Community comment management method and system

Abstract

The invention discloses a community comment management method and system, and relates to the technical field of data processing, the method can capture user interest changes in real time based on a dynamic interest map generated by registration information and short-term interaction behaviors, quickly screen high-correlation comments through association matching of interest levels, and improve the user experience. It is ensured that the recommended content is deeply matched with user preferences, and meanwhile the redundant data processing burden is greatly reduced. And through comprehensive analysis of a multi-modal quality evaluation system fused text, image and user behavior, and dynamic adjustment of community type differentiation weight distribution and real-time interactive feedback, false, contradictory or low-quality content is effectively identified, and compliance review and high-quality content exposure are enhanced. According to the embodiment, high-quality comments are accurately recommended through the dynamic interest map and multi-modal evaluation, the content ecology is optimized in real time, the timeliness is guaranteed, and the system stability is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a community comment management method and system. Background Art

[0002] With the explosive growth of user-generated content (UGC) in Internet communities, the massive amount of comments on Internet platforms is filled with low-quality and illegal information, resulting in a decline in user experience.

[0003] Existing community comment management solutions usually rely on long-term historical data to build user profiles, and sort and push comment content based on user profiles. However, sorting comment content and pushing information based on static user profiles cannot dynamically capture users' short-term behavioral preferences. Moreover, the push mechanism that solely relies on user profiles lacks multi-dimensional quality assessment of the comment content itself, resulting in implicit violations or low-quality content being misjudged as highly relevant content due to profile bias, and then pushed to users. Summary of the Invention

[0004] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a community comment management method and system, aiming to solve the above-mentioned problems recorded in the prior art.

[0005] A first aspect of the present invention is to provide a community comment management method, the method comprising: Generate an interest graph corresponding to the registered user based on the registration information and short-term interaction signal sequence corresponding to the registered user; Obtaining a work dataset corresponding to a target work published in an Internet community, and extracting a comment dataset corresponding to the target work from the work dataset; According to the interest graph corresponding to the registered user, the comment data set is divided into multiple comment data subsets, and a target data subset that best matches the interest graph is found among the multiple comment data subsets; According to pre-set community review rules, the quality of the multiple comments in the target data subset is evaluated, and the quality scores corresponding to the comments are output; According to the quality scores corresponding to the multiple review contents in the target data subset, the quality scores corresponding to the multiple review contents are sorted from high to low to obtain a quality sorting set; According to the ranking of the multiple review contents in the quality ranking set, the multiple review contents in the quality ranking set are pushed to the registered user.

[0006] According to one aspect of the above technical solution, the steps of dividing the review data set into multiple review data subsets according to the interest graph corresponding to the registered user, and searching for a target data subset that best matches the interest graph from the multiple review data subsets include: Extracting a plurality of interest nodes from the interest graph according to the interest graph corresponding to the registered user; Dividing the complete review data set into multiple review data subsets according to multiple interest nodes in the interest graph; Finding a target data subset that best matches an interest root node in the interest graph from among the plurality of comment data subsets; Among them, the interest root node is a common node of any multiple interest sub-nodes in the interest graph, that is, multiple interest sub-nodes associated with the interest root node that are directly and / or indirectly extended based on the interest root node.

[0007] According to one aspect of the above technical solution, the step of finding a target data subset that best matches an interest root node in the interest graph from among the plurality of comment data subsets includes: Identify the interest graph and output the graph relationship of the interest graph; According to the graph relationship of the interest graph, identifying an interest root node and its associated interest child nodes in the interest graph; Among the multiple comment data subsets, find the target data subset with the highest matching degree with the interest root node.

[0008] According to one aspect of the above technical solution, the steps of performing quality evaluation on multiple comments in the target data subset according to pre-set community comment rules and outputting quality scores corresponding to the comments include: Calling pre-set community comment rule data, decomposing the community comment rule data, and obtaining a number of community comment rules; According to the community review rules defined in the community review rule data, quality evaluation is performed on the multiple review contents in the target data subset respectively, and quality scores corresponding to the review contents are output.

[0009] According to one aspect of the above technical solution, the steps of respectively evaluating the quality of the plurality of comments in the target data subset according to the community comment rules defined in the community comment rule data and outputting the quality scores corresponding to the comments include: According to the community comment rules defined in the community comment rule data, perform multimodal analysis of text, image, and user behavior on multiple comment contents in the target data subset, and output text features, image features, and user behavior features; According to the community type label to which the target work belongs, initial weights are assigned to the text features, image features, and user behavior features respectively; Based on the real-time user interaction data after the comment content is published, the weight configuration corresponding to the comment content is updated online; The text features, image features, user behavior features and their corresponding updated weight configurations are imported into a pre-trained quality evaluation model, and the quality evaluation of the review content is performed using the quality evaluation model to output a quality score.

[0010] According to one aspect of the above technical solution, the step of online updating the weight configuration corresponding to the comment content based on the real-time user interaction data after the comment content is published includes: Based on the real-time interaction data of users on the comments, the interactive behaviors are classified into positive feedback and negative feedback, and the positive feedback intensity value and the negative feedback intensity value are calculated; Dynamically modifying initial weights of text features, image features, and user behavior features according to the positive feedback strength value and the negative feedback strength value; When there is a semantic contradiction between the text features and image features of the review content, the weight of the text features is reduced and the weight of the image features is increased; The attenuation coefficient is calculated based on the time difference between the comment content release time and the current time, and the weight of the comment content that exceeds the preset time threshold is attenuated. Based on the community type to which the target work belongs, the weights of image features or text features are adjusted for scenario-based reinforcement.

[0011] According to one aspect of the above technical solution, the steps further include: When the daily weight adjustment exceeds the preset fluctuation threshold, the automatic weight update will be frozen and the manual review process will be triggered.

[0012] A second aspect of the present invention is to provide a community comment management system, which is applied to the method described in the above technical solution, and the system includes: An interest graph generation module is used to generate an interest graph corresponding to a registered user based on the registration information and short-term interaction signal sequence corresponding to the registered user; A comment data extraction module is used to obtain a work dataset corresponding to a target work published in an Internet community, and extract a comment dataset corresponding to the target work from the work dataset; a target data search module, configured to divide the review data set into a plurality of review data subsets according to the interest graph corresponding to the registered user, and search for a target data subset that best matches the interest graph among the plurality of review data subsets; A review content evaluation module is used to evaluate the quality of multiple review contents in the target data subset according to pre-set community review rules and output a quality score corresponding to the review content; A quality score sorting module is used to sort the quality scores corresponding to the multiple review contents in the target data subset from high to low according to the quality scores corresponding to the multiple review contents, to obtain a quality sorting set; The comment content pushing module is used to push the multiple comment contents in the quality ranking set to the registered user according to the ranking of the multiple comment contents in the quality ranking set.

[0013] The third aspect of the present invention is to provide a readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the method described in the above technical solution.

[0014] The fourth aspect of the present invention is to provide an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in the above technical solution when executing the program.

[0015] Compared with the prior art, the community comment management method and system of the present invention have the following beneficial effects: This embodiment is based on a dynamic interest graph generated based on registration information and short-term interactive behavior. It can capture changes in user interests in real time, quickly screen highly relevant comments through correlation matching at the interest level, ensure that recommended content is deeply aligned with user preferences, and significantly reduce the burden of redundant data processing. In addition, through a comprehensive analysis of text, images, and user behavior integrated into a multimodal quality evaluation system, combined with the dynamic adjustment of differentiated weight distribution of community types and real-time interactive feedback, it can effectively identify false, contradictory, or low-quality content, strengthen compliance review, and expose high-quality content. This embodiment uses dynamic interest graphs and multimodal evaluations to accurately recommend high-quality comments, optimize the content ecosystem in real time, ensure timeliness, and enhance system stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments with reference to the accompanying drawings, in which: Figure 1 Schematic diagram of the process of a community comment management method in one embodiment of the present invention; Figure 2 This is a structural block diagram of a community comment management system in one embodiment of the present invention. DETAILED DESCRIPTION

[0017] To make the objectives, features, and advantages of the present invention more readily apparent, the following detailed description of specific embodiments of the present invention is provided in conjunction with the accompanying drawings. The accompanying drawings illustrate several embodiments of the present invention. However, the present invention may be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and comprehensive understanding of the present invention.

[0018] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly attached to the other element or there may be an intermediate element. When an element is referred to as being "connected to" another element, it may be directly connected to the other element or there may be an intermediate element. The terms "vertical," "horizontal," "left," "right," and similar expressions used herein are for illustrative purposes only.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one skilled in the art to which this invention pertains. The terms used in this specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0020] Example 1 See also Figure 1 The first embodiment of the present invention provides a community comment management method, which includes steps S10 to S60: Step S10: Generate an interest graph corresponding to the registered user based on the registration information corresponding to the registered user and the short-term interaction signal sequence.

[0021] First of all, it should be noted that the community shown in this embodiment can be an online communication community of an Internet platform such as a social platform, a short video platform, an online reading platform, a fan platform, etc., that is, a comment area. Registered users can comment on the content of the work, and can also comment on the comments posted by other users. The purpose of the method shown in this embodiment is to manage these comments so that community comments can be pushed to registered users in accordance with relevant regulations and certain flow rules, thereby creating a good community atmosphere.

[0022] Among them, in the process of community comment management, it is first necessary to generate an interest graph corresponding to the registered user based on the registration information corresponding to the registered user and the short-term interaction signal sequence generated in the historical period on the corresponding Internet platform. The interest graph is used to characterize the interest preferences of the registered user, so as to clarify the interest preferences of the registered user in advance.

[0023] Specifically, in this embodiment, the registration information corresponding to the registered user is stored in the background server and can be directly called, and the short-term interaction signal series is the summary data of a series of interactive behaviors performed by the registered user on the Internet platform, such as high-frequency browsing, liking, commenting and forwarding of a certain type of work and its corresponding comment content. In this embodiment, these interactive behaviors will be recorded, the frequency of occurrence of interactive behaviors will be identified, and the corresponding short-term interaction signal sequence will be generated according to the interactive behaviors. Finally, based on the registration information corresponding to the registered user and the short-term interaction signal sequence, an interest map corresponding to the registered user will be generated to clarify the interest preferences of the registered user.

[0024] Step S20 , obtaining a work dataset corresponding to a target work published in an Internet community, and extracting a comment dataset corresponding to the target work from the work dataset.

[0025] In this embodiment, after generating an interest graph corresponding to the registered user based on the registration information corresponding to the registered user and the short-term interaction signal sequence, a work dataset corresponding to the target work published in the Internet community will be obtained, and then a comment dataset corresponding to the target work will be extracted from the work dataset.

[0026] Among them, the work dataset contains the target work and its corresponding comments. The target work can be text, images and videos published by any registered user, or a random combination of them, and the corresponding comments can be the first-level comments published by any registered user based on the target work, as well as the second-level comments published by any registered user based on the first-level comments. In this embodiment, these comments are extracted from the work dataset to form a comment dataset.

[0027] It should be noted here that since there is a good hierarchical relationship between the multiple comments corresponding to the target work, when extracting comments from the work dataset to form a comment dataset, it is also necessary to maintain the hierarchical relationship between the multiple comments. This hierarchical relationship can be adjusted or not adjusted later, which depends entirely on the content quality of each comment. If it involves illegal content or content metaphor violations, the exposure of the comment content can be reduced or it can be directly deleted through background review.

[0028] Step S30: dividing the comment data set into a plurality of comment data subsets according to the interest graph corresponding to the registered user, and searching for a target data subset that best matches the interest graph among the plurality of comment data subsets.

[0029] Among them, after extracting the comment data set based on the work data set, in order to facilitate the classification of comments, the pre-extracted comment data set will be divided into multiple comment data subsets according to the interest graph corresponding to the registered user, and then the target data subset that best matches the pre-constructed interest graph will be found among the multiple comment data subsets, that is, the comment content that at least partially meets the user's interests will be found among a large number of comment contents.

[0030] In this embodiment, the steps of dividing the review data set into multiple review data subsets according to the interest graph corresponding to the registered user, and searching for a target data subset that best matches the interest graph among the multiple review data subsets include: Extracting a plurality of interest nodes from the interest graph according to the interest graph corresponding to the registered user; Dividing the complete review data set into multiple review data subsets according to multiple interest nodes in the interest graph; Finding a target data subset that best matches an interest root node in the interest graph from among the plurality of comment data subsets; Among them, the interest root node is a common node of any multiple interest sub-nodes in the interest graph, that is, multiple interest sub-nodes associated with the interest root node that are directly and / or indirectly extended based on the interest root node.

[0031] The step of searching for a target data subset that best matches an interest root node in the interest graph from among the plurality of comment data subsets includes: Identify the interest graph and output the graph relationship of the interest graph; According to the graph relationship of the interest graph, identifying an interest root node and its associated interest child nodes in the interest graph; Among the multiple comment data subsets, find the target data subset with the highest matching degree with the interest root node.

[0032] Specifically, in this embodiment, in the process of finding the target data subset that best matches the interest graph among multiple comment data subsets, the multiple interest nodes in the interest graph will be extracted first according to the interest graph corresponding to the registered user, and then the complete comment data set will be divided into multiple comment data subsets according to the multiple interest nodes in the interest graph. Specifically, the interest graph corresponding to the registered user will be identified, and the graph relationship corresponding to the interest graph will be output. According to the graph relationship, the interest root node in the interest graph and the interest sub-nodes associated with it are identified. The number of interest sub-nodes is usually multiple. Finally, among the multiple pre-divided comment data subsets, the target data subset with the best match with the interest root node is found to facilitate the subsequent classification of the comment content in the target interest subset.

[0033] More specifically, in the process of identifying the interest root node and its associated interest sub-nodes according to the graph relationship of the interest graph, the interest root node in the graph relationship will be identified first. There can be multiple interest root nodes, and then the number of interest sub-nodes associated with the interest root node will be counted. According to the number of interest sub-nodes associated with the interest root node, they will be sorted to determine the interest root node that best corresponds to the registered user's interest preference. Then, based on the interest root node, the target data subset that best matches the interest graph will be found in multiple comment data subsets, which is equivalent to further limiting the number of comment contents through the interest preference indicator.

[0034] Step S40 , performing quality evaluation on the multiple comments in the target data subset according to pre-set community comment rules, and outputting quality scores corresponding to the comment contents.

[0035] It should be noted here that in this embodiment, when managing community comments, a management model of first limiting the quantity and then evaluating the quality is adopted. The number of comment contents is first limited according to the interest preferences of registered users, and then the quality of the limited comment contents is evaluated, that is, the comment contents are evaluated, and finally the comment contents that are consistent with the registered users and have higher value are output.

[0036] Specifically, in this embodiment, when performing quality evaluation on multiple review contents in the target data subset, the quality evaluation is performed according to pre-set community evaluation rules, and finally a quality score corresponding to each review content is output.

[0037] Step S50 , sorting the quality scores corresponding to the multiple review contents in the target data subset from high to low according to the quality scores corresponding to the multiple review contents, to obtain a quality sorting set.

[0038] In this embodiment, the higher the quality score of the review content, the more exposure it can get on the Internet platform. By increasing the exposure, the review content can be recognized by more registered users.

[0039] Specifically, in this embodiment, after quality evaluation is performed on each comment content in the target data subset and a quality score is output, the corresponding multiple comment contents in the target data subset are sorted according to the quality score, and a quality sorted set is output. In this quality sorted set, there are multiple first-level comments that directly comment on the target work, and under each first-level comment, there are multiple second-level comments that comment on the first-level comments. There is a good hierarchical relationship between the first-level comments and the second-level comments, and this hierarchical relationship is further optimized after the number of comment contents is limited according to the interest preferences of registered users.

[0040] Step S60: Pushing the multiple review contents in the quality ranking set to the registered user according to the ranking of the multiple review contents in the quality ranking set.

[0041] In this embodiment, after the number of review contents is limited and their corresponding quality is evaluated and sorted, and a quality sorted set is output, a data set of high-quality review contents is obtained, and then multiple high-quality and relevant review contents in the quality sorted set are pushed to registered users.

[0042] Compared with the prior art, the community comment management method shown in this embodiment has the following beneficial effects: This embodiment is based on a dynamic interest graph generated based on registration information and short-term interactive behavior. It can capture changes in user interests in real time, quickly screen highly relevant comments through correlation matching at the interest level, ensure that recommended content is deeply aligned with user preferences, and significantly reduce the burden of redundant data processing. In addition, through a comprehensive analysis of text, images, and user behavior integrated into a multimodal quality evaluation system, combined with the dynamic adjustment of differentiated weight distribution of community types and real-time interactive feedback, it can effectively identify false, contradictory, or low-quality content, strengthen compliance review, and expose high-quality content. This embodiment uses dynamic interest graphs and multimodal evaluations to accurately recommend high-quality comments, optimize the content ecosystem in real time, ensure timeliness, and enhance system stability.

[0043] Example 2 The second embodiment of the present invention also provides a community comment management method, which is basically similar to the community comment management method shown in the first embodiment, except that: In this embodiment, the steps of performing quality evaluation on multiple comments in the target data subset according to pre-set community comment rules and outputting quality scores corresponding to the comments include: Calling pre-set community comment rule data, decomposing the community comment rule data, and obtaining a number of community comment rules; According to the community review rules defined in the community review rule data, quality evaluation is performed on the multiple review contents in the target data subset respectively, and quality scores corresponding to the review contents are output.

[0044] The step of respectively evaluating the quality of the plurality of comments in the target data subset according to the community comment rules defined in the community comment rule data and outputting the quality scores corresponding to the comment contents includes: According to the community comment rules defined in the community comment rule data, perform multimodal analysis of text, image, and user behavior on multiple comment contents in the target data subset, and output text features, image features, and user behavior features; According to the community type label to which the target work belongs, initial weights are assigned to the text features, image features, and user behavior features respectively; Based on the real-time user interaction data after the comment content is published, the weight configuration corresponding to the comment content is updated online; The text features, image features, user behavior features and their corresponding updated weight configurations are imported into a pre-trained quality evaluation model, and the quality evaluation of the review content is performed using the quality evaluation model to output a quality score.

[0045] The step of online updating the weight configuration corresponding to the comment content based on the real-time user interaction data after the comment content is published includes: Based on the real-time interaction data of users on the comments, the interactive behaviors are classified into positive feedback and negative feedback, and the positive feedback intensity value and the negative feedback intensity value are calculated; Dynamically modifying initial weights of text features, image features, and user behavior features according to the positive feedback strength value and the negative feedback strength value; When there is a semantic contradiction between the text features and image features of the review content, the weight of the text features is reduced and the weight of the image features is increased; The attenuation coefficient is calculated based on the time difference between the comment content release time and the current time, and the weight of the comment content that exceeds the preset time threshold is attenuated. Based on the community type to which the target work belongs, the weights of image features or text features are adjusted for scenario-based reinforcement.

[0046] In addition, the steps further include: When the daily weight adjustment exceeds the preset fluctuation threshold, the automatic weight update will be frozen and the manual review process will be triggered.

[0047] Specifically, in this embodiment: First, pre-set community comment rule data is retrieved from the platform's rule library. This rule data includes content compliance standards, quality assessment indicators, and community-specific constraints. The rule engine then decomposes complex rules into atomic rule units. For example, "image and text mismatch and abnormal likes" can be broken down into separate rules for checking image and text consistency and determining abnormal likes. These rules are then prioritized, ensuring that high-weighted rules (such as detecting politically sensitive keywords) are executed first, while low-weighted rules (such as checking the minimum word count) are processed later, thereby improving rule execution efficiency and coverage.

[0048] The process of multi-dimensional analysis of the comments in the target data subset includes: Text parsing: Extracting the sentiment polarity, topic distribution, and logical coherence score of the review text, and simultaneously detecting the matching degree of sensitive words; Image analysis: Identify sensitive objects (such as QR codes and exposed areas) and technical indicators (such as resolution and color saturation) in review images; User behavior analysis: collect statistics on real-time interaction data (like rate, reporting rate, secondary dissemination depth) and calculate standardized behavior indicators.

[0049] Then, based on the community type label of the target work, we assign initial weights to different modal features. For example, in e-commerce communities, image features are weighted at 0.6 to emphasize product authenticity verification, while in academic communities, text features are weighted at 0.8 to emphasize rigorous argumentation. This approach adapts to the core needs of different scenarios.

[0050] Among them, the weight configuration is dynamically adjusted based on real-time user interaction data, including: Quantification of positive and negative feedback: Calculate the strength of positive feedback (e.g., number of likes × 1 + number of reposts × 3) and negative feedback (e.g., number of reports × 5), and adjust the weights accordingly. For example, if the number of reports exceeds a threshold within an hour, the weight of the e-commerce community image will be reduced to 50% of its initial value. Cross-modal contradiction handling: When there is a semantic contradiction between text and image features, such as a text description of "strong battery life" but an image showing a low battery warning, the text weight is automatically reduced and the image weight is increased to prevent the spread of false content; Time-sensitive decay correction: Calculate the decay coefficient based on the time the comment is published, and exponentially decay the weight of comments that exceed the preset time limit, such as 72 hours, to prevent old content from occupying the traffic entrance for a long time.

[0051] Implement differentiated strengthening strategies for different community types: In e-commerce communities, if a comment image is clicked and enlarged multiple times, the image weight will be increased. In the academic community, if a comment cites authoritative literature, the text weight is increased to enhance credibility.

[0052] At the same time, a circuit breaker mechanism has been introduced for abnormal fluctuations: if the daily weight adjustment exceeds a preset threshold (e.g., ±30%), automatic updates are immediately frozen and a manual review process is triggered. After review, the administrator will confirm whether to resume automatic processing or permanently mark the abnormal status, thereby preventing rating distortion caused by malicious review manipulation or system failures.

[0053] This embodiment significantly improves the accuracy and adaptability of review quality assessments through multimodal dynamic scoring and real-time feedback-driven weighting adjustments. By integrating the dual advantages of a rule engine and machine learning models, it effectively identifies inconsistencies between text and images, sensitive content, and low-quality information. It also ensures the vitality of the content ecosystem and system stability through time-sensitive decay and circuit-breaker mechanisms, providing an efficient and reliable management solution for internet communities.

[0054] Example 3 See also Figure 2 A third embodiment of the present invention provides a community comment management system, which is applied to the method described in any of the above embodiments, and the system includes: An interest graph generation module 10 is configured to generate an interest graph corresponding to a registered user based on the registration information corresponding to the registered user and a short-term interaction signal sequence; The comment data extraction module 20 is used to obtain a work dataset corresponding to a target work published in an Internet community, and extract a comment dataset corresponding to the target work from the work dataset; a target data search module 30 for dividing the review data set into a plurality of review data subsets according to the interest graph corresponding to the registered user, and searching for a target data subset that best matches the interest graph among the plurality of review data subsets; A review content evaluation module 40 is configured to evaluate the quality of the multiple review contents in the target data subset according to pre-set community review rules and output a quality score corresponding to the review content; A quality score sorting module 50 is configured to sort the quality scores of the review contents in the target data subset from high to low according to the quality scores of the review contents, to obtain a quality sorting set; The comment content pushing module 60 is configured to push the multiple comment contents in the quality ranking set to the registered user according to the ranking of the multiple comment contents in the quality ranking set.

[0055] Compared with the prior art, the community comment management system shown in this embodiment has the following beneficial effects: This embodiment is based on a dynamic interest graph generated based on registration information and short-term interactive behavior. It can capture changes in user interests in real time, quickly screen highly relevant comments through correlation matching at the interest level, ensure that recommended content is deeply aligned with user preferences, and significantly reduce the burden of redundant data processing. In addition, through a comprehensive analysis of text, images, and user behavior integrated into a multimodal quality evaluation system, combined with the dynamic adjustment of differentiated weight distribution of community types and real-time interactive feedback, it can effectively identify false, contradictory, or low-quality content, strengthen compliance review, and expose high-quality content. This embodiment uses dynamic interest graphs and multimodal evaluations to accurately recommend high-quality comments, optimize the content ecosystem in real time, ensure timeliness, and enhance system stability.

[0056] Example 4 A fourth embodiment of the present invention provides a readable storage medium having computer instructions stored thereon, which implement the steps of the method described in the above embodiment when executed by a processor.

[0057] Example 5 A fifth embodiment of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in the above embodiment when executing the program.

[0058] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0059] Those skilled in the art will appreciate that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable storage medium for use by an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device), or for use in conjunction with such instruction execution system, apparatus, or device. For purposes of this specification, a "computer-readable storage medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by an instruction execution system, apparatus, or device, or in conjunction with such instruction execution system, apparatus, or device.

[0060] More specific examples (a non-exhaustive list) of computer-readable storage media include the following: an electrical connection with one or more wires (electronic device), a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable storage medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0061] It should be understood that various components of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the aforementioned embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one or a combination of the following technologies known in the art may be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0062] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0063] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A community comment management method, characterized in that: The method comprises: Generate an interest graph corresponding to the registered user based on the registration information and short-term interaction signal sequence corresponding to the registered user; Obtaining a work dataset corresponding to a target work published in an Internet community, and extracting a comment dataset corresponding to the target work from the work dataset; According to the interest graph corresponding to the registered user, the comment data set is divided into multiple comment data subsets, and a target data subset that best matches the interest graph is found among the multiple comment data subsets; According to pre-set community review rules, the quality of the multiple comments in the target data subset is evaluated, and the quality scores corresponding to the comments are output; According to the quality scores corresponding to the multiple review contents in the target data subset, the quality scores corresponding to the multiple review contents are sorted from high to low to obtain a quality sorting set; According to the ranking of the multiple review contents in the quality ranking set, the multiple review contents in the quality ranking set are pushed to the registered user.

2. The community comment management method according to claim 1, characterized in that: The steps of dividing the review data set into a plurality of review data subsets according to the interest graph corresponding to the registered user, and searching for a target data subset that best matches the interest graph from the plurality of review data subsets include: Extracting a plurality of interest nodes from the interest graph according to the interest graph corresponding to the registered user; Dividing the complete review data set into multiple review data subsets according to multiple interest nodes in the interest graph; Finding a target data subset that best matches an interest root node in the interest graph from among the plurality of comment data subsets; Among them, the interest root node is a common node of any multiple interest sub-nodes in the interest graph, that is, multiple interest sub-nodes associated with the interest root node that are directly and / or indirectly extended based on the interest root node.

3. The community comment management method according to claim 2, characterized in that: The step of searching for a target data subset that best matches an interest root node in the interest graph from among the plurality of comment data subsets includes: Identify the interest graph and output the graph relationship of the interest graph; According to the graph relationship of the interest graph, identifying an interest root node and its associated interest child nodes in the interest graph; Among the multiple comment data subsets, find the target data subset with the highest matching degree with the interest root node.

4. The community comment management method according to claim 1, characterized in that: The steps of evaluating the quality of the multiple comments in the target data subset according to the pre-set community comment rules and outputting the quality scores corresponding to the comments include: Calling pre-set community comment rule data, decomposing the community comment rule data, and obtaining a number of community comment rules; According to the community review rules defined in the community review rule data, quality evaluation is performed on the multiple review contents in the target data subset respectively, and quality scores corresponding to the review contents are output.

5. The community comment management method according to claim 4, characterized in that: The step of respectively evaluating the quality of a plurality of review contents in the target data subset according to the community review rules defined in the community review rule data and outputting quality scores corresponding to the review contents comprises: According to the community comment rules defined in the community comment rule data, perform multimodal analysis of text, image, and user behavior on multiple comment contents in the target data subset, and output text features, image features, and user behavior features; According to the community type label to which the target work belongs, initial weights are assigned to the text features, image features, and user behavior features respectively; Based on the real-time user interaction data after the comment content is published, the weight configuration corresponding to the comment content is updated online; The text features, image features, user behavior features and their corresponding updated weight configurations are imported into a pre-trained quality evaluation model, and the quality evaluation of the review content is performed using the quality evaluation model to output a quality score.

6. The community comment management method according to claim 5, characterized in that: The step of online updating the weight configuration corresponding to the comment content based on the real-time user interaction data after the comment content is published includes: Based on the real-time interaction data of users on the comments, the interactive behaviors are classified into positive feedback and negative feedback, and the positive feedback intensity value and the negative feedback intensity value are calculated; Dynamically modifying initial weights of text features, image features, and user behavior features according to the positive feedback strength value and the negative feedback strength value; When there is a semantic contradiction between the text features and image features of the review content, the weight of the text features is reduced and the weight of the image features is increased; The attenuation coefficient is calculated based on the time difference between the comment content release time and the current time, and the weight of the comment content that exceeds the preset time threshold is attenuated. Based on the community type to which the target work belongs, the weights of image features or text features are adjusted for scenario-based reinforcement.

7. The community comment management method according to claim 6, characterized in that: The steps also include: When the daily weight adjustment exceeds the preset fluctuation threshold, the automatic weight update will be frozen and the manual review process will be triggered.

8. A community comment management system, characterized in that: The method according to any one of claims 1 to 7, wherein the system comprises: An interest graph generation module is used to generate an interest graph corresponding to a registered user based on the registration information and short-term interaction signal sequence corresponding to the registered user; A comment data extraction module is used to obtain a work dataset corresponding to a target work published in an Internet community, and extract a comment dataset corresponding to the target work from the work dataset; a target data search module, configured to divide the review data set into a plurality of review data subsets according to the interest graph corresponding to the registered user, and search for a target data subset that best matches the interest graph among the plurality of review data subsets; A review content evaluation module is used to evaluate the quality of multiple review contents in the target data subset according to pre-set community review rules and output a quality score corresponding to the review content; A quality score sorting module is used to sort the quality scores corresponding to the multiple review contents in the target data subset from high to low according to the quality scores corresponding to the multiple review contents, to obtain a quality sorting set; The comment content pushing module is used to push the multiple comment contents in the quality ranking set to the registered user according to the ranking of the multiple comment contents in the quality ranking set.

9. A readable storage medium having computer instructions stored thereon, characterized in that: When the instruction is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the method according to any one of claims 1 to 7 are implemented.

Images