Information correlation analysis method and system based on network comment big data

Abstract

The application provides an information correlation analysis method and system based on network comment big data, relates to the technical field of big data, and first extracts a transient propagation segment from network comment big data, identifies a unit trigger node and extracts features, generates a correlation unit to form a comment correlation unit set, extracts unit correlation information of the correlation unit, analyzes an evolution influence relationship to form a comment correlation dynamic factor evolution sequence, uses the correlation unit as a topological node, determines dynamic correlation parameters and directions by using a dynamic factor evolution process, constructs a comment dynamic correlation topology, analyzes comment dynamic correlation topology structure integration into a comment correlation evolution trend and transmits the comment correlation evolution trend to a terminal, so that network comment correlation and dynamic evolution can be deeply analyzed, timely comment dynamics can be mastered, scientific and reasonable decisions can be made, and the efficiency and accuracy of comment analysis are improved.

Description

Technical Field

[0001] This application relates to the field of big data technology, and more specifically, to a method and system for information correlation analysis based on big data of online comments. Background Technology

[0002] Currently, the analysis methods for online comments mainly focus on simple classification of comment content, sentiment analysis, and preliminary tracing of dissemination paths. However, these existing methods have several limitations. On the one hand, the analysis of the comment dissemination process is not in-depth enough, often only focusing on the overall dissemination trend while ignoring the various key nodes in the dissemination process and their complex interrelationships. For example, different triggering nodes may play different roles in promoting the spread of comments, but existing methods struggle to accurately identify these nodes and their characteristics. On the other hand, existing methods lack a comprehensive grasp of the dynamic evolution of comments, cannot accurately analyze the evolutionary influence relationships between different comment-related units, cannot promptly capture dynamic changes in the comment dissemination process, and are unable to accurately judge the evolutionary trend of comment associations. Summary of the Invention

[0003] In view of this, the purpose of this application is to provide a method and system for information correlation analysis based on big data of online comments.

[0004] According to a first aspect of this application, a method for information correlation analysis based on big data of online comments is provided, the method comprising:

[0005] Extracting instantaneous propagation segments from online comment big data, identifying unit trigger nodes in these instantaneous propagation segments, extracting propagation trigger features, content trigger features, and subject trigger features from these unit trigger nodes, determining the unit type based on the extracted trigger features, generating associated units corresponding to each unit trigger node, and integrating all associated units to form a set of comment associated units;

[0006] Based on the associated units in the comment associated unit set, the unit association information of each associated unit is extracted, the evolutionary influence relationship between the unit association information is analyzed, the degree of evolutionary influence between different associated units is calculated, the dynamic factor evolution type is determined according to the degree of evolutionary influence, and all dynamic factor evolution processes are integrated to form the comment associated dynamic factor evolution sequence.

[0007] The associated units in the set of associated units in the comment association are used as topology nodes. The dynamic factor evolution process in the dynamic factor evolution sequence of the comment association is used to determine the dynamic association strength parameters and dynamic association direction between each topology node. The topology nodes are connected according to the dynamic association strength parameters and dynamic association direction to form the comment dynamic association topology.

[0008] The analysis examines the distribution of topological nodes, the evolution of dynamic association strength parameters between nodes, and the propagation flow corresponding to the dynamic association direction in the dynamic association topology of comments. The topological node distribution, the evolution of dynamic association strength parameters, and the propagation flow are integrated into the comment association evolution trend, and the comment association evolution trend is transmitted to the comment analysis terminal.

[0009] According to a second aspect of this application, an information association analysis system based on big data of online comments is provided. The system includes a machine-readable storage medium and a processor. The machine-readable storage medium stores machine-executable instructions. When the processor executes the machine-executable instructions, the system implements the aforementioned information association analysis method based on big data of online comments.

[0010] According to a third aspect of this application, a computer-readable storage medium is provided, wherein computer-executable instructions are stored therein, and when the computer-executable instructions are executed, the aforementioned information association analysis method based on big data of online comments is implemented.

[0011] Based on any of the above aspects, the technical effect of this application is as follows:

[0012] By accurately extracting instantaneous propagation fragments from online comment big data and identifying unit trigger nodes, we can extract propagation trigger, content trigger, and subject trigger features. This allows us to accurately determine unit types and generate associated units, which are then integrated into a set of comment-related units. Based on this set, we extract unit association information, analyze evolutionary influence relationships, calculate the degree of evolutionary influence to determine the dynamic factor evolution type, and form a dynamic factor evolution sequence for comment association. This sequence presents the dynamic evolution process between comment-related units. Next, using associated units as topological nodes, we determine the dynamic association strength parameters and directions using the dynamic factor evolution process, constructing a dynamic association topology for comments. Finally, we analyze the distribution of topological nodes, the evolutionary laws of dynamic association strength parameters, and the propagation evolution flow, integrating this into a comment association evolution trend and transmitting it to the comment analysis terminal. This helps to grasp comment dynamics in a timely manner, make scientific and reasonable decisions, and improve the efficiency and accuracy of comment analysis. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 A flowchart illustrating the information association analysis method based on big data of online comments provided in an embodiment of this application is shown.

[0015] Figure 2 This paper illustrates a schematic diagram of the component structure of an information association analysis system based on big data of online comments provided in an embodiment of this application. Detailed Implementation

[0016] The embodiments of this application are described below with reference to the accompanying drawings. It should be understood that the embodiments described below with reference to the accompanying drawings are exemplary descriptions for explaining the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions of the embodiments of this application.

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings. The technical solutions of the embodiments of this application and the technical effects produced by the technical solutions of this application will be explained below through the description of several exemplary embodiments. It should be noted that the following embodiments can be referenced, borrowed from, or combined with each other, and the same terms, similar features, and similar implementation steps in different embodiments will not be described again.

[0018] Figure 1 This illustration shows a flowchart of the information association analysis method and system based on big data of online comments provided in an embodiment of this application. It should be understood that in other embodiments, the order of some steps in the information association analysis method based on big data of online comments in this embodiment can be shared according to actual needs, or some steps can be omitted or maintained. The detailed steps of the information association analysis method based on big data of online comments include:

[0019] Step S110: Extract instantaneous propagation segments from the big data of online comments, identify unit trigger nodes in the instantaneous propagation segments, extract propagation trigger features, content trigger features and subject trigger features of the unit trigger nodes, determine the unit type based on the extracted trigger features, generate associated units corresponding to each unit trigger node, and integrate all associated units to form a set of comment associated units.

[0020] In this embodiment, firstly, after obtaining full data authorization from relevant entities through a legitimate and compliant public data interface, big data on online comments is acquired. This data covers publicly available comments from social media platforms, news media, forums, and other channels, encompassing all relevant comments within a preset time frame. Privacy-sensitive data is anonymized by replacing users' real identity information with unique anonymous identifiers. For comment data containing personal geographical locations, aggregation processing technology is used to categorize them into preset granular geographical regions before storage, ensuring that individual identities cannot be deduced from the data.

[0021] Step S111: Filter comment data containing information on a surge in instantaneous dissemination from the big data of online comments. The comment data containing information on a surge in instantaneous dissemination covers records of surges in the posting, forwarding, and receiving of comment content within a short period of time.

[0022] In this embodiment, each comment in the large dataset of online comments is traversed, and the posting time, forwarding time, receiving time, and corresponding quantity information of each comment are extracted. For each comment, the rate of change of posting quantity, forwarding quantity, and receiving quantity within a preset short time window are calculated. The rate of change is calculated as follows: the quantity in the current time window minus the quantity in the previous time window, then divided by the quantity in the previous time window. If the rate of change of posting quantity, forwarding quantity, or receiving quantity of a comment exceeds a preset surge threshold, the comment is determined to contain instantaneous surge information and is filtered out. The filtered comment data includes posting surge records (recording the rapid increase in the number of posts), forwarding surge records (recording the rapid increase in the number of forwards), and receiving surge records (recording the rapid increase in the number of receivings) within the short time window.

[0023] Step S112: Based on the timestamp sequence in the instantaneous propagation surge information, divide the comment data containing the instantaneous propagation surge information into multiple consecutive instantaneous propagation periods. Each instantaneous propagation period corresponds to a comment instantaneous propagation process, forming instantaneous propagation segments in multiple comment propagation processes.

[0024] In this embodiment, all timestamps in the comment data containing information about the surge in instantaneous propagation are first extracted. These timestamps record the surge times of comment posting, forwarding, and receiving, and each timestamp contains specific date and time information. Then, all extracted timestamps are sorted by numerical value to obtain an ordered timestamp sequence, arranged in ascending order. Next, the difference between any two adjacent timestamps in the ordered timestamp sequence is calculated, and all adjacent timestamp differences are recorded. Subsequently, the distribution of adjacent timestamp differences is analyzed to find the boundary point where the difference value is greater than other differences. The difference corresponding to this boundary point can distinguish different instantaneous propagation periods. Specifically, the analysis method is as follows: the frequency of occurrence of all adjacent timestamp differences is statistically analyzed, and the average and standard deviation of the differences are calculated. If a certain difference is greater than the average plus a preset multiple of the standard deviation, then this difference is determined to be a boundary point, because this difference is significantly greater than the differences of other adjacent timestamps, indicating a sudden change in the time interval between the two timestamps, corresponding to different instantaneous propagation periods.

[0025] Step S1121: Using the dividing point as the boundary, divide the ordered timestamp sequence into multiple consecutive time subsequences, each time subsequence corresponding to an instantaneous propagation period.

[0026] In this embodiment, after determining the dividing point, the ordered timestamp sequence is divided according to the dividing point. For example, if the ordered timestamp sequence is [T1, T2, T3, T4, T5, T6], and the dividing point is the difference between T3 and T4, then the ordered timestamp sequence is divided into two time subsequences: [T1, T2, T3] and [T4, T5, T6]. Each time subsequence corresponds to an instantaneous propagation period. The instantaneous propagation period corresponding to the first time subsequence is the time range from T1 to T3, and the second is the time range from T4 to T6.

[0027] Step S1122: For each time subsequence, extract all comment data within that time subsequence. The comment data includes records of posting surges, forwarding surges, and receiving surges during that instantaneous propagation period.

[0028] In this embodiment, all comment data is traversed, and the timestamp of each comment is compared with the time range of each time subsequence. If the timestamp of a comment falls within the time range of a certain time subsequence, the comment is extracted to the temporary storage area corresponding to that time subsequence. For example, the time range of the time subsequence [T1, T2, T3] is from T1 to T3. If the timestamp of a comment is T2, it is extracted to the temporary storage area. Ultimately, the temporary storage area corresponding to each time subsequence contains all posting surge records, forwarding surge records, and receiving surge records within that instantaneous propagation period.

[0029] Step S1123: Bind the comment data corresponding to each time subsequence to the identifier of the instantaneous propagation period to form an instantaneous propagation segment corresponding to each instantaneous propagation period.

[0030] In this embodiment, a unique identifier is assigned to each instantaneous propagation period, such as the combination of the start and end times of the time subsequence, like [T1-T3]. Then, the comment data corresponding to the time subsequence is bound to the identifier and stored in the same data structure, such as a dictionary or a data table. Each data structure corresponds to an instantaneous propagation segment, containing the identifier and the corresponding comment data.

[0031] Step S1124: Verify the integrity of the comment data in each instantaneous propagation segment, requiring each instantaneous propagation segment to contain all relevant post surge records, forward surge records, and receive surge records within that instantaneous propagation period.

[0032] In this embodiment, each instantaneous propagation segment is traversed, and the number of surges in posting, forwarding, and receiving records contained therein is counted. At the same time, the number of relevant records in the original network comment big data during that instantaneous propagation period is compared with the number of relevant records in that instantaneous propagation period. If there is a difference in the number, the comment data during that instantaneous propagation period is extracted again until the number is consistent, ensuring that each instantaneous propagation segment contains complete relevant records.

[0033] Step S1125: Number the verified instantaneous propagation segments, assign consecutive numbers to the instantaneous propagation segments according to the time sequence of the instantaneous propagation periods, establish a mapping relationship table between the instantaneous propagation segment numbers and the corresponding instantaneous propagation period identifiers, and record the instantaneous propagation period identifier, time subsequence range, and included comment data type for each number.

[0034] In this embodiment, each instantaneous propagation segment is assigned a consecutive number starting from 1 according to the chronological order of the instantaneous propagation periods. For example, the first instantaneous propagation segment is numbered 1, the second is numbered 2, and so on. Then, a mapping table is established. Each row in the table contains a number, an instantaneous propagation period identifier, a time subsequence range, and the type of comment data included. For example, the instantaneous propagation period identifier corresponding to number 1 is [T1-T3], the time subsequence range is T1 to T3, and the type of comment data included is posting surge record, forwarding surge record, and receiving surge record. This mapping table is stored in a designated storage location for subsequent steps to retrieve.

[0035] Step S113: Traverse each instantaneous propagation segment and identify the entity that triggers the propagation surge in the instantaneous propagation segment. The entity that triggers the propagation surge includes the entity that publishes the surge, the entity that forwards the surge, and the entity that receives the surge. Treat each entity that triggers the propagation surge as a unit trigger node in the instantaneous propagation segment.

[0036] In this embodiment, each instantaneous propagation segment is traversed, and the anonymous identifier of the publishing entity is extracted from the publishing surge records. The number of publications by each publishing entity is counted. If the number of publications by a certain publishing entity exceeds a preset publishing quantity threshold, the publishing entity is determined to be a publishing surge entity. The anonymous identifier of the forwarding entity is extracted from the forwarding surge records, and the number of forwards by each forwarding entity is counted. If the number of forwards by a certain forwarding entity exceeds a preset forwarding quantity threshold, the forwarding entity is determined to be a forwarding surge entity. The anonymous identifier of the receiving entity is extracted from the receiving surge records, and the number of receptions by each receiving entity is counted. If the number of receptions by a certain receiving entity exceeds a preset receiving quantity threshold, the receiving entity is determined to be a receiving surge entity. All publishing surge entities, forwarding surge entities, and receiving surge entities are used as unit trigger nodes in the instantaneous propagation segment. Each unit trigger node is an anonymous identifier, representing the entity that caused the propagation surge.

[0037] Step S114: For each unit trigger node, extract the information on the change in propagation speed, the expansion of propagation range, and the increase in propagation level when the unit trigger node causes a propagation surge, and integrate the information on the change in propagation speed, the expansion of propagation range, and the increase in propagation level into the propagation triggering feature of the unit trigger node.

[0038] In this embodiment, for each unit trigger node, the propagation speed change information when the unit trigger node causes a propagation surge is first extracted. Specifically, the propagation speed of comments before and after the propagation surge is recorded. The propagation speed is calculated using the number of comments propagated per unit time. For example, if the number of comments propagated per unit time before the propagation surge is V1, and the number of comments propagated per unit time after the propagation surge is V2, then the propagation speed change information is V2-V1. Then, the propagation range expansion information is extracted. The propagation range is represented by the geographical area or user group range covered by the comment propagation. The propagation range before and after the propagation surge is recorded, and the expansion area or expansion value of the propagation range is calculated. The number of users is used as an example. For instance, if the area of ​​propagation before the surge is S1 and after the surge is S2, then the propagation range expansion information is S2-S1. Next, the propagation level increase information is extracted. The propagation level is represented by the number of user levels through which the comment propagation passes. The number of propagation levels before and after the propagation surge caused by the unit trigger node is recorded. For example, if the number of propagation levels before the surge is L1 and after the surge is L2, then the propagation level increase information is L2-L1. Finally, the propagation speed change information, propagation range expansion information, and propagation level increase information are integrated into the propagation trigger feature of the unit trigger node. This propagation trigger feature contains information in three dimensions, with each dimension corresponding to one type of information.

[0039] Step S115: Extract the text topic change information, text sentiment change information, and text keyword density change information of the comment content corresponding to the surge in dissemination caused by each unit trigger node, and integrate the text topic change information, text sentiment change information, and text keyword density change information into the content trigger feature of the unit trigger node.

[0040] In this embodiment, for each unit trigger node, the text topic change information of the corresponding comment content when the unit trigger node causes a surge in dissemination is first extracted. Specifically, a text topic recognition algorithm is used to identify the text topic of the comment content before and after the unit trigger node causes a surge in dissemination. For example, the text topic before the surge is topic A, and the topic after the surge is topic B. If topic A and topic B are different, the text topic change information is the change from topic A to topic B. Then, text sentiment change information is extracted, and a text sentiment analysis algorithm is used to analyze the text sentiment tendency of the comment content before and after the unit trigger node causes a surge in dissemination. For example, the sentiment tendency before the surge is positive. If the subsequent surge in dissemination is negative, then the change in text sentiment is a shift from positive to negative. Next, the text keyword density change information is extracted, and the text keyword density of the comments before and after the surge in dissemination triggered by the unit's trigger node is calculated. The text keyword density is the ratio of the number of keywords to the total number of words in the text. For example, if the keyword density before the surge in dissemination is D1 and after the surge in dissemination is D2, then the text keyword density change information is D2-D1. Finally, the text topic change information, text sentiment change information, and text keyword density change information are integrated into the content trigger feature of the unit's trigger node. This content trigger feature contains information in three dimensions, with each dimension corresponding to one type of information.

[0041] Step S116: Extract the subject activity change information, subject association network change information, and subject propagation habit change information when each unit trigger node triggers a propagation surge, and integrate the subject activity change information, subject association network change information, and subject propagation habit change information into the subject triggering feature of the unit trigger node.

[0042] In this embodiment, for each unit trigger node, firstly, the subject activity change information when the unit trigger node triggers a propagation surge is extracted. Subject activity is calculated by the number of posts, forwards, and interactions of the unit trigger node within a unit time. The subject activity before and after the propagation surge is recorded. For example, if the subject activity before the propagation surge is A1 and after the propagation surge is A2, then the subject activity change information is A2-A1. Then, the subject association network change information is extracted. The subject association network is represented by the number of associated subjects and the activity of associated subjects of the unit trigger node. The number of associated subjects and the activity of associated subjects before and after the propagation surge are recorded. The change in the number of associated subjects and the change in the activity of associated subjects are calculated. For example, if the number of associated subjects before the propagation surge is N1 and after the propagation surge is N2, then the change in the number of associated subjects is N2-N1; if the average activity of associated subjects before the propagation surge is A3 and after the propagation surge is A4, then the change in the average activity of associated subjects is A4-A3. The subject association network change information package... The data includes changes in the number of associated entities and changes in the average activity level of associated entities. Next, information on changes in entity dissemination habits is extracted. These habits are represented by the dissemination time, content type, and channel of the trigger node for that unit. The data records the dissemination time, content type, and channel before and after a surge in dissemination triggered by the trigger node. For example, if the dissemination time before the surge was during the day and after was at night, the dissemination time change represents a change from day to night; if the content type before the surge was text and after was video, the content type change represents a change from text to video; if the dissemination channel before the surge was social platform A and after was social platform B, the channel change represents a change from social platform A to social platform B. The information on changes in entity dissemination habits includes changes in dissemination time, content type, and channel. Finally, the information on changes in entity activity, entity association network, and entity dissemination habits is integrated into the entity trigger feature of the trigger node for that unit. This trigger feature contains information in multiple dimensions, with each dimension corresponding to one type of information.

[0043] Step S117: Collect the propagation trigger features, content trigger features, and subject trigger features of all unit trigger nodes. Classify the collected propagation trigger features, content trigger features, and subject trigger features of all unit trigger nodes. Determine the unit type based on the unit trigger attributes reflected by the trigger features. The unit type includes propagation trigger type units, content trigger type units, and subject trigger type units.

[0044] In this embodiment, the propagation trigger features, content trigger features, and subject trigger features of all unit trigger nodes are first collected and stored in the same dataset. Each unit trigger node corresponds to a data entry containing three trigger features. Then, the three trigger features of all unit trigger nodes are classified. Specifically, a feature classification algorithm is used to analyze the weights of the three trigger features of each unit trigger node. If the weight of the propagation trigger feature is the highest, the unit type of the unit trigger node is determined to be a propagation trigger type unit; if the weight of the content trigger feature is the highest, it is determined to be a content trigger type unit; if the weight of the subject trigger feature is the highest, it is determined to be a subject trigger type unit. The feature classification algorithm is obtained through training, and the training data consists of the three trigger features of unit trigger nodes with clearly defined unit types in history.

[0045] Step S118: Assign a unique unit identifier to each unit trigger node, bind the propagation trigger feature, content trigger feature, and main trigger feature of the unit trigger node with the corresponding unit identifier and unit type, and generate the associated unit corresponding to each unit trigger node.

[0046] In this embodiment, a unique unit identifier is assigned to each unit trigger node, such as a unique identifier represented by a string, like "Unit_001". Then, the propagation trigger feature, content trigger feature, and main trigger feature of each unit trigger node are bound to the corresponding unit identifier and unit type and stored in the same data structure, such as a dictionary. Each data structure corresponds to an associated unit, containing the unit identifier, propagation trigger feature, content trigger feature, main trigger feature, and unit type.

[0047] Step S119: Collect all associated units corresponding to unit trigger nodes, sort the associated units according to the time sequence of instantaneous propagation segments, and form a set of comment associated units.

[0048] In this embodiment, all associated units corresponding to unit trigger nodes are collected, and then the associated units are sorted according to the time sequence of the instantaneous propagation segments corresponding to each associated unit. The sorting method is from early to late time. For example, the associated unit corresponding to the first instantaneous propagation segment is listed first, the second is listed later, and so on. All sorted associated units are stored in the same data set to form a comment associated unit set.

[0049] Step S1110: Verify the completeness of information for each associated unit. Each associated unit is required to include the corresponding unit trigger node identifier, propagation trigger feature, content trigger feature, subject trigger feature, unit type, and unit identifier. If there are associated units with missing information, return to the corresponding extraction step to extract the relevant information.

[0050] In this embodiment, each associated unit is traversed to check whether it contains unit trigger node identifier, propagation trigger feature, content trigger feature, main trigger feature, unit type and unit identifier. If any are missing, the process returns to the corresponding extraction step. For example, if the propagation trigger feature is missing, the process returns to step S114 to re-extract the propagation trigger feature of the unit trigger node until the information of all associated units is complete.

[0051] Step S120: Based on the associated units in the comment associated unit set, extract the unit association information of each associated unit, analyze the evolutionary influence relationship between the unit association information, calculate the degree of evolutionary influence between different associated units, determine the dynamic factor evolution type according to the degree of evolutionary influence, and integrate all dynamic factor evolution processes to form the comment associated dynamic factor evolution sequence.

[0052] In this embodiment, firstly, the propagation trigger features, content trigger features, and subject trigger features of each associated unit are extracted from the comment associated unit set. These three trigger features are used as the unit association information for each associated unit. The unit association information of each associated unit is a multi-dimensional feature set containing three trigger features. Then, any two associated units are selected, and the unit association information of the first associated unit is compared with that of the second associated unit. The associated features are identified, including information about consistent propagation speed trends, related text topics, and overlapping subject association networks. Specifically, the comparison method is as follows: for propagation speed change information in the propagation trigger features, if the propagation speed change trends of the two associated units are the same (e.g., both are increasing or both are decreasing), then the propagation speed change trends are considered consistent. For text topic change information in the content trigger features, if the text topics of the two associated units are related (e.g., belong to the same major category), then the text topics are considered related. For subject association network change information in the subject trigger features, if the associated subjects of the two associated units overlap, then the subject association networks are considered overlapping.

[0053] Step S121: For the identified associated features, analyze the evolutionary driving effect of the associated features of the first associated unit on the associated features of the second associated unit, and at the same time analyze the evolutionary feedback effect of the associated features of the second associated unit on the associated features of the first associated unit, and determine the evolutionary influence relationship between the two types of unit association information.

[0054] In this embodiment, for each identified associated feature, the evolutionary driving effect of the associated feature of the first associated unit on the associated feature of the second associated unit is analyzed. For example, if the propagation speed of the first associated unit shows an upward trend, and the propagation speed of the second associated unit also shows an upward trend, and the propagation speed change of the first associated unit occurs before the second, then it is determined that the propagation speed change of the first associated unit has an evolutionary driving effect on the propagation speed change of the second associated unit. At the same time, the evolutionary feedback effect of the associated feature of the second associated unit on the associated feature of the first associated unit is analyzed. For example, if the propagation speed change of the second associated unit further intensifies after the propagation speed change of the first associated unit, then it is determined that the propagation speed change of the second associated unit has an evolutionary feedback effect on the propagation speed change of the first associated unit. By analyzing the driving and feedback effects of all associated features, the evolutionary influence relationship between the two types of unit association information is determined. This evolutionary influence relationship is bidirectional, including influences in both driving and feedback directions.

[0055] Step S122: Based on the evolutionary influence relationship, count the number of evolutionary interactions between the two related units during the propagation process due to the related feature terms, and record the evolutionary delay and evolutionary range changes when each evolutionary interaction occurs.

[0056] In this embodiment, based on the determined evolutionary influence relationship, the number of evolutionary interactions between two associated units due to their associated feature items is counted throughout all instantaneous propagation periods. Each evolutionary interaction corresponds to one push or feedback action. Then, the evolutionary delay at each interaction is recorded. The evolutionary delay is the time interval from the change in the associated feature item of the first associated unit to the change in the associated feature item of the second associated unit. For example, if the change in the associated feature item of the first associated unit occurs at time T1 and the change occurs at time T2, then the evolutionary delay is T2-T1. Simultaneously, the evolutionary range change at each interaction is recorded. The evolutionary range change is the change in propagation range before and after the change in the associated feature item of the second associated unit. For example, if the propagation range before the interaction is S1 and after is S2, then the evolutionary range change is S2-S1.

[0057] Step S123: Combine the number of evolutionary interactions, evolutionary delay, and changes in evolutionary range to determine the degree of evolutionary influence between different related units. This degree of evolutionary influence is comprehensively reflected by the frequency of evolutionary interactions, the length of evolutionary delay, and the magnitude of changes in evolutionary range.

[0058] In this embodiment, firstly, the number of evolutionary interactions between any two associated units is counted throughout all instantaneous propagation periods. The ratio of the number of evolutionary interactions to the total number of instantaneous propagation periods is calculated to obtain the frequency of evolutionary interactions. The higher the frequency of evolutionary interactions, the more frequent the evolutionary association between the two associated units. Then, the time difference from the first associated unit to the second associated unit is recorded at each evolutionary interaction. All time differences are averaged to obtain the average evolutionary delay. The shorter the average evolutionary delay, the higher the evolutionary efficiency between the two associated units. Next, the evolutionary range information of the second associated unit before and after each evolutionary interaction is extracted. The absolute value of the difference between the evolutionary range after each interaction and the evolutionary range before each interaction is calculated. The absolute values ​​of all differences are summed to obtain the total change in evolutionary range. The larger the total change in evolutionary range, the greater the influence of the first associated unit on the evolutionary range of the second associated unit.

[0059] Step S1231: Normalize the frequency of evolutionary interactions, the average evolutionary delay, and the total change in the evolutionary range to obtain normalized values ​​for the frequency of evolutionary interactions, the average evolutionary delay, and the total change in the evolutionary range. The normalization of the average evolutionary delay is performed using a reciprocal form and scaled to a preset range to reflect that the shorter the average evolutionary delay, the greater the impact.

[0060] In this embodiment, the frequency of evolutionary interactions is normalized by dividing the frequency of evolutionary interactions by the maximum value of the evolutionary interaction frequencies of all associated unit pairs to obtain a normalized evolutionary interaction frequency value. This normalized frequency value falls within a preset range (e.g., [0, 1]), with higher frequencies resulting in larger values. The average evolutionary delay is also normalized by using a reciprocal form and scaling it to a preset range. Specifically, the maximum average evolutionary delay of all associated unit pairs is calculated, and then the reciprocal of the average evolutionary delay is divided by the reciprocal of this maximum value to obtain a normalized average evolutionary delay value. This normalized average evolutionary delay value falls within a preset range, with shorter delays resulting in larger values. Finally, the total change in the evolutionary range is normalized by dividing the total change in the evolutionary range by the maximum value of the total change in the evolutionary range of all associated unit pairs to obtain a normalized total change in the evolutionary range. This normalized total change in the evolutionary range falls within a preset range, with larger changes resulting in larger values.

[0061] Step S1232: Based on the relative importance of the frequency of evolutionary interactions, the average evolutionary delay, and the total change in the evolutionary range to the degree of influence on evolution, assign corresponding weights to the normalized values ​​of evolutionary interaction frequency, normalized average evolutionary delay, and normalized total change in the evolutionary range. The weight assignment is obtained based on training with historical data.

[0062] In this embodiment, for example, multiple sets of historical comment-related evolution data are first collected. Each set of historical comment-related evolution data includes the frequency of evolutionary interactions, the average evolutionary delay, the total change in the evolutionary range, and the corresponding actual evolutionary impact assessment results. The actual evolutionary impact assessment results are values ​​assessed by humans or experts based on the historical comment evolution, and the range is a preset interval (e.g., [0, 1]). Then, the frequency of evolutionary interactions, the average evolutionary delay, and the total change in the evolutionary range in each set of historical comment-related evolution data are standardized to eliminate the differences in the dimensions of different data dimensions. The standardization method is to subtract the mean of each data point and then divide it by its standard deviation. Next, the frequency of standardized evolutionary interactions, average evolutionary delay, and total change in evolutionary range were used as input variables, and the actual evolutionary impact assessment results were used as output variables to construct a weighted analysis model. This weighted analysis model is a linear regression model with the form Y=w1*X1+w2*X2+w3*X3+b, where Y is the actual evolutionary impact assessment result, X1 is the frequency of standardized evolutionary interactions, X2 is the average evolutionary delay after standardization, X3 is the total change in evolutionary range after standardization, w1, w2, and w3 are weight parameters, and b is a bias parameter.

[0063] Step S12321: Train the weighted analysis model using the collected historical comment association evolution data, adjust the parameters in the weighted analysis model so that the weighted analysis model can accurately output the corresponding actual evolutionary impact assessment results based on the input variables.

[0064] In this embodiment, the collected historical comment association evolution data is divided into a training set and a validation set. The training set is used to train the model, and the validation set is used to verify the model's performance. During training, the gradient descent algorithm is used to adjust the weight parameters w1, w2, w3 and the bias parameter b in the weight analysis model. The mean squared error function is used as the loss function. The mean squared error between the model's predicted output and the actual evolutionary impact assessment result is calculated. The parameters are iteratively adjusted to minimize the mean squared error until the model converges.

[0065] Step S12322: After training is completed, extract the coefficients corresponding to the input variables in the weight analysis model. These coefficients reflect the contribution of each input variable to the output variable.

[0066] In this embodiment, after the weighted analysis model is trained, the weight parameters w1, w2, and w3 corresponding to the input variables are extracted. These parameters represent the contribution of each input variable to the output variable. The value of w1 reflects the contribution of the frequency of evolutionary interaction to the degree of influence of evolution, w2 reflects the contribution of the average evolutionary delay, and w3 reflects the contribution of the total change in the evolutionary range.

[0067] Step S12323: Normalize the extracted coefficients so that the sum of the normalized coefficients is 1. The normalized coefficients are the weights corresponding to the frequency of evolutionary interaction, the average evolutionary delay, and the total change in the evolutionary range.

[0068] In this embodiment, the sum of the absolute values ​​of the extracted coefficients w1, w2, and w3 is calculated, and then each coefficient is divided by the sum of the absolute values ​​to obtain the normalized coefficients. For example, the normalized w1' = w1 / (|w1|+|w2|+|w3|), w2' = w2 / (|w1|+|w2|+|w3|), and w3' = w3 / (|w1|+|w2|+|w3|). These normalized coefficients are the weights corresponding to the frequency of evolutionary interaction, the average evolutionary delay, and the total change in the evolutionary range. The sum of the weights is 1.

[0069] Step S12324: Select some historical comment-related evolution data that did not participate in the training, multiply the standardized input variables by the corresponding weights and sum them to obtain the predicted value of the degree of evolutionary influence. Compare the predicted value of the degree of evolutionary influence with the actual evolutionary influence assessment results to generate a difference index. When the difference index is less than or equal to the difference threshold, it is determined that the weight allocation scheme meets the requirements.

[0070] In this embodiment, a portion of historical comment-related evolutionary data that was not used in training is selected as test data. The frequency of evolutionary interactions, average evolutionary latency, and total change in evolutionary range in the test data are standardized. Then, the standardized input variables are multiplied by their corresponding weights and summed to obtain the predicted value of the evolutionary impact. For example, the predicted value = w1'*X1' + w2'*X2' + w3'*X3', where X1', X2', and X3' are the standardized input variables of the test data. Then, the difference index between the predicted value and the actual evolutionary impact assessment result is calculated. The difference index uses the mean absolute error or root mean square error, for example, the mean absolute error = the average of |predicted value - actual value|. When the difference index is less than or equal to a preset difference threshold, the weight allocation scheme is deemed to meet the requirements; if it is greater than the threshold, the model needs to be retrained.

[0071] Step S12325: When the difference index is greater than the difference threshold, start the parameter optimization process of the weight analysis model, retrain the model and extract the weight parameters until the difference index is less than or equal to the difference threshold.

[0072] In this embodiment, when the difference index is greater than the difference threshold, the training parameters of the weight analysis model are adjusted, such as the learning rate and the number of iterations. Then, the model is retrained using the training set. After training, the weight parameters are extracted and normalized to obtain new weights. The difference index is then calculated using the test data. This process is repeated until the difference index is less than or equal to the difference threshold.

[0073] Step S12326: After the weight allocation scheme is verified, store the frequency weight parameter of evolution interaction, the average evolution delay weight parameter, and the total change in evolution range weight parameter into the weight parameter database.

[0074] In this embodiment, after the weight allocation scheme is verified, the frequency weight parameter w1' of evolutionary interaction, the average evolution delay weight parameter w2', and the total change in evolution range weight parameter w3' are stored in the weight parameter database. Each record in the database contains three weight parameters and corresponding verification information, such as difference index and training data identifier.

[0075] Step S12327: Establish a complete parameter traceability record in the weight parameter database, including training data identifier, data processing method, model configuration information, training parameter settings, verification results and final weight parameter values.

[0076] In this embodiment, a complete parameter traceability record is established for each weight parameter record in the weight parameter database. The record includes training data identifier (used to identify the training data used), data processing method (specific method of standardization processing), model configuration information (structural parameters of the weight analysis model), training parameter settings (learning rate, number of iterations, etc.), verification results (value of difference index), and final weight parameter values ​​(w1', w2', w3') for subsequent querying and tracing.

[0077] Step S1233: Multiply the normalized evolutionary interaction frequency value, the normalized average evolutionary delay value, and the normalized total change in evolutionary range value by their respective weights to obtain three weighted values.

[0078] In this embodiment, the normalized evolutionary interaction frequency value is multiplied by the evolutionary interaction frequency weight parameter to obtain the first weighted value; the normalized average evolutionary delay value is multiplied by the average evolutionary delay weight parameter to obtain the second weighted value; and the normalized total change in evolutionary range value is multiplied by the total change in evolutionary range weight parameter to obtain the third weighted value. The three weighted values ​​correspond to the contributions of the three features to the degree of influence on evolution, respectively.

[0079] Step S1234: Add the three weighted values ​​together to obtain a comprehensive score of the degree of evolutionary influence. The larger the comprehensive score, the higher the degree of evolutionary influence between different related units.

[0080] In this embodiment, the three weighted values ​​are added together to obtain a comprehensive score of the degree of evolutionary influence. For example, the comprehensive score = weighted value 1 + weighted value 2 + weighted value 3. The numerical range of the comprehensive score is a preset interval (such as [0, 1]). The larger the value, the higher the degree of evolutionary influence between different related units.

[0081] Step S124: Determine the dynamic factor evolution type based on the feature type corresponding to the degree of evolutionary influence. If the degree of evolutionary influence originates from the association of propagation triggering features, the dynamic factor evolution type is propagation triggering dynamic factor evolution; if the degree of evolutionary influence originates from the association of content triggering features, the dynamic factor evolution type is content triggering dynamic factor evolution; if the degree of evolutionary influence originates from the association of subject triggering features, the dynamic factor evolution type is subject triggering dynamic factor evolution.

[0082] In this embodiment, the source of the evolutionary influence is analyzed, that is, it is determined whether the association between propagation triggering features, content triggering features, or subject triggering features contributes the most to the evolutionary influence. The specific method is as follows: compare the magnitude of the three weighted values. If the first weighted value (corresponding to the propagation triggering feature) is the largest, it is determined that the evolutionary influence originates from the association of the propagation triggering feature, and the dynamic factor evolution type is propagation triggering dynamic factor evolution; if the second weighted value (corresponding to the content triggering feature) is the largest, it is determined that the association originates from the content triggering feature, and the dynamic factor evolution type is content triggering dynamic factor evolution; if the third weighted value (corresponding to the subject triggering feature) is the largest, it is determined that the association originates from the subject triggering feature, and the dynamic factor evolution type is subject triggering dynamic factor evolution.

[0083] Step S125: Assign a unique evolution identifier to each dynamic factor evolution type, bind the identifier of the associated unit, the description of the degree of evolutionary influence and the corresponding evolution identifier, and generate the dynamic factor evolution process corresponding to each evolutionary influence relationship.

[0084] In this embodiment, a unique evolution identifier is assigned to each dynamic factor evolution type. For example, the identifier for propagation-triggered dynamic factor evolution is "Evolve_Type_001", the identifier for content-triggered dynamic factor evolution is "Evolve_Type_002", and the identifier for subject-triggered dynamic factor evolution is "Evolve_Type_003". Then, the identifiers of the two associated units, the description of the degree of evolutionary influence (e.g., the comprehensive score value), and the corresponding evolution identifiers are bound and stored in the same data structure. Each data structure corresponds to a dynamic factor evolution process, including the identifiers of the two associated units, the description of the degree of evolutionary influence, the dynamic factor evolution type, and the evolution identifier.

[0085] Step S126: Collect all dynamic factor evolution processes, sort all collected dynamic factor evolution processes according to the chronological order of their evolution, and form a comment-related dynamic factor evolution sequence.

[0086] In this embodiment, all dynamic factor evolution processes are collected, and then sorted according to the chronological order of their occurrence. The sorting method is from earliest to latest, for example, the first evolution process is listed first, the second is listed later, and so on. All sorted dynamic factor evolution processes are stored in the same dataset to form a comment-related dynamic factor evolution sequence.

[0087] Step S127: Verify each dynamic factor evolution process in the comment-related dynamic factor evolution sequence. Each dynamic factor evolution process is required to include the corresponding two related unit identifiers, evolutionary influence relationship description, evolutionary influence degree description, dynamic factor evolution type and evolution identifier. If there are any dynamic factor evolution processes with missing information, return to the corresponding generation step to supplement and improve the relevant information.

[0088] In this embodiment, each dynamic factor evolution process is traversed to check whether it contains two associated unit identifiers, an evolutionary influence relationship description, an evolutionary influence degree description, a dynamic factor evolution type, and an evolution identifier. If any are missing, the process returns to the corresponding generation step. For example, if the evolutionary influence relationship description is missing, the process returns to step S121 to re-analyze the evolutionary influence relationship until the information of all dynamic factor evolution processes is complete.

[0089] Step S130: Take the associated units in the comment association unit set as topology nodes, use the dynamic factor evolution process in the comment association dynamic factor evolution sequence to determine the dynamic association strength parameters and dynamic association directions between each topology node, and connect each topology node according to the dynamic association strength parameters and dynamic association directions to form a comment dynamic association topology.

[0090] In this embodiment, all associated units are first extracted from the comment association unit set, and a unique topology node identifier is assigned to each associated unit, such as a unique identifier represented by a string, like "Node_001". Each associated unit is then treated as a topology node. Next, the comment association dynamic factor evolution sequence is invoked to extract the identifiers of the two topology nodes associated with each dynamic factor evolution process, the dynamic factor evolution type, and a description of the degree of evolutionary influence.

[0091] Step S131: Based on the evolution rules corresponding to the dynamic factor evolution type, the description of the degree of evolutionary influence is converted into a dynamic correlation strength parameter that can be quantified. This dynamic correlation strength parameter is positively correlated with the degree of evolutionary influence. The higher the degree of evolutionary influence, the larger the dynamic correlation strength parameter.

[0092] In this embodiment, the degree of evolutionary influence is described as a dynamic correlation strength parameter according to the evolutionary rules corresponding to the dynamic factor evolution type. Specifically, the rules are as follows: for propagation-triggered dynamic factor evolution, the higher the degree of evolutionary influence, the larger the dynamic correlation strength parameter. The conversion method is to multiply the comprehensive score of the degree of evolutionary influence by a preset propagation-triggered coefficient. For content-triggered dynamic factor evolution, the conversion method is to multiply the comprehensive score by a preset content-triggered coefficient. For subject-triggered dynamic factor evolution, the conversion method is to multiply the comprehensive score by a preset subject-triggered coefficient. The magnitudes of the three coefficients are obtained by training based on historical data, reflecting the degree of influence of different dynamic factor evolution types on the correlation strength. The numerical range of the dynamic correlation strength parameter is a preset interval (e.g., [0, 100]). The larger the value, the higher the correlation strength.

[0093] Step S132: Analyze the evolution time of the two topological nodes during the comment propagation process. If the evolution time value of the first topological node is less than that of the second topological node, then the dynamic association direction is determined to be from the first topological node to the second topological node; if the evolution time value of the second topological node is less than that of the first topological node, then the dynamic association direction is determined to be from the second topological node to the first topological node.

[0094] In this embodiment, the evolution time of two topological nodes during the comment propagation process is analyzed. The evolution time is the time when the unit triggering node corresponding to the topological node causes a propagation surge. If the evolution time of the first topological node is less than that of the second topological node, the dynamic association direction is from the first to the second; if the evolution time of the second is less than that of the first, the direction is from the second to the first. The dynamic association direction is represented by an arrow. For example, Node_001 to Node_002 means the direction is from Node_001 to Node_002.

[0095] Step S133: Create a dynamic connection record for each pair of topological nodes that have an evolutionary association. The dynamic connection record contains the identifiers of the two topological nodes, the dynamic association strength parameter, and the dynamic association direction.

[0096] In this embodiment, a dynamic connection relationship record is created for each pair of topological nodes with evolutionary association. Each record contains the identifiers of the two topological nodes, the dynamic association strength parameter, and the dynamic association direction. For example, the record content is "Node_001——Node_002, the strength parameter is a preset value, and the direction is from Node_001 to Node_002". All dynamic connection relationship records are stored in the same data set.

[0097] Step S134: Organize all topology nodes and their corresponding dynamic connection records according to the evolution time sequence of the topology nodes. First, place the topology nodes whose evolution time is earlier in the evolution time sequence of all topology nodes, and then connect the subsequent topology nodes in sequence according to the dynamic association direction.

[0098] In this embodiment, the topology nodes are sorted according to their evolution time, with those that evolved earlier placed first and those that evolved later placed last. Then, subsequent topology nodes are connected sequentially according to the dynamic association direction. For example, the first topology node has the earliest evolution time and is placed at the front. Then, the second topology node is connected according to the dynamic association direction, then the third, and so on, to form a preliminary topology structure.

[0099] Step S135: During the connection process, if multiple topology nodes point to the same subsequent topology node at the same time, the display priority of the connection is adjusted according to the dynamic association strength parameter between each topology node and the subsequent topology node. Connections with larger dynamic association strength parameters have higher display priority.

[0100] In this embodiment, during the connection process, if multiple topology nodes simultaneously point to the same subsequent topology node, for example, Node_001 and Node_002 simultaneously point to Node_003, the dynamic association strength parameter between Node_001 and Node_003 and the dynamic association strength parameter between Node_002 and Node_003 are compared. Connections with larger values ​​have higher display priority. For example, if the strength parameter between Node_001 and Node_003 is a preset value, and the same is true for Node_002 and Node_003, and the former is larger, then the connection from Node_001 to Node_003 has higher display priority and is displayed first in the topology structure.

[0101] Step S136: Perform structural processing on the organized topology nodes and dynamic connection relationships, eliminate duplicate connections between topology nodes, and retain the connection relationship with the largest dynamic association strength parameter value.

[0102] In this embodiment, the topology nodes and dynamic connection relationships after organization are structurally processed to check for duplicate connections, that is, whether there are multiple connection relationships between two topology nodes. If so, the dynamic association strength parameters of these connection relationships are compared, the connection relationship with the largest value is retained, and other duplicate connection relationships are deleted. For example, there are two connection relationships between Node_001 and Node_002, with the strength parameters being preset values ​​and preset values ​​respectively. In this case, the connection relationship with the largest value is retained, and the other one is deleted.

[0103] Step S137: Based on the changing trend of the dynamic association strength parameter in the dynamic factor evolution sequence of the comment association, dynamically adjust the processed topological nodes and connection relationships. If the dynamic association strength parameter changes during the subsequent evolution process, the dynamic association strength parameter of the corresponding connection relationship is updated synchronously.

[0104] In this embodiment, the dynamic correlation strength parameters of all dynamic factors in the dynamic factor evolution sequence are first extracted and arranged in chronological order to form a dynamic correlation strength parameter evolution sequence. Then, the changing trend of this evolution sequence is analyzed to identify the parameter rising interval, parameter falling interval, and parameter stable interval, and to determine the changing pattern of the dynamic correlation strength parameters in different time periods.

[0105] Step S1371: For the processed topological nodes and connections, establish a mapping relationship between each connection and the corresponding parameter in the dynamic association strength parameter evolution sequence, and clarify the parameter change period corresponding to each connection.

[0106] In this embodiment, for each processed connection, a mapping relationship is established between it and the corresponding parameter in the dynamic association strength parameter evolution sequence. For example, the connection Node_001 to Node_002 corresponds to the parameters P1, P2, and P3 in the evolution sequence, which correspond to different time periods. The parameter change period corresponding to each connection is clarified. For example, P1 corresponds to the time period T1-T2, P2 corresponds to the time period T2-T3, and so on.

[0107] Step S1372: Monitor the parameter changes in the subsequent evolution process of the dynamic association strength parameter evolution sequence in real time. If any parameter corresponding to any connection relationship is detected to change, record the values ​​before and after the parameter change.

[0108] In this embodiment, the parameter changes in the subsequent evolution process of the dynamic association strength parameter evolution sequence are monitored in real time. For example, the latest parameter value is read once every preset time interval. If the parameter corresponding to any connection relationship is detected to change, the value before the parameter change and the value after the change are recorded. For example, the value before the change is a preset value and the value after the change is a preset value.

[0109] Step S1373: Update the dynamic association strength parameter of the connection relationship according to the parameter change value. If the parameter value increases, increase the value of the dynamic association strength parameter of the corresponding connection relationship; if the parameter value decreases, decrease the value of the dynamic association strength parameter of the corresponding connection relationship.

[0110] In this embodiment, the dynamic association strength parameter of the connection relationship is updated according to the change value of the parameter. If the value after the change is greater than the value before the change, the value of the dynamic association strength parameter of the connection relationship is increased. For example, if the value before the change is a preset value and the value after the change is a preset value, then the value after the change is updated. If the value after the change is less than the value before the change, the parameter value is decreased. For example, if the value before the change is a preset value and the value after the change is a preset value, then the value after the change is updated.

[0111] Step S1374: While updating the parameters, check whether there are other connections between the two topological nodes associated with the connection relationship. If so, compare the updated dynamic association strength parameter with the dynamic association strength parameter of other connections, and retain the connection relationship with the largest dynamic association strength parameter value.

[0112] In this embodiment, while updating the parameters, it checks whether there are other connections between the two topological nodes associated with the connection relationship. If there are, for example, if there is another connection between Node_001 and Node_002 in addition to the current connection relationship, then the updated dynamic association strength parameter is compared with the parameter of the other connection relationship. The connection relationship with the largest value is retained, and the connection relationship with the smaller value is deleted. For example, if the updated parameter is a preset value and the other parameter is a preset value, then the connection relationship with the preset value is retained, and the other parameter is deleted.

[0113] Step S1375: Record the time of each dynamic adjustment, the identifier of the adjusted connection relationship, the values ​​of the dynamic correlation strength parameters before and after the adjustment, and the reason for the adjustment, to form a dynamic adjustment log.

[0114] In this embodiment, the time of each dynamic adjustment (specific date and time), the identifier of the connection relationship to be adjusted (e.g., Node_001-Node_002), the values ​​of the dynamic association strength parameters before and after the adjustment (e.g., preset values ​​before and after the adjustment), and the reason for the adjustment (e.g., parameter changes, values ​​increase or decrease) are recorded. The above information is stored in the dynamic adjustment log, and each log record corresponds to one dynamic adjustment.

[0115] Step S1376: Based on the dynamic adjustment log, analyze the impact of parameter changes on the topology. If parameter changes cause the topology to become unbalanced, perform a secondary adjustment on the relevant connections to keep the topology stable.

[0116] In this embodiment, the impact of dynamic adjustment log analysis parameter changes on the topology is analyzed. For example, the number of connections of each topology node in the topology changes after the statistical parameters change. If the number of connections of a certain topology node suddenly increases or decreases significantly, causing the topology to become unbalanced, a secondary adjustment is made to the relevant connection relationships. For example, the display priority of the connection relationships is adjusted or some connection relationships are deleted to keep the topology stable.

[0117] Step S1377: After completing all dynamic adjustments, generate an adjusted topology document, which contains all topology node identifiers, connection relationship identifiers, current dynamic association strength parameters, and dynamic association directions.

[0118] In this embodiment, after all dynamic adjustments are completed, an adjusted topology document is generated. The document contains all topology node identifiers (e.g., Node_001, Node_002, etc.), connection relationship identifiers (e.g., Node_001-Node_002), current dynamic association strength parameters (e.g., preset values), and dynamic association directions (e.g., from Node_001 to Node_002). The document is then stored in a specified storage location.

[0119] Step S1378: Compare the adjusted topology document with the original topology document, record the structural differences before and after the adjustment, and form a topology adjustment comparison table.

[0120] In this embodiment, the adjusted topology document is compared with the one before the adjustment, and the structural differences before and after the adjustment are recorded, such as changes in the number of connections of topology nodes, changes in dynamic association strength parameters, and changes in dynamic association directions, etc., to form a topology adjustment comparison table. Each row in the table corresponds to a topology node or connection relationship, and records the difference information before and after the adjustment.

[0121] Step S138: Integrate the dynamically adjusted topology nodes and connections into a dynamic mesh structure to form a dynamic relational topology for comments.

[0122] In this embodiment, all dynamically adjusted topology nodes and connection relationships are integrated to form a dynamic mesh structure. The topology nodes are the nodes in the mesh structure, the connection relationships are the edges between the nodes, the weight of the edge is the dynamic association strength parameter, and the direction of the edge is the dynamic association direction. This dynamic mesh structure is the comment dynamic association topology.

[0123] Step S139: Detect whether each topology node has a corresponding dynamic connection relationship when there are associated nodes. Each dynamic connection relationship includes a complete topology node identifier, dynamic association strength parameters, and dynamic association direction. If a structural missing is detected, an adjustment process is automatically triggered to supplement and improve it.

[0124] In this embodiment, it is detected whether each topological node has a corresponding dynamic connection relationship when there are associated nodes. That is, if a topological node has an evolutionary association with other topological nodes, it is checked whether there is a corresponding connection relationship. At the same time, it is checked whether each dynamic connection relationship contains a complete topological node identifier, dynamic association strength parameters, and dynamic association direction. If a structural missing is detected, such as a topological node having associated nodes but no connection relationship, an adjustment process is automatically triggered to re-extract the dynamic factor evolution process of the associated node, determine the dynamic association strength parameters and direction, and supplement the corresponding connection relationship until the connection relationship of all topological nodes is complete.

[0125] Step S140: Analyze the topological node distribution, the evolution law of dynamic association strength parameters between nodes, and the propagation evolution flow corresponding to the dynamic association direction in the dynamic association topology. Integrate the topological node distribution, the evolution law of dynamic association strength parameters, and the propagation evolution flow into the comment association evolution situation, and transmit the comment association evolution situation to the comment analysis terminal.

[0126] In this embodiment, all topological nodes in the dynamic association topology of comments are traversed first, and the evolution time, evolution subject, and evolution content summary information of each topological node are recorded. The evolution subject is the anonymous identifier of the unit trigger node, and the evolution content summary information is a brief description of the comment content.

[0127] Step S141: Based on the evolution time, the topology nodes are divided into multiple evolution time interval groups. Each evolution time interval group contains all the topology nodes within that time interval. The number of topology nodes in each evolution time interval group is counted to form the topology node time distribution statistics.

[0128] In this embodiment, the evolution time information of all topological nodes is first extracted to determine the earliest and latest evolution times, for example, the earliest is T1 and the latest is T2. Then, the difference between the earliest and latest evolution times is calculated, and multiple evolution time intervals of equal duration are divided according to the size of the difference. Each evolution time interval has the same duration. For example, if the difference is T2-T1=preset duration, a preset number of evolution time intervals with a preset duration are divided, namely [T1, T1+preset value], [T1+preset value, T1+2*preset value], ..., [T1+(preset number-1)*preset value, T2].

[0129] Step S1411: Assign a unique interval identifier to each evolution time interval and record the start and end evolution times of each evolution time interval.

[0130] In this embodiment, a unique interval identifier is assigned to each evolution time interval, for example, the combination of the start time and end time of the interval is used as the identifier, such as [T1, T1+preset value]. Then, the start evolution time and end evolution time of each evolution time interval are recorded. For example, the start of the interval [T1, T1+preset value] is T1, and the end is T1+preset value.

[0131] Step S1412: Traverse each topology node, compare the evolution time of the topology node with the start evolution time and end evolution time of each evolution time interval, and determine the evolution time interval to which the evolution time of the topology node belongs.

[0132] In this embodiment, each topology node is traversed, and its evolution time is compared with the start and end times of each evolution time interval. If the evolution time is greater than or equal to the start time and less than or equal to the end time, the evolution time of the topology node is determined to belong to the evolution time interval. For example, if the evolution time of the topology node is T1 + preset value / 2, then it belongs to the interval [T1, T1 + preset value].

[0133] Step S1413: Assign the topology nodes to the evolution time interval group corresponding to their respective evolution time intervals. Each evolution time interval group contains all topology nodes whose evolution time belongs to that evolution time interval.

[0134] In this embodiment, each topology node is assigned to the evolution time interval group corresponding to its evolution time interval. For example, the topology node with an evolution time of T1 + preset value / 2 is assigned to the evolution time interval group corresponding to the interval [T1, T1 + preset value]. Each evolution time interval group contains all the topology nodes in that interval.

[0135] Step S1414: For each evolution time interval group, count the number of topological nodes within that evolution time interval group and record the counted value.

[0136] In this embodiment, for each evolution time interval group, the number of topological nodes contained therein is counted. For example, if the evolution time interval group corresponding to the interval [T1, T1+preset value] contains a preset number of topological nodes, then the number value is the preset number, and the number value of each evolution time interval group is recorded.

[0137] Step S1415: Organize the interval identifier, start evolution time, end evolution time and corresponding number of topological nodes for each evolution time interval group into a table, and sort the contents of the table according to the order of the evolution time intervals to form the statistical results of the time distribution of topological nodes.

[0138] In this embodiment, the interval identifier, start evolution time, end evolution time, and number of topological nodes for each evolution time interval group are organized into a table. The columns of the table include the interval identifier, start time, end time, and number of nodes. Then, the table contents are sorted according to the order of the evolution time intervals, for example, sorted from smallest to largest by start time, to form the statistical results of the time distribution of topological nodes.

[0139] Step S1416: Compare the statistical results of the time distribution of topological nodes with the evolution time information of all topological nodes, and determine whether the total number of topological nodes in the statistical results is consistent with the actual total number of topological nodes extracted. If there is a difference in the number, repeat the division and statistical steps.

[0140] In this embodiment, the number of nodes in all evolution time interval groups in the topology node time distribution statistics results are added together to obtain the total number of topology nodes in the statistics results. At the same time, the total number of topology nodes actually extracted is counted. If the two are consistent, the result is correct. If there is a difference in the number, for example, the statistical result is a preset value and the actual value is a preset value, the division and statistics steps are re-executed to check whether there are any topology nodes that have not been assigned to any evolution time interval group or have been assigned incorrectly, until the number is consistent.

[0141] Step S142: Based on the domain to which the evolutionary subject belongs, the topological nodes are divided into multiple evolutionary domain groups. Each evolutionary domain group contains all the topological nodes in that evolutionary domain. The number of topological nodes in each evolutionary domain group is counted to form the topological node domain distribution statistics.

[0142] In this embodiment, the domain classification of the evolving subject is first determined, such as political, economic, social, and cultural domains. Then, a domain is assigned to the evolving subject (unit trigger node) of each topological node. For example, if the evolution content of a unit trigger node involves the economy, then its domain is the economic domain. Next, the topological nodes are divided into multiple evolution domain groups according to their domains. Each evolution domain group contains all the topological nodes within that domain. For example, the economic domain group contains all topological nodes whose evolving subjects belong to the economic domain. Then, the number of topological nodes in each evolution domain group is counted. For example, the economic domain group contains a preset number of topological nodes. The domain identifier and the number of topological nodes in each evolution domain group are compiled into a table to form the topological node domain distribution statistics.

[0143] Step S143: Integrate the statistical results of the time distribution of topology nodes with the statistical results of the domain distribution of topology nodes to obtain the distribution of topology nodes.

[0144] In this embodiment, the statistical results of the time distribution of topological nodes are integrated with the statistical results of the domain distribution of topological nodes. For example, based on the statistical results of the time distribution, the number of topological nodes in each domain within each evolution time interval group is added to form a more detailed distribution of topological nodes. For example, the number of political domain nodes in the interval [T1, T1+preset value] is a preset value, the number of economic domain nodes is a preset value, and the number of social domain nodes is a preset value. The above information is integrated to obtain the distribution of topological nodes.

[0145] Step S144: Extract the dynamic association strength parameters of all dynamic connections in the dynamic association topology of the comments. Arrange the dynamic association strength parameters into an evolution sequence according to the order of evolution time. Analyze the increase and decrease of the dynamic association strength parameters in the evolution sequence to determine the rising stage, falling stage and stable stage of the dynamic association strength parameters in different evolution stages, and form the evolution law of the dynamic association strength parameters between nodes.

[0146] In this embodiment, the dynamic association strength parameters of all dynamic connections in the dynamic association topology of comments are first extracted. Then, the parameters are arranged into an evolutionary sequence according to the evolution time of each connection. For example, the parameter of the first connection is P1, the second is P2, and so on, forming the sequence [P1, P2, P3, ..., Pn]. Then, the increase and decrease of parameters in the evolutionary sequence are analyzed. For example, the parameter gradually increases from P1 to P5, which is determined to be the rising stage; the parameter gradually decreases from P5 to P10, which is determined to be the falling stage; and the parameter remains unchanged from P10 to P15, which is determined to be the stable stage. By analyzing the changes of all parameters, the rising, falling, and stable stages of the dynamic association strength parameters in different evolutionary stages are determined, forming the evolution law of the dynamic association strength parameters between nodes.

[0147] Step S145: Traverse all dynamic association directions in the comment dynamic association topology, count the number of topology nodes and the number of dynamic connections in each dynamic association direction, and determine the main direction of propagation and evolution. If the number of topology nodes and the number of dynamic connections in any direction are greater than those in other directions, then that direction is the main direction of propagation and evolution.

[0148] In this embodiment, all dynamic association directions are traversed, and the number of topological nodes and the number of dynamic connections in each direction are counted. For example, the number of topological nodes and the number of connections in direction A are preset values; the number of nodes and the number of connections in direction B are preset values; and the number of nodes and the number of connections in direction C are preset values. Then, the number of nodes and the number of connections in direction A are both greater than those in other directions, and direction A is determined to be the main propagation and evolution direction.

[0149] Step S146: Analyze the domain coverage of the main propagation and evolution flow, count the topological node proportion of each domain in the main propagation and evolution flow, and determine the domain penetration of the main propagation and evolution flow.

[0150] In this embodiment, the scope of the domains covered by the main propagation and evolution flow is analyzed, that is, the domains to which the topological nodes in the flow belong. Then, the proportion of topological nodes in each domain is counted. For example, the proportion of the topological nodes in the main propagation and evolution flow belonging to the political domain is a preset value, the economic domain is a preset value, and the social domain is a preset value. The domain penetration of the main propagation and evolution flow is determined, that is, it mainly penetrates the political domain, followed by the economic and social domains.

[0151] Step S147: Integrate the main propagation evolution flow and domain penetration to obtain the propagation evolution flow corresponding to the dynamic correlation direction.

[0152] In this embodiment, the main propagation evolution flow and domain penetration are integrated. For example, the main propagation evolution flow is direction A, and the domain penetration is a preset value for the proportion of the political domain, a preset value for the economic domain, and a preset value for the social domain. After integration, the propagation evolution flow corresponding to the dynamic associated direction is obtained, which includes the main direction and domain penetration information.

[0153] Step S148: Organize the topological node distribution, the evolution law of dynamic association strength parameters between nodes, and the propagation evolution flow according to a preset structured format to form a comment association evolution status text. This structured format includes a distribution statistics module, an intensity evolution module, and a flow analysis module. The distribution statistics module corresponds to the analysis results of the topological node distribution, the intensity evolution module corresponds to the analysis results of the evolution law of dynamic association strength parameters between nodes, and the flow analysis module corresponds to the analysis results of the propagation evolution flow.

[0154] In this embodiment, the distribution, evolution, and propagation flow of topological nodes are organized according to a preset structured format. This structured format includes three modules: a distribution statistics module, an intensity evolution module, and a flow analysis module. The distribution statistics module contains statistical results of the temporal and domain distribution of topological nodes, presented in tabular form. The intensity evolution module contains the evolution sequence of dynamic correlation strength parameters and the division of each stage, presented in line graph form (the trend of the line graph is described in the text). The flow analysis module contains the main propagation flow and domain penetration, presented in textual description. The contents of these modules are integrated to form a comment correlation evolution trend text, with the text describing the information of each module in detail.

[0155] Step S149: Transmit the text related to the evolution of the comment to the comment analysis terminal.

[0156] In this embodiment, the generated text of the comment association evolution is transmitted to the comment analysis terminal via the network. During the transmission process, encryption technology is used to protect the security of the text content, ensuring that the text can reach the comment analysis terminal accurately and securely for comment analysts to view and analyze.

[0157] Figure 2 This application illustrates an information association analysis system 100 based on big data of online comments, including a processor 1001, a memory 1003, and program code stored in the memory 1003. The processor 1001 executes the program code to implement the steps of the information association analysis method based on big data of online comments.

[0158] Figure 2 The illustrated information association analysis system 100 based on big data of online comments includes a processor 1001 and a memory 1003. The processor 1001 and the memory 1003 are connected, for example, via a bus 1002. Optionally, the information association analysis system 100 based on big data of online comments may further include a transceiver 1004, which can be used for data interaction between this information association analysis system and other information association analysis systems based on big data of online comments, such as sending and / or receiving data. It should be noted that in actual scheduling, the transceiver 1004 is not limited to one, and the structure of this information association analysis system 100 based on big data of online comments does not constitute a limitation on the embodiments of this application.

[0159] The memory 1003 is used to store program code for executing the embodiments of this application, and its execution is controlled by the processor 1001. The processor 1001 is used to execute the program code stored in the memory 1003 to implement the steps shown in the foregoing method embodiments.

[0160] This application provides a computer-readable storage medium storing program code, which, when executed by a processor, can implement the steps and corresponding content of the aforementioned method embodiments.

[0161] It should be understood that although arrows indicate various operation steps in the flowcharts of the embodiments of this application, the order in which these steps are implemented is not limited to the order indicated by the arrows. Unless explicitly stated herein, in some implementation scenarios of the embodiments of this application, the implementation steps in each flowchart may be executed in other orders based on requirements. Furthermore, some or all steps in each flowchart may include multiple sub-steps or multiple stages depending on the actual implementation scenario. Some or all of these sub-steps or stages may be executed at the same time, and each sub-step or stage may also be executed at different times. In scenarios where execution times differ, the execution order of these sub-steps or stages can be flexibly configured based on requirements, and the embodiments of this application do not limit this.

[0162] The above description is only an optional implementation method for some implementation scenarios of this application. It should be noted that for those skilled in the art, other similar implementation methods based on the technical concept of this application, without departing from the technical concept of this application, also fall within the protection scope of the embodiments of this application.

Claims

1. A method for information correlation analysis based on big data of online comments, characterized in that, The method includes: Extracting instantaneous propagation segments from online comment big data, identifying unit trigger nodes in these instantaneous propagation segments, extracting propagation trigger features, content trigger features, and subject trigger features from these unit trigger nodes, determining the unit type based on the extracted trigger features, generating associated units corresponding to each unit trigger node, and integrating all associated units to form a set of comment associated units; Based on the associated units in the comment associated unit set, the unit association information of each associated unit is extracted, the evolutionary influence relationship between the unit association information is analyzed, the degree of evolutionary influence between different associated units is calculated, the dynamic factor evolution type is determined according to the degree of evolutionary influence, and all dynamic factor evolution processes are integrated to form the comment associated dynamic factor evolution sequence. The associated units in the set of associated units in the comment association are used as topology nodes. The dynamic factor evolution process in the dynamic factor evolution sequence of the comment association is used to determine the dynamic association strength parameters and dynamic association direction between each topology node. The topology nodes are connected according to the dynamic association strength parameters and dynamic association direction to form the comment dynamic association topology. The analysis examines the distribution of topological nodes, the evolution of dynamic association strength parameters between nodes, and the propagation flow corresponding to the dynamic association direction in the dynamic association topology of comments. The topological node distribution, the evolution of dynamic association strength parameters, and the propagation flow are integrated into the comment association evolution trend, and the comment association evolution trend is transmitted to the comment analysis terminal. The process of using the associated units in the comment association unit set as topology nodes, determining the dynamic association strength parameters and dynamic association directions between each topology node using the dynamic factor evolution process in the comment association dynamic factor evolution sequence, and connecting each topology node according to the dynamic association strength parameters and dynamic association directions to form a comment dynamic association topology includes: Extract all associated units from the set of associated comment units, assign a unique topology node identifier to each associated unit, and treat each associated unit as a topology node; The dynamic factor evolution sequence associated with the comments is invoked, and the identifiers of the two topological nodes associated with each dynamic factor evolution process, the dynamic factor evolution type, and the degree of evolution impact are extracted. Based on the evolution rules corresponding to the dynamic factor evolution type, the description of the degree of evolutionary influence is converted into a dynamic correlation strength parameter that can be quantified. The dynamic correlation strength parameter is positively correlated with the degree of evolutionary influence; the higher the degree of evolutionary influence, the larger the dynamic correlation strength parameter. Analyze the evolution time of two topological nodes during the comment propagation process. If the evolution time value of the first topological node is less than that of the second topological node, then the dynamic association direction is determined to be from the first topological node to the second topological node; if the evolution time value of the second topological node is less than that of the first topological node, then the dynamic association direction is determined to be from the second topological node to the first topological node. For each pair of topological nodes with evolutionary association, a dynamic connection record is created, which includes the identifiers of the two topological nodes, the dynamic association strength parameter, and the dynamic association direction. Organize all topology nodes and their corresponding dynamic connection records according to the evolution time sequence of the topology nodes. First, place the topology nodes whose evolution time is earlier in the evolution time sequence of all topology nodes, and then connect the subsequent topology nodes in sequence according to the dynamic association direction. During the connection process, if multiple topology nodes point to the same subsequent topology node at the same time, the display priority of the connection is adjusted according to the dynamic association strength parameter between each topology node and the subsequent topology node. Connections with larger dynamic association strength parameters have higher display priority. Structural processing is performed on the reorganized topology nodes and dynamic connection relationships to eliminate duplicate connections between topology nodes and retain the connection relationship with the largest dynamic association strength parameter value. Based on the changing trend of dynamic association strength parameters in the dynamic factor evolution sequence of comment association, the processed topological nodes and connection relationships are dynamically adjusted. If the dynamic association strength parameters change in the subsequent evolution process, the dynamic association strength parameters of the corresponding connection relationships are updated synchronously. The dynamically adjusted topology nodes and connections are integrated into a dynamic mesh structure to form a dynamic relational topology for comments. The system checks whether each topology node has a corresponding dynamic connection relationship when there are associated nodes. Each dynamic connection relationship includes a complete topology node identifier, dynamic association strength parameters, and dynamic association direction. If a structural missing structure is detected, an adjustment process is automatically triggered to supplement and improve it.

2. The information correlation analysis method based on big data of online comments according to claim 1, characterized in that, The process involves extracting instantaneous propagation segments from big data of online comments, identifying unit trigger nodes within these segments, extracting propagation trigger features, content trigger features, and subject trigger features from these unit trigger nodes, determining the unit type based on the extracted trigger features, generating associated units corresponding to each unit trigger node, and integrating all associated units to form a set of comment associated units, including: The system filters comment data containing information about a sudden surge in dissemination from big data on online comments. This comment data includes records of a surge in the posting, forwarding, and receiving of comment content within a short period of time. Based on the timestamp sequence in the instantaneous propagation surge information, the comment data containing the instantaneous propagation surge information is divided into multiple consecutive instantaneous propagation periods. Each instantaneous propagation period corresponds to a segment of instantaneous comment propagation process, forming instantaneous propagation segments in multiple comment propagation processes. Traverse each instantaneous propagation segment and identify the entity that triggers the propagation surge in the instantaneous propagation segment. The entity that triggers the propagation surge includes the entity that publishes the surge, the entity that forwards the surge, and the entity that receives the surge. Each entity that triggers the propagation surge is regarded as a unit trigger node in the instantaneous propagation segment. For each unit trigger node, extract the information on the change in propagation speed, the expansion of propagation range, and the increase in propagation levels when the unit trigger node causes a propagation surge. Integrate the information on the change in propagation speed, the expansion of propagation range, and the increase in propagation levels into the propagation triggering features of the unit trigger node. Extract the text topic change information, text sentiment change information, and text keyword density change information of the comment content corresponding to the surge in dissemination caused by each unit trigger node, and integrate the text topic change information, text sentiment change information, and text keyword density change information into the content trigger feature of the unit trigger node; Extract information on changes in subject activity, subject association network, and subject propagation habits when each unit triggers a surge in propagation. Integrate these information into the subject triggering features of the unit triggering node. Collect the propagation trigger features, content trigger features, and subject trigger features of all unit trigger nodes. Classify the collected propagation trigger features, content trigger features, and subject trigger features of all unit trigger nodes. Determine the unit type based on the unit trigger attributes reflected by the trigger features. The unit type includes propagation trigger type units, content trigger type units, and subject trigger type units. Assign a unique unit identifier to each unit trigger node, and bind the propagation trigger characteristics, content trigger characteristics, and main trigger characteristics of the unit trigger node with the corresponding unit identifier and unit type to generate the associated unit corresponding to each unit trigger node; Collect all associated units corresponding to unit trigger nodes, sort the associated units according to the time sequence of instantaneous propagation segments, and form a set of comment associated units; Verify the completeness of information for each associated unit. Each associated unit must include the corresponding unit trigger node identifier, propagation trigger feature, content trigger feature, subject trigger feature, unit type, and unit identifier. If there are associated units with missing information, return to the corresponding extraction step to extract the relevant information.

3. The information correlation analysis method based on big data of online comments according to claim 1, characterized in that, The process involves extracting unit association information from the set of associated comment units, analyzing the evolutionary influence relationships between these units, calculating the degree of evolutionary influence between different associated units, determining the dynamic factor evolution type based on the degree of influence, and integrating all dynamic factor evolution processes to form a comment-related dynamic factor evolution sequence, including: Extract the propagation trigger features, content trigger features, and subject trigger features of each associated unit from the set of associated comment units, and use the three trigger features of each associated unit as the unit association information of that associated unit; Select any two associated units, compare the unit association information of the first associated unit with the unit association information of the second associated unit, and identify the features that are associated between the two types of unit association information. The features that are associated include information that the propagation speed changes in a consistent trend, information that is related to the text topic, and information that the main association network overlaps. For the identified association features, we analyze the evolutionary driving effect of the association features of the first association unit on the association features of the second association unit, and at the same time analyze the evolutionary feedback effect of the association features of the second association unit on the association features of the first association unit, so as to determine the evolutionary influence relationship between the association information of the two types of units. Based on the evolutionary influence relationship, the number of evolutionary interactions between two related units due to the related feature terms during the propagation process is counted, and the evolutionary delay and evolutionary range changes at the time of each evolutionary interaction are recorded. By combining the number of evolutionary interactions, evolutionary delays, and changes in evolutionary range, the degree of evolutionary influence between different related units is determined. The degree of evolutionary influence is comprehensively reflected by the frequency of evolutionary interactions, the length of evolutionary delays, and the magnitude of changes in evolutionary range. Based on the characteristic type corresponding to the degree of evolutionary influence, the dynamic factor evolution type is determined. If the degree of evolutionary influence originates from the correlation of propagation triggering features, the dynamic factor evolution type is propagation triggering dynamic factor evolution; if the degree of evolutionary influence originates from the correlation of content triggering features, the dynamic factor evolution type is content triggering dynamic factor evolution; if the degree of evolutionary influence originates from the correlation of subject triggering features, the dynamic factor evolution type is subject triggering dynamic factor evolution. A unique evolution identifier is assigned to each type of dynamic factor evolution. The identifiers of associated units, the description of the degree of evolutionary influence, and the corresponding evolution identifiers are bound together to generate the dynamic factor evolution process corresponding to each evolutionary influence relationship. Collect all dynamic factor evolution processes, sort all collected dynamic factor evolution processes according to the chronological order of their occurrence, and form a comment-related dynamic factor evolution sequence; Verify the evolution process of each dynamic factor in the dynamic factor evolution sequence associated with the comment. Each dynamic factor evolution process is required to include the corresponding two associated unit identifiers, the description of the evolutionary influence relationship, the description of the degree of evolutionary influence, the dynamic factor evolution type, and the evolution identifier. If there are any dynamic factor evolution processes with missing information, return to the corresponding generation step to supplement and improve the relevant information.

4. The information correlation analysis method based on big data of online comments according to claim 1, characterized in that, The analysis analyzes the topological node distribution, the evolution of dynamic association strength parameters between nodes, and the propagation flow direction corresponding to the dynamic association direction in the dynamic association topology. It integrates the topological node distribution, the evolution of dynamic association strength parameters, and the propagation flow direction into a comment association evolution trend, and transmits this trend to the comment analysis terminal, including: Traverse all topological nodes in the dynamic association topology of comments, and record the evolution time, evolution subject, and evolution content summary information of each topological node; Based on the evolution time, the topological nodes are divided into multiple evolution time interval groups. Each evolution time interval group contains all the topological nodes within that time interval. The number of topological nodes in each evolution time interval group is counted to form the statistical results of the time distribution of topological nodes. Based on the domain to which the evolutionary subject belongs, the topological nodes are divided into multiple evolutionary domain groups. Each evolutionary domain group contains all the topological nodes in that evolutionary domain. The number of topological nodes in each evolutionary domain group is counted to form the statistical results of the domain distribution of topological nodes. By integrating the statistical results of the temporal distribution of topology nodes with the statistical results of the domain distribution of topology nodes, the distribution of topology nodes is obtained; Extract the dynamic association strength parameters of all dynamic connections in the dynamic association topology of comments, arrange the dynamic association strength parameters into an evolution sequence according to the evolution time, analyze the increase and decrease of the dynamic association strength parameters in the evolution sequence, determine the rising stage, falling stage and stable stage of the dynamic association strength parameters in different evolution stages, and form the evolution law of dynamic association strength parameters between nodes. Traverse all dynamic association directions in the dynamic association topology of the comments, count the number of topology nodes and the number of dynamic connections in each dynamic association direction, and determine the main direction of propagation and evolution. If the number of topology nodes and the number of dynamic connections in any one direction are greater than those in other directions, then that direction is the main direction of propagation and evolution. Analyze the domain coverage of the main propagation and evolution flow, count the topological node proportion of each domain in the main propagation and evolution flow, and determine the domain penetration of the main propagation and evolution flow. By integrating the main propagation evolution trends and domain penetration, we can obtain the propagation evolution trends corresponding to the dynamic correlation directions; The topological node distribution, the evolution of dynamic association strength parameters between nodes, and the propagation evolution flow are organized according to a preset structured format to form a comment association evolution status text. The structured format includes a distribution statistics module, an intensity evolution module, and a flow analysis module. The distribution statistics module corresponds to the analysis results of the topological node distribution, the intensity evolution module corresponds to the analysis results of the evolution of dynamic association strength parameters between nodes, and the flow analysis module corresponds to the analysis results of the propagation evolution flow. The text describing the evolution of comments is transmitted to the comment analysis terminal.

5. The information correlation analysis method based on big data of online comments according to claim 2, characterized in that, The process involves dividing the comment data containing the instantaneous propagation surge information into multiple consecutive instantaneous propagation periods based on the timestamp sequence in the instantaneous propagation surge information. Each instantaneous propagation period corresponds to a segment of instantaneous comment propagation, forming multiple instantaneous propagation segments within the comment propagation process, including: Extract all timestamps from comment data containing information about a sudden surge in dissemination, where the timestamps record the times of surges in posting, forwarding, and receiving of comment data. All timestamps in the extracted comment data containing information about the instantaneous spread of information are sorted according to their numerical values ​​to obtain an ordered timestamp sequence; Calculate the difference between two adjacent timestamps in an ordered timestamp sequence and record all adjacent timestamp differences; Analyze the distribution of differences between adjacent timestamps, find the boundary point where the difference value is greater than other difference values, and the difference value corresponding to the boundary point can distinguish different instantaneous propagation periods; Using the dividing point as the boundary, the ordered timestamp sequence is divided into multiple consecutive time subsequences, each time subsequence corresponding to an instantaneous propagation period; For each time subsequence, all comment data within that time subsequence is extracted. The comment data includes records of posting surges, forwarding surges, and receiving surges during that instantaneous propagation period. The comment data corresponding to each time subsequence is bound to the identifier of that instantaneous propagation period to form an instantaneous propagation segment corresponding to each instantaneous propagation period; Verify the integrity of comment data in each instantaneous propagation segment, requiring each instantaneous propagation segment to include all relevant surge records of posting, forwarding, and receiving during that instantaneous propagation period; The verified instantaneous propagation segments are numbered, and consecutive numbers are assigned to the instantaneous propagation segments according to the time sequence of the instantaneous propagation period. A mapping table is established between the instantaneous propagation segment numbers and the corresponding instantaneous propagation period identifiers, recording the instantaneous propagation period identifier, time subsequence range, and the type of comment data contained in each number.

6. The information association analysis method based on big data of online comments according to claim 3, characterized in that, The determination of the degree of evolutionary influence between different associated units by combining the number of evolutionary interactions, evolutionary delay, and changes in evolutionary range includes: The number of evolutionary interactions between any two associated units is counted in all instantaneous propagation periods. The ratio of the number of evolutionary interactions to the total number of instantaneous propagation periods is calculated to obtain the frequency of evolutionary interactions. The higher the frequency of evolutionary interactions, the more frequent the evolutionary association between the two associated units. Record the time difference from the first associated unit to the second associated unit when each evolutionary interaction occurs, and average all time differences to obtain the average evolutionary delay. The shorter the average evolutionary delay, the higher the evolutionary efficiency between the two associated units. Extract the evolution range information of the second associated unit before and after each evolution interaction, calculate the absolute value of the difference between the evolution range after each evolution interaction and the evolution range before the evolution interaction, sum all the absolute values ​​of the differences to obtain the total change in evolution range. The larger the total change in evolution range, the greater the influence of the first associated unit on the evolution range of the second associated unit. The frequency of evolutionary interactions, the average evolutionary delay, and the total change in evolutionary range were normalized to obtain normalized values ​​for the frequency of evolutionary interactions, the average evolutionary delay, and the total change in evolutionary range. The normalization of the average evolutionary delay was performed by taking the inverse form and scaling it to [0, 1] to reflect that the shorter the average evolutionary delay, the greater the impact. Based on the relative importance of the frequency of evolutionary interactions, the average evolutionary delay, and the total change in the evolutionary range to the degree of influence on evolution, corresponding weights are assigned to the normalized values ​​of evolutionary interaction frequency, normalized average evolutionary delay, and normalized total change in the evolutionary range. The weight assignment is obtained based on training with historical data. The normalized evolutionary interaction frequency, the normalized average evolutionary delay, and the normalized total change in evolutionary range are multiplied by their respective weights to obtain three weighted values. The three weighted values ​​are added together to obtain a comprehensive score of the degree of evolutionary influence. The higher the comprehensive score, the greater the degree of evolutionary influence between different related units.

7. The information correlation analysis method based on big data of online comments according to claim 1, characterized in that, The process involves dynamically adjusting the processed topological nodes and connections based on the changing trend of the dynamic association strength parameter in the dynamic factor evolution sequence of the comment association. If the dynamic association strength parameter changes during subsequent evolution, the dynamic association strength parameter of the corresponding connection is updated synchronously, including: Extract the dynamic association strength parameters of all dynamic factors in the dynamic factor evolution sequence of comment association, arrange them in chronological order of evolution time, and form a dynamic association strength parameter evolution sequence; Analyze the changing trend of the dynamic correlation strength parameter evolution sequence, identify the parameter rising interval, parameter falling interval and parameter stable interval, and determine the changing pattern of the dynamic correlation strength parameter in different time periods; For the processed topological nodes and connections, establish a mapping relationship between each connection and the corresponding parameters in the evolution sequence of dynamic association strength parameters, and clarify the parameter change period for each connection. Real-time monitoring of parameter changes in the subsequent evolution process of the dynamic correlation strength parameter evolution sequence; if a change is detected in the parameter corresponding to any connection relationship, the values ​​before and after the parameter change are recorded. Based on the changes in the parameter values, update the dynamic association strength parameter of the connection relationship. If the parameter value increases, increase the value of the dynamic association strength parameter of the corresponding connection relationship; if the parameter value decreases, decrease the value of the dynamic association strength parameter of the corresponding connection relationship. While updating the parameters, check whether there are other connections between the two topological nodes associated with the connection relationship. If there are, compare the updated dynamic association strength parameter with the dynamic association strength parameter of other connections, and retain the connection relationship with the largest dynamic association strength parameter value. Record the time of each dynamic adjustment, the identifier of the adjusted connection relationship, the values ​​of the dynamic correlation strength parameters before and after the adjustment, and the reason for the adjustment to form a dynamic adjustment log; Based on the dynamic adjustment log, the impact of parameter changes on the topology is analyzed. If the parameter changes cause the topology to become unbalanced, the relevant connection relationships are adjusted a second time to keep the topology stable. After all dynamic adjustments are completed, an adjusted topology document is generated, which includes all topology node identifiers, connection relationship identifiers, current dynamic association strength parameters, and dynamic association directions. Compare the adjusted topology document with the original topology document, record the structural differences before and after the adjustment, and form a topology adjustment comparison table.

8. The information correlation analysis method based on big data of online comments according to claim 4, characterized in that, The process involves dividing the topological nodes into multiple evolution time interval groups based on their evolution time. Each evolution time interval group contains all the topological nodes within that time interval. The number of topological nodes within each evolution time interval group is counted to form a statistical result of the temporal distribution of topological nodes, including: Extract the evolution time information of all topological nodes and determine the earliest and latest evolution times in the evolution time information; Calculate the difference between the earliest and latest evolution times, and divide the evolution time into multiple equal-length intervals based on the magnitude of the difference, with each interval having the same duration. Assign a unique interval identifier to each evolution time interval and record the start and end evolution times of each evolution time interval; Traverse each topology node, compare the evolution time of the topology node with the start evolution time and end evolution time of each evolution time interval, and determine the evolution time interval to which the evolution time of the topology node belongs. The topology nodes are assigned to the evolution time interval group corresponding to their respective evolution time intervals. Each evolution time interval group contains all topology nodes whose evolution time belongs to that evolution time interval. For each evolution time interval group, count the number of topological nodes within that evolution time interval group and record the counted value; The interval identifier, start evolution time, end evolution time, and corresponding number of topological nodes for each evolution time interval group are compiled into a table. The contents of the table are sorted according to the order of the evolution time intervals to form the statistical results of the time distribution of topological nodes. The statistical results of the time distribution of topological nodes are compared with the evolution time information of all topological nodes. It is determined whether the total number of topological nodes in the statistical results is consistent with the total number of topological nodes actually extracted. If there is a difference in the number, the division and statistical steps are repeated.

9. An information correlation analysis system based on big data of online comments, characterized in that, The method includes a processor and a computer-readable storage medium storing machine-executable instructions, which, when executed by the processor, implement the information association analysis method based on big data of online comments as described in any one of claims 1-8.

Images