking deeply about a certain content. In some embodiments, the TF-IDF method can be combined to analyze the keywords of user interactions to further infer their reading interests.

[0105] In this embodiment, the system infers the user's reading interests and cognitive state based on time monitoring, scrolling monitoring, and interaction monitoring data. Alternatively, a Bayesian classifier or Hidden Markov Model (HMM) can be used to predict the user's future reading needs based on their historical behavior and dynamically adjust the order of content recommendations.

[0106] Specifically, in one possible implementation, the system can calculate the distribution of users' reading interests in different categories of content:

[0107]

[0108] Where: P(C i |X) represents the user's preference for category C i The probability of interest; P(X|C i ) indicates that in category C i The probability that a user performs certain actions under the given conditions; P(C i ) is the prior probability of the category; P(X) is the probability distribution of the user's overall behavior.

[0109] Through this formula, we can estimate users' preferences for content on different topics, thereby optimizing the recommendation algorithm and improving the accuracy of personalized reading experience.

[0110] In some embodiments, eye tracking technology can be combined to analyze the user's gaze movement trajectory to obtain more accurate reading attention data. In addition, sentiment analysis methods can be used to analyze the user's emotional fluctuations during reading, such as through facial expression recognition or voice emotion analysis, to determine the user's reaction to specific content and thus adjust the recommendation strategy.

[0111] As another possible implementation method, reinforcement learning algorithms can be used to adjust the recommendation model in real time, so that the system can be optimized based on users' immediate feedback and gradually improve the recommendation effect.

[0112] The user behavior monitoring process comprehensively collects and analyzes user reading behaviors through four key sub-processes: time monitoring, scrolling monitoring, interaction monitoring, and reading interest inference. Based on this data, the system can accurately infer user reading patterns and, combined with machine learning algorithms, optimize personalized recommendation strategies, making the reading management system more intelligent and efficient.

[0113] S3, information gain calculation step, based on the information entropy calculation result and user behavior data, calculating the user's information gain for different contents of the electronic document, which is used to evaluate the user's reading depth;

[0114] The user behavior analysis and recommendation strategy adjustment step (S3) builds on the user behavior monitoring in step S2 by further mining user behavior data to identify users' reading interests, content preferences, and interaction patterns, and dynamically optimizes the recommendation strategy accordingly. The core goal of this step is to improve the matching of recommended content, enabling users to more quickly access documents that meet their needs, thereby enhancing their reading experience.

[0115] In this step, the system conducts a multi-dimensional analysis of various user behavior data within the document, including reading time distribution, scrolling behavior patterns, interaction frequency, and scrollback operations. Furthermore, the system combines machine learning and statistical inference techniques, such as Bayesian classification, hidden Markov models (HMMs), and reinforcement learning, to dynamically adjust recommendation strategies to ensure that recommended content remains highly consistent with user interests.

[0116] In this embodiment, the interest pattern analysis unit processes the user behavior data collected in step S2 and constructs a user reading interest model through feature extraction, classification, and pattern recognition. The system first extracts the user's behavior characteristics in different content categories and represents them as a behavior feature vector:

[0117] X=(x1,x2,...,x n );

[0118] Where: x1, x2, ..., x n Represents behavioral data of different dimensions, such as the dwell time, number of clicks, scrolling rate, etc. of a certain content.

[0119] Generally, a user’s interest in a certain type of content can be calculated through a Bayesian classifier to calculate its interest probability, using the “reading interest distribution” disclosed by S2.

[0120] In one possible implementation, the system continuously monitors changes in user interests and updates the user interest model based on new behavioral data. For example, if a user frequently scrolls back to a certain type of content over a period of time, the system can adjust the interest weight of that category, giving it a higher priority in subsequent recommendations.

[0121] In addition, in some embodiments, user behavior patterns can be classified through clustering algorithms (such as K-means) to identify different types of reading habits, such as fast browsing, in-depth reading, and skipping reading, thereby optimizing personalized recommendation strategies.

[0122] Based on the interest pattern analysis, in this embodiment, the personalized recommendation strategy adjustment unit is responsible for adjusting the recommended content according to the dynamic changes in user interests. Generally speaking, the recommendation strategy adjustment mainly includes operations such as content screening, sequence optimization, and dynamic update.

[0123] Specifically, the system first uses the interest model to filter the content in the document library and remove content that the user may not be interested in; then, based on the user's interest weights in different content categories, the recommended content is prioritized to ensure that the most relevant content is presented first; finally, the system dynamically updates the recommendation list based on real-time behavioral data to match the user's immediate needs.

[0124] In some embodiments, the system can combine the Hidden Markov Model (HMM) to predict the user's next reading behavior and make recommendations in advance. The calculation method is as follows:

[0125]

[0126] Where: P(X t+1 |X t ) represents the user's current behavior X t After that, the next possible reading behavior; P(X t+1 |S i ) represents the state S i Next, the user performs behavior X t+1 The probability of P(S i |X t ) is in the current behavior X t Next, the user is in state S i The probability of; N1 is the number of all possible reading states.

[0127] This model can be used to predict users' stay time and interaction methods on different content, so as to adjust the recommended content in advance and improve the accuracy of recommendations.

[0128] In this embodiment, the recommendation algorithm optimization unit is responsible for continuously optimizing the recommendation strategy to ensure that the recommendation system can continuously adjust the content recommendation weight based on user feedback. In one possible implementation, the system uses a reinforcement learning algorithm to continuously update the recommendation strategy based on real-time user feedback.

[0129] Reinforcement learning can be achieved through the Q-learning method:

[0130]

[0131] Where: Q(s t ,a t ) represents the state s t Next, perform action a t The Q value is the value of the recommended content; α is the learning rate, which determines the impact of new data on the model; r t is the user's immediate feedback on the recommended content (such as clicks, favorites, etc.); γ is the discount factor, which is used to measure the impact on future revenue; is in the next state s t+1 Next, take the optimal action a t+1 The maximum Q value.

[0132] In some embodiments, reinforcement learning can be combined with a multi-armed bandit algorithm to find the optimal balance between exploring new content and leveraging existing interest data. For example, in some cases, the system can recommend content that the user has not read before to test their interest and thus optimize long-term recommendation effectiveness.

[0133] Another possible implementation approach is to combine natural language processing (NLP) technology to analyze the notes and highlights that users add during reading to identify their key points of interest. For example, using the TF-IDF algorithm or topic modeling (LDA), we can identify the keywords that users are most interested in within a document and adjust recommended content accordingly.

[0134] Furthermore, in some embodiments, the system can incorporate sentiment analysis, utilizing facial expression recognition or voice emotion analysis, to identify users' emotional responses to different content. For example, if a user displays high concentration or positive emotions on certain content, the system can further optimize the recommendation weight of the relevant content.

[0135] The user behavior analysis and recommendation strategy adjustment steps involve deeply analyzing the behavioral data collected by S2 to establish a user interest model and optimize personalized recommendation strategies based on this model. The system combines Bayesian classification, hidden Markov models (HMMs), reinforcement learning, and other technologies to predict user reading interests and dynamically adjust recommended content. Furthermore, by combining natural language processing (NLP) technology and sentiment analysis, the system can further enhance the accuracy of recommendation strategies to optimize the user's overall reading experience.

[0136] S4, Bayesian reasoning progress calculation step, based on user behavior data and information gain calculation results, using Bayesian reasoning method to update the user's reading status and dynamically calculate the user's reading progress;

[0137] Step S4 of user feedback processing and recommendation strategy adaptive optimization is based on the user behavior analysis and recommendation strategy adjustment in step S3, and further optimizes the recommendation strategy based on user feedback information.

[0138] Generally speaking, user feedback can be divided into two categories: explicit feedback and implicit feedback. Explicit feedback includes interactions that clearly express user preferences, such as likes, favorites, and comments, while implicit feedback includes behavioral characteristics that do not require active user interaction, such as reading time, scrolling speed, click frequency, and scroll back times.

[0139] As an option, this step processes user feedback data to calculate the quality score of the recommended content, and combines reinforcement learning, adaptive optimization and other methods to enable dynamic adjustment of the recommendation strategy to improve the personalized accuracy of the recommended content.

[0140] Specifically, the core technologies of this step include:

[0141] Collect and classify user feedback data, and summarize and organize explicit and implicit feedback;

[0142] Recommendation effect evaluation, using a weighted scoring function to calculate the comprehensive score of the recommended content;

[0143] Reinforcement learning optimization, adjust recommendation weights based on user feedback data, and optimize strategies;

[0144] Recommendation diversity control, using the multi-armed bandit (MAB) method to prevent over-concentration of recommended content;

[0145] Adaptive parameter adjustment, combined with policy gradient optimization, enables the recommendation algorithm to adapt to changes in user interests.

[0146] In this embodiment, the user feedback data collection unit is responsible for classifying and storing various types of interaction data generated by users during the electronic document reading process to form a complete feedback data set.

[0147] In general, user feedback can be represented as a feature vector:

[0148] F=(f1,f2,...,f7);

[0149] Among them: f1 is the number of likes, the number of times the user likes the content; f2 is the number of collections, the number of times the user bookmarks the document; f3 is the number of comments, the number of times the user comments; f4 is the dwell time, the time the user spends reading the content (unit: seconds); f5 is the scrolling speed, the speed at which the user scrolls the page (unit: pixels / second); f6 is the number of rollbacks, the number of times the user returns to the previously read content; f7 is the bounce rate, the probability that the user leaves without reading in depth; F is the original recommendation ranking factor, which represents the weight that has not been adjusted by time decay.

[0150] In one possible implementation, the system can use principal component analysis (PCA) to reduce the dimensionality of the feedback data to identify the features that have the greatest impact on optimizing the recommendation strategy. For example, if a user's dwell time and number of rollbacks show a strong correlation in historical data, the system can reduce some redundant features to improve computational efficiency.

[0151] In some embodiments, the time-effectiveness weight of the feedback data may be calculated in combination with a time decay factor to reduce the impact of outdated data:

[0152] F ′ =e -λt F;

[0153] Among them: F ′ is the feedback feature vector after time decay processing; λ is the time decay factor, which indicates the degree of data decay over time; t is the timestamp of the feedback data, which is the difference from the current time; e is the base of the natural logarithm; F is the original recommendation ranking factor, which indicates the weight that has not been adjusted by time decay.

[0154] In this embodiment, the recommendation effect evaluation unit quantifies the quality of the recommended content to determine whether the current recommendation strategy needs to be adjusted.

[0155] Alternatively, the recommendation effect can be calculated using a weighted scoring function as follows:

[0156]

[0157] Where: S is the comprehensive score of the recommended content; f i is the value of the i-th feedback feature; w i is the weight of the feedback feature, which indicates the influence of the feature on the optimization of the recommendation strategy.

[0158] In some embodiments, the weight w i Adaptive adjustments can be made based on historical user interaction data. For example, if the system finds that users spend a long time on a certain type of content and leave many comments, it will automatically increase the weights corresponding to the dwell time and number of comments to enhance the influence of this feature.

[0159] In addition, in one possible implementation, the system may use regression analysis to calculate the contribution of different feedback features to the final score, ensuring that the optimized weight distribution is more reasonable.

[0160] In this embodiment, the recommendation strategy optimization unit combines the reinforcement learning method to continuously adjust the parameters of the recommendation strategy according to user feedback data.

[0161] As an option, the Q-learning method can be used to update the quality score of the recommended content (the specific formula is given in step S3 and will not be repeated here).

[0162] In another implementation, the policy gradient method can be combined to make the recommendation strategy more intelligent by optimizing the policy function π(a|s). The goal of policy gradient optimization is to find the policy parameter θ that can maximize long-term user satisfaction:

[0163]

[0164] Where: J(θ) is the target optimization function, which measures the overall performance of the recommendation strategy; is the policy gradient; θ is the parameter vector of the recommended strategy; π θ (a|s) represents the probability of selecting recommended action a in state s; Q(s,a) is the Q value of the recommended content, that is, the predicted user satisfaction; is the mathematical expectation, which is used to calculate the weighted average gradient under all possible strategies; is the logarithmic policy gradient, which represents the logarithm of the policy probability of taking action a in a given state s, with respect to the parameter θ.

[0165] In this embodiment, to avoid excessive concentration of recommended content in certain specific categories, the system combines the multi-armed bandit (MAB) method to strike a balance between exploring new content and utilizing existing data.

[0166] As an option, the system may adopt a UCB (upper confidence bound) strategy to calculate the priority of the recommended content (the formula is given in step S3 and will not be repeated here).

[0167] In some embodiments, the range of the exploration factor c can be dynamically adjusted to match the changing interests of different users. For example, if the system detects that a user's recent interests are relatively stable, the exploration factor can be lowered to reduce the frequency of new content recommendations. Conversely, if the user's behavior patterns vary significantly, the exploration factor can be appropriately increased to increase the proportion of new content recommendations.

[0168] In addition, in another possible implementation, a temperature softening (Softmax) strategy can be used to adjust the recommendation probability:

[0169] Where: P(a i ) is to select recommended content a i The probability of Q(a i ) is the quality score of the content; j is the index variable in the recommendation candidate set, which is used to traverse all the recommended content a jThe normalized probability distribution is used; τ is the temperature parameter, which controls the degree of exploration. The larger the value, the more random the recommendation strategy.

[0170] User feedback processing and adaptive optimization of recommendation strategies (S4) achieves dynamic optimization of recommendation strategies through the collection of user feedback data, evaluation of recommendation effects, reinforcement learning optimization, and diversity control.

[0171] The system combines Q-learning, policy gradient, UCB, Softmax and other algorithms to continuously adjust the recommendation strategy so that it can be adaptively optimized according to changes in user interests, thereby improving the personalized experience of electronic document reading.

[0172] S5, personalized optimization step, based on the reading progress calculation results, adjusts the personalized reading recommendation strategy and optimizes the order of subsequent content recommendations to match the user's reading habits and understanding level;

[0173] Step S5 of optimizing the display of recommendation results and user interaction is based on the user feedback processing and adaptive optimization of recommendation strategies in step S4, and further optimizes the display of recommended content and enhances the user interaction experience.

[0174] Generally speaking, the presentation of recommended content has a direct impact on user reading behavior and feedback. Therefore, in this step, the system not only considers the accuracy of the recommended content, but also focuses on the visualization of the recommended information, content sorting, dynamic adjustment mechanisms, and personalized interaction optimization, thereby improving user reading interest and satisfaction.

[0175] As an option, this step combines methods such as interface layout optimization, interactive recommendation, reinforcement learning regulation, and personalized information annotation to make the display of recommended content more intelligent and meet the reading preferences of different users.

[0176] Specifically, the core technologies of this step include:

[0177] Optimize the ranking of recommended content and adjust the display order based on user interest weight;

[0178] Personalized recommendation display, dynamically adjusting content presentation based on user historical data;

[0179] Interactive recommendation mechanism to improve user experience through sliding, clicking, folding, etc.

[0180] Dynamic interface adaptation, adaptively adjusting the recommendation module according to device type and screen size;

[0181] Information annotation and visualization enhancement improve information comprehension through labels, charts, etc.

[0182] In this embodiment, the recommendation result display unit optimizes the order of the content by calculating the personalized weight score of the recommended content, so as to ensure that the content that best meets the user's interests can be displayed first.

[0183] Generally, the ranking of recommended content is affected by multiple factors, including historical user interaction data, features of current recommended content, real-time changes in user feedback, etc. Therefore, in this embodiment, the ranking score of recommended content can be calculated using a weighted scoring function.

[0184] As an option, the aforementioned weighted scoring function can be used to calculate the “weighted scoring function calculation”:

[0185] In addition, in another possible implementation, a collaborative filtering (CF) method may be combined to calculate the ranking scores of recommended content based on the preferences of similar users.

[0186] In some embodiments, a time decay factor may be incorporated to make the most recent user behavior have a greater impact on the ranking score. The time decay factor is calculated in the same manner as in step S4 and the formula is not repeated, but it still applies to this embodiment.

[0187] In this embodiment, in order to improve the user experience, the system can dynamically adjust the display method of recommended content based on the user's reading habits.

[0188] As an option, recommended content can be displayed in waterfall or pagination mode, and the scrolling method can be adjusted according to user behavior. For example:

[0189] If the user is accustomed to fast browsing, the recommendation list adopts continuous loading mode to improve content fluency;

[0190] If users are more interested in specific categories, partition recommendations can be used to display content of different categories in different modules.

[0191] In some embodiments, recommended content can be categorized based on the user's areas of interest and differentiated using different card styles. For example:

[0192] News content can be displayed in the form of title + summary;

[0193] Technical documents can be displayed in a directory format to facilitate quick location of key information;

[0194] Multimedia content can be presented in the form of video preview + text introduction to improve readability.

[0195] In this embodiment, in order to enhance the user's active interaction ability, the recommendation system provides a series of interactive recommendation methods, including:

[0196] Slide to select recommendations: Users can adjust the recommended content by sliding left and right, similar to the "card-style recommendation" mode;

[0197] Foldable recommendations: Users can expand or collapse certain types of content to reduce information overload;

[0198] User-defined recommendations: Users can adjust recommendation parameters, such as selecting preferred content categories, recommendation frequency, etc.

[0199] In one possible implementation, the system could incorporate reinforcement learning (RL) to continuously optimize interaction methods based on user behavior. For example, if the system detects that users tend to scroll through content, it would increase the proportion of "swiping recommendations" and reduce the use of "paged loading."

[0200] In this embodiment, in order to adapt to the reading needs of different devices, the recommendation system can automatically adjust the interface layout.

[0201] As an option, the system can dynamically adjust the display of recommended content based on screen size (ScreenSize) and resolution (Resolution). For example:

[0202] On the PC side, recommended content can use a grid layout to increase information density;

[0203] On mobile phones, recommended content adopts a single-column flow layout to improve reading continuity.

[0204] In some embodiments, the system can combine gesture interactions to optimize the recommendation experience, such as:

[0205] Pair-zoom can be used to adjust the font size of recommended content;

[0206] Long press can be used to mark important content and trigger in-depth recommendations.

[0207] In this embodiment, in order to improve the readability of the recommended content, the system may add information tagging and visualization elements to the recommended content.

[0208] Recommendations may optionally include:

[0209] Keyword highlighting, such as highlighting terms that users are interested in;

[0210] Hotness indicators use icons to indicate the popularity of content;

[0211] Reason for recommendation: Display the “Recommendation Reason” below the recommended content, such as “Because you follow XX topic”.

[0212] In some embodiments, the system can combine data visualization technology to present the statistical information of recommended content in the form of charts. For example:

[0213] The bar chart shows the reading popularity trend of the article;

[0214] The word cloud diagram intuitively presents the user's keywords of interest.

[0215] This embodiment achieves intelligent display of recommended content and optimization of user interaction through technologies such as recommended content sorting optimization, personalized display, interactive recommendation, interface adaptive adjustment, and information visualization.

[0216] The system combines reinforcement learning, collaborative filtering, and interface adaptation algorithms to continuously adjust the recommendation method so that it can be dynamically optimized according to the user's reading habits, thereby improving the user's personalized experience and the effectiveness of content recommendations.

[0217] The electronic document reading management apparatus described below and the electronic document reading management method described above may refer to each other.

[0218] Please see the attached Figure 2 The present invention also provides an electronic document reading management device, comprising:

[0219] Document information entropy calculation module, used to calculate the complexity of electronic document text and set reading weight;

[0220] User behavior monitoring module, used to collect user reading behavior data, including scrolling, staying, returning, highlighting and note-taking behavior data;

[0221] The information gain calculation module is used to calculate the depth of understanding of document content based on user interaction behavior and adjust the reading progress weight; the Bayesian reasoning progress calculation module is used to dynamically update the reading progress based on user behavior data;

[0222] The personalized optimization module is used to adjust the reading recommendation strategy based on the user's historical reading behavior.

[0223] The device of this embodiment can be used to execute the above method embodiment, and its principles and technical effects are similar, so they will not be repeated here.

[0224] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for managing the reading of electronic documents, characterized in that: The following steps are involved: S1, document information entropy calculation step, analyzing the text content of the electronic document based on the information entropy calculation method to determine the complexity of different contents; S2, user behavior data collection step, collecting user behavior data on scrolling, staying, returning, highlighting, and taking notes when reading electronic documents; S3, information gain calculation step, based on the information entropy calculation result and user behavior data, calculating the user's information gain for different contents of the electronic document, which is used to evaluate the user's reading depth; S4, Bayesian reasoning progress calculation step, based on user behavior data and information gain calculation results, using Bayesian reasoning method to update the user's reading status and dynamically calculate the user's reading progress; S5, personalized optimization step, based on the reading progress calculation results, adjusts the personalized reading recommendation strategy and optimizes the order of subsequent content recommendations to match the user's reading habits and understanding level; The document information entropy calculation step includes: Parse the text content of electronic documents and calculate the information entropy value of each text paragraph based on the statistical model of the text content; According to the information entropy value, the complexity of each text is determined, and different reading weights are assigned to content of different complexity; Use an entropy-based algorithm to mark important parts of the document for optimization in subsequent user behavior analysis; The Bayesian reasoning progress calculation step includes: Set the user's reading prior probability; Calculate the relationship between the user's reading behavior pattern on a certain page and the degree of reading completion; Use Bayesian inference to calculate the probability that a user actually reads a page; The reading progress is dynamically adjusted based on the reading probability of different pages, and the reading progress calculation results are used to dynamically adjust subsequent content recommendations.

2. The electronic document reading management method according to claim 1, characterized in that: The information gain calculation step comprises: Based on the complexity analysis results in the document information entropy calculation step, combined with the dwell time, scrolling behavior, and highlighting in the user behavior data collection step, the user's information gain for each content paragraph in the document is calculated; Evaluate the user's understanding of different content based on the information gain value and dynamically adjust the user's reading progress based on the results; For content with high information gain, more detailed follow-up content is recommended, while for content with low information gain, relatively brief or summary content is recommended.

3. The electronic document reading management method according to claim 1, characterized in that: The information gain calculation step comprises: Calculate the initial information entropy of the document; Calculate the change in text information entropy after the user reads; Use information gain calculation method to measure user reading depth; According to the information gain results, adjust the contribution weight of the page to the reading progress.

4. The electronic document reading management method according to claim 1, characterized in that: The personalized optimization step includes: Adjust reading time allocation strategy based on user reading behavior patterns; Predict the best learning path for the user's unread parts and adjust the recommendation order. The recommended reading path is dynamically adjusted based on the user's reading behavior; Generate personalized reading reports for users and provide review and reinforcement learning suggestions.

5. An electronic document reading management device, characterized in that: The method for managing the reading of an electronic document according to any one of claims 1 to 4 comprises: Document information entropy calculation module, used to calculate the complexity of electronic document text and set reading weight; User behavior monitoring module, used to collect user reading behavior data, including scrolling, staying, returning, highlighting and note-taking behavior data; The information gain calculation module is used to calculate the depth of understanding of document content based on user interaction behavior and adjust the reading progress weight; Bayesian reasoning progress calculation module, used to dynamically update reading progress based on user behavior data; Personalized optimization module, used to adjust reading recommendation strategies based on user historical reading behavior; The document information entropy calculation module includes: A text parsing unit for analyzing the text structure of electronic documents; The word frequency statistics unit is used to count the occurrence probability of words in the document; Information entropy calculation unit, used to calculate the complexity of the text and set the reading time weight; The Bayesian reasoning progress calculation module includes: Behavior probability calculation unit, used to calculate the relationship between user reading behavior and reading completion; Bayesian reasoning unit, used to calculate the user's true reading probability using the Bayesian method; The progress dynamic adjustment unit is used to dynamically calculate the reading progress based on the reading probability and optimize the learning path.

6. The electronic document reading management device according to claim 5, characterized in that: The user behavior monitoring module includes: Time monitoring unit, used to record the user's stay time on the page; Scroll monitoring unit, used to monitor the user's scrolling behavior and page turning speed; The interaction monitoring unit is used to record the user's highlighting, annotation and note-taking behavior data.

Images