Social media topic polarization degree quantification method and device
By building a user opinion network and using a random walk algorithm to quantify the degree of topic polarization of different social media platforms, the problem of difficulty in quantifying topic polarization between different platforms in the existing technology is solved, and the precise quantification of social media topic polarization is achieved.
Patent Information
- Application Number
- CN202411710517.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-27
- Publication Date
- 2025-06-06
AI Technical Summary
The prior art is difficult to quantify the degree of topic polarization between different social media platforms and ignores the differences between different platforms, making it difficult to comprehensively evaluate the degree of topic polarization of social media.
By building a user opinion network, using the user's comment relationship under polarized topics, combining the random walk algorithm to construct quantitative indicators of social media topic polarization, and quantifying the degree of topic polarization on different social media platforms.
It has achieved the quantification of the polarization degree of different social media platforms and topics, which can effectively identify the polarization of user groups, provide a unified scale of polarization quantization, and improve the accuracy of quantization.
Smart Images
Figure CN120106087A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a method and device for quantifying the degree of polarization of social media topics. Background Art
[0002] With the development of Internet technology and the popularity of smart phones, social media has become the main channel for social interaction, information dissemination and cultural exchange around the world. In this process, social media platforms widely use complex information filtering and personalized recommendation algorithms to present users with content that is more in line with their interests and opinions. Among them, the information filtering mechanism screens massive amounts of information based on the user's historical behavior, preferences and interaction patterns, and selects content that may arouse the user's interest. However, this mechanism of personalized recommendation and information filtering may lead to the formation of information cocoons, exposing users to more information that is consistent with their existing views, while information with different or opposite views is almost not displayed. Users may form a preference for single information on social media platforms, making them more inclined to interact with information that is consistent with their existing beliefs and opinions, while they may show a lower acceptance of challenging or dissenting views, thereby triggering social media topic polarization.
[0003] Although existing technologies can detect the existence of social media topic polarization and propose coping strategies, quantifying the degree of social media topic polarization is a major challenge in polarization research. Social media platforms disseminate information in different forms, such as video and text, which poses a challenge to quantifying polarization. Existing research related to the quantification of social media topic polarization only focuses on the degree of polarization of a single social media platform, ignoring the differences between different social media platforms, making it difficult to quantify the degree of polarization between different social media platforms. Summary of the invention
[0004] In view of the current difficulty in quantifying the polarization of topics on different social media platforms, the present invention provides a method and device for quantifying the degree of polarization of social media topics. The present invention constructs a user opinion network through the comment relationship of users under polarized topics, and performs random walks to construct a social media topic polarization quantitative index to quantify the degree of topic polarization on social media platforms.
[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:
[0006] A method for quantifying the degree of polarization of social media topics, comprising the following steps:
[0007] S1: Calculate user inclination based on the text content of user first-level comments under posts on polarized topics on social media platforms.
[0008] S2: By analyzing the relationship between user comments involving polarized topics, an undirected user opinion network is constructed. The nodes in the network represent users and are assigned user preference attributes. The edges represent the comments of users on the same post, and the weight of the edge represents the number of common comments made by users on the same post.
[0009] S3: Constructing a quantitative metric for topic polarization in social media by performing random walks in user opinion networks.
[0010] S4: Analyze the polarization level of different topics on social media platforms through the quantitative indicators constructed in S3.
[0011] Furthermore, the step S1 includes the following steps:
[0012] S1.1: The tendency of polarized topic posts is manually judged by the titles of the posts, i.e., negative tendency, neutral tendency, and positive tendency, which are represented by -1, 0, and 1 respectively.
[0013] S1.2: Calculate the user's inclination towards the topic based on the text sentiment inclination of the user's first-level comments under the polarized topic post. When the text sentiment inclination is positive, the user's comment inherits the post's inclination; otherwise, the user's comment inherits the opposite inclination of the post. When the text sentiment inclination is neutral, the user's comment inherits the neutral inclination. Finally, the user's inclination is determined by the inclination that appears most frequently among all the comments on the polarized topic.
[0014] Furthermore, the step S2 specifically includes:
[0015] Based on the first-level comments of users on polarized topics, an undirected user opinion network is constructed. The nodes of the network represent users, and the nodes are given the tendency attributes calculated by step S1. The edges represent the comments of users on the same post, and the edge weights represent the number of times users have jointly commented on the same post.
[0016] Furthermore, the step S3 includes the following steps:
[0017] S3.1: In the user opinion network constructed in step S2, select nodes as initial nodes i in turn for random walk. The transition probability of random walk is based on the normalized value of all edge weights of the current node, that is, P(x|y)=(w yx -w min ) / (w max -w min ), where wyx represents the weight of the edge between node y and node x, w min represents the minimum weight of the edge containing node y, w max Indicates the maximum weight of the edge containing node y. In the random walk process, the number of steps S required to traverse all tendencies in the user opinion network i, represents the tendency coverage time of node i. Finally, after performing R rounds of random walks with node i as the initial node, the average tendency coverage time S i (t) represents the final tendency coverage time of node i, where n represents the number of rounds, i.e.
[0018] S3.2: Construct a zero model network corresponding to the user opinion network, that is, keep the original user opinion network structure unchanged, and randomly disrupt the node inclination and edge weights in the network. In the obtained zero model network, repeat the random walk operation in step S3.1.
[0019] S3.3: Calculate the average tendency coverage time of all nodes in the original user opinion network and the zero model network respectively Where N represents the number of nodes in the user opinion network. Finally, by calculating the average tendency coverage time of the original user opinion network minus the average tendency coverage time of the null model network, the normalized tendency coverage time of the original user opinion network is obtained: NCT = (S i (t)-μ(t) null ) / μ(t) org , as a quantitative indicator of topic polarization. The larger the value of NCT, the more serious the polarization of user groups in the user opinion network.
[0020] Further, the step S4 includes the following steps:
[0021] S4.1: Based on the polarization quantitative indicators obtained in S3, calculate the polarization index values under the topics of different social media platforms, and compare and analyze these indicators to determine the degree of polarization.
[0022] S4.2: Based on the differences in polarization levels, compare the polarization levels of different platforms and topics, identify platforms and topics with higher or lower polarization levels, and use this to determine the degree of user polarization on different platforms and which topics are more likely to cause stronger opinion divisions on social media.
[0023] A second aspect of the present invention relates to a device for quantifying the degree of polarization of social media topics, comprising a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, they are used to implement a method for quantifying the degree of polarization of social media topics of the present invention.
[0024] A third aspect of the present invention relates to a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements a method for quantifying the degree of polarization of social media topics of the present invention.
[0025] The technical concept of the present invention is: a method for quantifying the degree of polarization of social media topics. In steps S1 and S2, the user opinion network is constructed by using the user's comments under polarized topic posts; in step S3, the polarization quantification index of social media topics is constructed by performing random walks in the user opinion network; in S4, the social media platform is analyzed by the constructed index to quantify the degree of polarization of platform topics. The topic polarization quantification index proposed by the present invention can be applied to social media platforms that disseminate information in different forms, such as video social media, blog social media, picture sharing social media, etc. This quantitative index can effectively utilize the text information and preference information of users of social media platforms in the process of browsing polarized information, and make full use of existing resources to achieve the quantification of social media topic polarization.
[0026] The beneficial effects of the present invention are as follows: 1) a social media topic polarization quantification index is constructed, which can achieve a unified scale of polarization quantification for social media platforms that disseminate information in different forms. 2) by combining user comment relationships and comment text content to build a network, the accuracy of social media topic polarization quantification is higher. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Figure 1 A framework diagram of a method for quantifying the degree of polarization of social media topics provided by the present invention.
[0028] Figure 2 A comparison chart of the quantified degree of polarization of social media topics on multiple platforms provided by the present invention.
[0029] Figure 3 It is a schematic diagram of the device of the present invention. DETAILED DESCRIPTION
[0030] The specific implementation modes of the present invention are further described in detail below in conjunction with the accompanying drawings.
[0031] Example 1
[0032] Reference Figure 1 A method for quantifying the degree of polarization of social media topics is proposed. Taking the degree of polarization of multiple topics on two platforms, Bilibili and YouTube, as examples, the method includes the following steps:
[0033] S1: Calculate user inclination based on the text content of user first-level comments under multiple polarized topic posts.
[0034] S2: By analyzing the relationship between user comments involving multiple polarized topics, an undirected user opinion network is constructed. The nodes in the network represent users and are given tendency attributes. The edges represent the comments of users on the same post, and the weight of the edge represents the number of common comments made by users on the same post.
[0035] S3: Calculate the average tendency coverage time of all nodes in the original user opinion network and the zero model network respectively Where N represents the number of nodes in the user opinion network. Finally, by calculating the normalized tendency coverage time of the original user opinion network, multiple NCTs, a quantitative index of topic polarization is obtained. The larger the value of NCT, the more serious the polarization of the user group in the user opinion network.
[0036] S4: Analyze the polarization level of different topics on social media platforms through the quantitative indicators constructed in S3.
[0037] Furthermore, the step S1 includes the following steps:
[0038] S1.1: By analyzing the titles of polarized topic posts, we manually judge the tendency of the posts, using -1, 0, and 1 to represent negative, neutral, and positive tendencies, respectively.
[0039] S1.2: Calculate the user's inclination towards the topic based on the text sentiment inclination of the user's first-level comments under the polarized topic post. Specifically, suppose user u has published N comments related to topic T, where N pT Represents the tendency of the comment. For the comment posted by user u, if the sentiment tendency of the text content is positive, the user comment inherits the tendency of the post. Otherwise, the user comment inherits the opposite tendency of the post. When the sentiment tendency of the text content is neutral, the user comment inherits the neutral tendency. Therefore, for the N comments posted by user u under the post involving topic T, the tendency is N pT =-1 (negative tendency), N pT = 0 (neutral) and N pT = 1 (positive tendency), where {N pT =-1}+{N pT =0}+{N pT =1} = N. Finally, the tendency of user u on topic T can be regarded as the tendency that appears most frequently in his personally generated comments, i.e., p u =max{N pT Among them, the sentiment tendency is completed based on the NLTK sentiment analysis package.
[0040] Furthermore, the step S2 specifically includes:
[0041] Based on the first-level comments of users under posts involving polarized topics, an undirected user opinion network is constructed. The nodes in the network represent users, and the nodes are given the user tendency attribute p calculated in step S1. u , the edge represents the user’s comments on the same post, and the edge weight represents the number of times the user has made common comments on the same post.
[0042] Furthermore, the step S3 includes the following steps:
[0043] S3.1: First, in the user opinion network constructed in step S2, nodes are selected as initial nodes i for random walks. The transition probability is determined by the weight of the edge between the node and its neighbors, as shown in Formula 1.
[0044]
[0045] where w yx represents the weight of the edge between node y and node x, w min represents the minimum weight of the edge containing node y, w max represents the maximum weight of the edge containing node y. Secondly, the number of steps S required for random walk to traverse all tendencies in the user opinion network i , represents the tendency coverage time of node i. Finally, after performing R rounds of random walks with node i as the initial node, the average tendency coverage time S i (t) represents the final tendency coverage time of node i, where n represents the number of rounds, as shown in Formula 2.
[0046]
[0047] S3.2: Construct a zero model network corresponding to the user opinion network, that is, keep the structure of the original user opinion network unchanged, and randomly disrupt the node inclinations and edge weights in the network. In the resulting zero model network, repeat the random walk operation of S3.1.
[0048] S3.3: Calculate the average tendency coverage time μ(t) of all nodes in the original user opinion network and the zero model network respectively by formula 3: org 、μ(t) null Finally, by calculating the normalized tendency coverage time NCT of the original user opinion network, as shown in Formula 4, the quantitative index of topic polarization is obtained. The larger the value of NCT, the more serious the polarization of the user group in the user opinion network.
[0049]
[0050] NCT=(S i (t)-μ(t) null ) / μ(t) org (4)
[0051] Further, the step S4 includes the following steps:
[0052] S4.1: Based on the polarization quantitative indicators obtained in S3, calculate the polarization index values under the topics of different social media platforms, and compare and analyze these indicators to determine the distribution of polarization degrees.
[0053] S4.2: Based on the difference in polarization, compare the polarization intensity of topics on the two platforms of Bilibili and YouTube, identify the platforms and topics with higher or lower polarization, and thus judge the degree of user polarization on different platforms and which topics are likely to cause stronger opinion differentiation on social media. Figure 2 As shown, on the Bilibili platform, the median coverage time of NCT on three different topics is greater than 0 and exceeds that of YouTube, reflecting the relatively strong polarization in user discussions.
[0054] In view of the difficulties in the current field of quantification of social media topic polarization, the present invention designs a method for quantifying the degree of polarization of social media topics. A user opinion network is constructed based on user comments on polarized topic posts, and node tendency attributes are assigned. Finally, a polarization quantification index of topics in social media is constructed by random walks in the user opinion network and the corresponding null model network, and the degree of topic polarization on the social media platform is quantified.
[0055] Example 2
[0056] This embodiment relates to a device for quantifying the degree of polarization of social media topics, such as Figure 3 , including a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, they are used to implement a method for quantifying the degree of polarization of social media topics in Example 1.
[0057] Example 3
[0058] This embodiment 3 relates to a computer-readable storage medium on which a program is stored. When the program is executed by a processor, a method for quantifying the degree of polarization of social media topics in embodiment 1 is implemented.
[0059] The contents described in the embodiments of this specification are merely an enumeration of the implementation forms of the inventive concept. The protection scope of the present invention should not be regarded as limited to the specific forms described in the embodiments. The protection scope of the present invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.
Claims
1. A method for quantifying the degree of polarization of social media topics, characterized by: The following steps are involved: S1: Calculate user inclination based on the text content of user first-level comments under polarized topic posts; S2: By analyzing the first-level comment relationships of users involved in polarized topics, an undirected user opinion network is constructed. The nodes in the network represent users and are given tendency attributes. The edges represent the comments of users on the same post, and the weight of the edge represents the number of common comments made by users on the same post. S3: Constructing a quantitative indicator of topic polarization in social media by performing random walks in user opinion networks; S4: Analyze the polarization level of different topics on social media platforms through the quantitative indicators constructed in S3.
2. A method for quantifying the degree of polarization of social media topics as claimed in claim 1, characterized in that: The step S1 specifically includes: S1.1: The tendency of posts is manually judged through the titles of polarized topic posts, i.e. negative tendency, neutral tendency, and positive tendency, which are represented by -1, 0, and 1 respectively; S1.2: Calculate the user's inclination towards the topic based on the text sentiment inclination of the user's first-level comments under the polarized topic post; when the text sentiment inclination is positive, the user's comment inherits the post's inclination; otherwise, the user's comment inherits the opposite inclination of the post; when the text sentiment inclination is neutral, the user's comment inherits the neutral inclination; finally, the user's inclination is determined by the tendency that appears most frequently among all the comments on the polarized topic.
3. A method for quantifying the degree of polarization of social media topics as claimed in claim 1, characterized in that: The step S2 specifically includes: Based on the user's comments on polarized topics, an undirected user opinion network is constructed; the nodes of the network represent users, and the nodes are given the tendency attributes calculated by step S1; the edges represent the user's comments on the same post, and the edge weight represents the number of times the user has made joint comments on the same post.
4. A method for quantifying the degree of polarization of social media topics according to claim 1: said step S3 specifically comprises: S3.1: In the user opinion network constructed in step S2, select nodes as initial nodes i in turn for random walk; Its transition probability is determined by the weight of the edge between the node and its neighbors, as shown in Formula 1, where w yx represents the weight of the edge between node y and node x, w min represents the minimum weight of the edge containing node y, w max represents the maximum weight of the edge containing node y; secondly, the number of steps S required for random walk to traverse all tendencies in the user opinion network i , represents the tendency coverage time of node i; finally, after performing R rounds of random walks with node i as the initial node, the average tendency coverage time S i (t) represents the final tendency coverage time of node i, where n represents the number of rounds, as shown in Formula 2; S3.2: Construct a zero model network corresponding to the user opinion network, that is, keep the original user opinion network structure unchanged, randomly disrupt the node inclination and edge weights in the network; repeat the random walk operation in step S3.1 in the obtained zero model network; S3.3: Calculate the average tendency coverage time μ(t) of all nodes in the original user opinion network and the zero model network respectively by formula (3): org 、μ(t) null ; Finally, by calculating the normalized tendency coverage time NCT of the original user opinion network, as shown in formula (4), the quantitative index of topic polarization is obtained; the larger the value of NCT, the more serious the polarization of the user group in the user opinion network; NCT=(S i (t)-μ(t) null ) / μ(t) org (4)。 5. A method for quantifying the degree of polarization of social media topics according to claim 1: said step S4 specifically comprises: S4.1: Based on the polarization quantitative indicators obtained in S3, calculate the polarization index values under the topics of different social media platforms, and conduct comparative analysis on these indicators to determine the distribution of polarization degrees; S4.2: Based on the differences in polarization, compare the polarization intensity of topics on the platforms and identify platforms and topics with higher or lower polarization levels. This will help determine the degree of user polarization on different platforms and which topics are more likely to cause stronger opinion divisions on social media.
6. A device for quantifying the degree of polarization of social media topics, characterized in that: The invention comprises a memory and one or more processors, wherein the memory stores executable codes, and when the one or more processors execute the executable codes, they are used to implement a method for quantifying the degree of polarization of social media topics according to any one of claims 1 to 5.
7. A computer-readable storage medium, characterized in that: A program is stored thereon, and when the program is executed by a processor, a method for quantifying the degree of polarization of social media topics as described in any one of claims 1-5 is implemented.