Dynamic capturing method for cultural transmission public opinion data

Abstract

The invention relates to the field of data and technology, and discloses a cultural transmission public opinion data dynamic capturing method, which comprises the following steps: step 1, dynamic source configuration and monitoring; 2, intelligent incremental grabbing scheduling is carried out; 3, carrying out multi-modal data fusion processing; 4, aggregating and tracing the event data; 5, data credibility filtering is carried out; and 6, historical data backtracking analysis is carried out. According to the cultural transmission public opinion data dynamic capturing method, through dynamic source configuration and a rule self-learning mechanism and target knowledge base dynamic expansion in monitoring, the problem that a traditional method depends on manual preset rules is solved; when a platform page structure is changed, the system automatically detects differences, generates a new analysis rule, and supplements emerging cultural terms based on user search behaviors; and in combination with an anti-crawling confrontation strategy, self-adaptive capture of multi-source data across social media, video platforms, forums and the like is realized, and the problem of dynamic change of data sources is effectively solved.

Description

Technical Field

[0001] The present invention relates to the field of data and technology, and specifically to a method for dynamically capturing cultural communication public opinion data. Background Art

[0002] During the dissemination process of cultural products (such as movies, variety shows, intangible cultural heritage projects, online literature, etc.), public opinion data such as user comments, social topics, and media reports are scattered across multiple platforms (Weibo, Douyin, Bilibili, Douban, news websites, etc.), and the data forms cover multimodal structures such as text, short videos, ratings, and barrage.

[0003] According to an Internet data processing system and method mentioned in the invention patent with Chinese patent application number 202310648491.X, the Internet data processing system and method can solve the problem that traditional data management is difficult to analyze and manage streaming data, and the Internet data processing system and method have limited data collection functions for processing Internet data.

[0004] In practical applications, the dissemination of cultural events is sudden and cross-platform. Existing technologies rely on manually preset rules and cannot adapt to changes in data sources and fluctuations in public opinion. In addition, commercial public opinion systems lack a knowledge base in the cultural field, which leads to semantic misjudgments between common and popular words (such as mislabeling "50-cent special effects" as a politically sensitive word). Therefore, it is necessary to propose a dynamic capture method for cultural communication public opinion data to solve the above problems. Summary of the Invention

[0005] (1) Technical problems solved

[0006] In response to the shortcomings of the existing technology, the present invention provides a method for dynamically capturing cultural communication public opinion data, which has the advantages of adaptive multi-source platform, support for dynamic incremental capture, and integration of cultural semantic understanding.

[0007] (2) Technical solution

[0008] In order to achieve the above-mentioned adaptive multi-source platform, support dynamic incremental capture, and integrate cultural semantic understanding, the present invention provides the following technical solutions: a method for dynamically capturing cultural communication public opinion data, comprising the following steps:

[0009] Step 1: Dynamic source configuration and monitoring:

[0010] Used to build a target knowledge base for cultural communication, including keywords and synonyms for IP names, artists, work identifiers, and cultural symbols;

[0011] Configure the three-layer resolution protocol for each target platform (social media / video platform / forum / news site):

[0012] (1) API interface authentication and field mapping rules;

[0013] (2) Operation path script of dynamic page rendering engine (HeadlessBrowser);

[0014] (3) XPath / CSS positioning rule tree for static HTML parser;

[0015] Real-time monitoring of platform page structure changes, triggering rule failure alarms;

[0016] Step 2: Intelligent incremental crawling scheduling:

[0017] Based on the cultural event life cycle model, data collection intensity levels are divided (warm-up period / explosion period / decline period);

[0018] Dynamically adjust the collection frequency through the popularity prediction algorithm (LSTM neural network + public opinion dissemination SIR model);

[0019] Build a two-layer priority queue:

[0020] (1) Level 1 queue: responds to real-time external event signals (sudden changes in the hot search list, third-party public opinion warnings);

[0021] (2) Secondary cohort: executed according to the preset cultural target collection plan;

[0022] Step 3: Multimodal data fusion processing:

[0023] Perform cross-modal unified encoding on the original data stream:

[0024] (1) Text comments: remove advertisements, garbled text, and non-target language content;

[0025] (2) Short video / live broadcast: extract ASR speech-to-text and barrage stream;

[0026] (3) Image and text content: OCR recognition and overlay image and text correlation analysis;

[0027] Deep semantic annotation based on pre-trained models in cultural fields;

[0028] Step 4: Event-based data aggregation and tracing:

[0029] Construct a cultural communication event graph: cultural entities are nodes and communication behaviors (forwarding / secondary creation / controversy) are edges;

[0030] Identify cross-platform event clusters using spatiotemporal clustering algorithms (DBSCAN + Dynamic Time Warping DTW);

[0031] Generate a weighted propagation path graph, where edge weights include voice volume, sentiment polarity, and user influence factor;

[0032] Step 5: Data credibility filtering:

[0033] Eliminate online gunman accounts: Based on posting frequency / content duplication / device ID cluster analysis;

[0034] Authority weighting: official media weight = 1.0, individual user weight = 0.2-0.8 (based on number of followers);

[0035] Step 6: Retrospective analysis of historical data: Periodically rerun historical data and supplement associated edges based on new events.

[0036] Preferably, in step 1:

[0037] The rule self-learning mechanism includes:

[0038] (1) When the parsing failure rate is greater than the threshold, the DOM tree comparison module is started to locate the differences in the page structure between the new and old versions;

[0039] (2) Automatically generate candidate XPaths based on the difference regions and update the rule base after verifying the accuracy rate >95% through historical samples;

[0040] Dynamic expansion of the target knowledge base: Automatically supplement emerging cultural terms through user search behavior analysis and industry report mining.

[0041] Preferably, in step 2:

[0042] Anti-climbing strategies include:

[0043] (1) Traffic control: simulate human operation interval (2-10 seconds random delay) + IP proxy pool rotation;

[0044] (2) Behavioral camouflage: Automatically generate legitimate HTTP headers (including Referer / Cookie updates);

[0045] (3) Captcha cracking: integrated image and text captcha recognition model and slider trajectory simulator;

[0046] Interrupted and resumed data collection mechanism: when the capture is interrupted, the last data fingerprint (such as the largest ID on Weibo) is recorded, and after recovery, the data is incrementally synchronized based on the fingerprint.

[0047] Preferably, the deep semantic annotation in step 3 includes:

[0048] Multi-dimensional analysis of cultural emotions:

[0049] (1) Basic emotions: based on the domain emotion dictionary (expanding cultural scene words such as "feelings" and "cultural confidence");

[0050] (2) Aesthetic evaluation: Identify evaluations in dimensions such as vision / narrative / performance (e.g., "excellent costumes, makeup, and props", "拖沓plot").

[0051] (3) Values analysis: Detect sensitive expressions related to ethnic / historical / gender issues.

[0052] Cultural entity linking refers to:

[0053] (1) Link character names and IP terms in comments to knowledge base entities.

[0054] (2) Identify metaphorical expressions (e.g., "Fa Hai doesn't understand love" refers to the cultural symbol of "The Legend of White Snake").

[0055] New language and dialect processing:

[0056] (1) Integrate minority language word segmenters (Tibetan / Uyghur).

[0057] (2) Build a dialect conversion model (semantic alignment between Cantonese and Mandarin).

[0058] Real-time update of Internet buzzwords:

[0059] (1) Discover new abbreviations through bullet screen frequency statistics (e.g., "nbcs" = nobody cares), and dynamically update the term library.

[0060] (2) Disambiguate polysemous words in context (e.g., "虐" means the plot is heart-breaking or the actor is injured).

[0061] Preferably, the event graph construction in step 4 includes:

[0062] Propagation influence quantification algorithm: Edge weight W = α * (number of forwards) + β * (core user influence value) + γ * (emotional intensity), where α, β, γ are industry adjustment coefficients, and the core user influence value is calculated based on the number of fans, authentication level, and historical propagation power.

[0063] Derivative event detection:

[0064] (1) Identify sub-topic branches through the topic model (LDA) (e.g., "casting controversy" gives rise to "digging into the actor's historical remarks").

[0065] (2) Locate the turning point of public opinion through emotional migration analysis (e.g., the detonation period when positive reviews turn into negative reviews).

[0066] Preferably, the final output:

[0067] Dynamic public opinion dashboard:

[0068] (1) Multidimensional indicators of cultural entities (volume trend, emotional cloud map, platform penetration rate).

[0069] (2) List of high-influence nodes (TOP-K opinion leaders, key dissemination articles);

[0070] Crisis warning signs:

[0071] (1) Generate an early warning when the sudden negative emotion cluster exceeds the threshold;

[0072] (2) Mark the cross-circle diffusion risk index of controversial events (e.g., the second dimension circle → mainstream media).

[0073] (3) Beneficial effects

[0074] Compared with the existing technology, the present invention provides a method for dynamically capturing cultural communication public opinion data, which has the following beneficial effects:

[0075] 1. This method for dynamically capturing cultural communication public opinion data solves the problem of traditional methods relying on manually preset rules through dynamic source configuration and monitoring of rule self-learning mechanisms (such as automatic generation of parsing rules through DOM tree comparison) and dynamic expansion of the target knowledge base. When the platform page structure changes, the system automatically detects differences and generates new parsing rules (with a verification accuracy rate of >95%), and supplements emerging cultural terms (such as "national tide" and "metaverse") based on user search behavior. Combined with anti-crawling strategies (IP proxy pool rotation and verification code cracking), it realizes adaptive capture of multi-source data across social media, video platforms, forums, etc., effectively coping with the problem of dynamic changes in data sources.

[0076] 2. This method for dynamically capturing cultural communication public opinion data dynamically adjusts the collection frequency based on the cultural event life cycle model (warm-up period / outburst period / decline period) and the heat prediction algorithm (LSTM+public opinion communication SIR model); responds to hot search mutation events in real time through a double-layer priority queue; combines the spatiotemporal clustering algorithm (DBSCAN+dynamic time warping DTW) to identify cross-platform event clusters (such as a certain IP controversy spreading from Weibo to Douyin), and constructs an event map with cultural entities as nodes and communication behaviors as edges; edge weights integrate voice volume, emotional polarity and user influence factors (such as official media weight = 1.0) to accurately quantify the public opinion dissemination path and solve the problem of tracking sudden cultural events.

[0077] 3. The dynamic capture method of cultural communication public opinion data realizes deep semantic annotation through multimodal data fusion processing and cultural field pre-training model: precise sentiment analysis: expand the scene vocabulary such as "cultural confidence" and "excellent costumes and props", distinguish basic emotions, aesthetic evaluations and values ​​(such as national / historical issues); (1) semantic disambiguation and entity linking: identify metaphorical expressions (such as "Fa Hai does not understand love" refers to "The Legend of White Snake") and dialects / network terms (Cantonese translation, dynamic update of "nbcs" and other terms); (2) trusted data filtering: eliminate network gunmen based on release frequency / device ID clustering, and improve data reliability by combining authority weighting; (3) avoid the misjudgment of cultural terms by general public opinion systems (such as mislabeling "50-cent special effects" as politically sensitive words), and significantly improve analysis accuracy.

[0078] 4. This method of dynamically capturing cultural communication public opinion data integrates multi-dimensional indicators (volume trends, emotional cloud maps, platform penetration) and high-influence nodes (TOP-K opinion leaders) by ultimately outputting a dynamic public opinion dashboard. It marks the risk of controversial events spreading across circles (such as the secondary circle → mainstream media) through crisis warning signals (such as negative emotional clustering exceeding the threshold), assisting users in quickly locating turning points in public opinion (such as the detonation period when positive reviews turn to negative ones). It also combines historical data to backtrack and supplement related edges, forming a closed-loop analysis system to provide data support for cultural communication strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 This is a structural flow chart of the data dynamic crawling method of the present invention;

[0080] Figure 2 This is a flow chart of the dynamic source configuration and monitoring structure of the present invention;

[0081] Figure 3 This is a flow chart of the intelligent incremental crawling scheduling structure of the present invention;

[0082] Figure 4 This is a flow chart of the temporal data aggregation and traceability structure of the present invention. DETAILED DESCRIPTION

[0083] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0084] See also Figure 1-4 A method for dynamically capturing cultural communication public opinion data includes the following steps:

[0085] Step 1: Dynamic source configuration and monitoring:

[0086] Used to build a target knowledge base for cultural communication, including keywords and synonyms for IP names, artists, work identifiers, and cultural symbols;

[0087] Configure the three-layer resolution protocol for each target platform (social media / video platform / forum / news site):

[0088] (1) API interface authentication and field mapping rules;

[0089] (2) Operation path script of dynamic page rendering engine (HeadlessBrowser);

[0090] (3) XPath / CSS positioning rule tree for static HTML parser;

[0091] Real-time monitoring of platform page structure changes, triggering rule failure alarms;

[0092] Step 2: Intelligent incremental crawling scheduling:

[0093] Based on the cultural event life cycle model, data collection intensity levels are divided (warm-up period / explosion period / decline period);

[0094] Dynamically adjust the collection frequency through the popularity prediction algorithm (LSTM neural network + public opinion dissemination SIR model);

[0095] Build a two-layer priority queue:

[0096] (1) Level 1 queue: responds to real-time external event signals (sudden changes in the hot search list, third-party public opinion warnings);

[0097] (2) Secondary cohort: executed according to the preset cultural target collection plan;

[0098] Step 3: Multimodal data fusion processing:

[0099] Perform cross-modal unified encoding on the original data stream:

[0100] (1) Text comments: remove advertisements, garbled text, and non-target language content;

[0101] (2) Short video / live broadcast: extract ASR speech-to-text and barrage stream;

[0102] (3) Image and text content: OCR recognition and overlay image and text correlation analysis;

[0103] Deep semantic annotation based on pre-trained models in cultural fields;

[0104] Step 4: Event-based data aggregation and tracing:

[0105] Construct a cultural communication event graph: cultural entities are nodes and communication behaviors (forwarding / secondary creation / controversy) are edges;

[0106] Identify cross-platform event clusters using spatiotemporal clustering algorithms (DBSCAN + Dynamic Time Warping DTW);

[0107] Generate a weighted propagation path graph, where edge weights include voice volume, sentiment polarity, and user influence factor;

[0108] Step 5: Data credibility filtering:

[0109] Eliminate online gunman accounts: Based on posting frequency / content duplication / device ID cluster analysis;

[0110] Authority weighting: official media weight = 1.0, individual user weight = 0.2-0.8 (based on number of followers);

[0111] Step 6: Retrospective analysis of historical data: Periodically rerun historical data and supplement associated edges based on new events.

[0112] Case example: Long-term tracking of intangible cultural heritage dissemination (ongoing events):

[0113] (1) Application background: Monitor the Mazu cultural communication project (throughout 2023), focusing on capturing: discussions in multiple dialect areas (Minnan dialect / Teochew dialect) during sacrificial activities; cultural controversies caused by the export of cultural and creative products; and retrospective analysis of historical public opinion data.

[0114] Long-term data backtracking mechanism

[0115] (2) Re-run the data of Mazu’s birthday in Meizhou in 2022 to identify the periodic pattern:

[0116] (if the current date is within [5 days before or after the 23rd day of the third lunar month]: #Mazu's Birthday;

[0117] The gripping strength is automatically increased to P1 level)

[0118] (3) Cross-language processing: Chaoshan dialect comment “Mazu Baohaohe” (bless everyone) → converted to entity label: [praying behavior] [positive]; establish religious terminology library: {“golden body”: “statue of deity”, “parade”: “parade ceremony”}

[0119] (4) Overseas public opinion monitoring: Capturing English comments on YouTube / TikTok to identify cultural conflicts (Original article: Why Chinese worship is a bit too much? → Tags: [Cultural misunderstanding][Need for popular science])

[0120] (5) Dynamic graph output: (pie

[0121] TitleMazu culture voice platform distribution

[0122] Weibo: 35

[0123] TikTok: 28

[0124] "Overseas Platform": 22

[0125] "Dialect Forum": 15)

[0126] (flowchartTB

[0127] Core Event [Meizhou Sacrifice Ceremony] --> Domestic Communication [Douyin Challenge "Meet Mazu"]

[0128] Core Event: Overseas Spread [TikTok's Second Creation "MazuDance"]

[0129] Spreading overseas --> Controversy [Foreign media misinterprets "idolatry"]

[0130] Controversial Incident --> Official Response [State Council Information Office Press Conference Clarification])

[0131] In summary, this method for dynamically capturing cultural communication public opinion data solves the problem of traditional methods relying on manually preset rules through the self-learning mechanism of rules in dynamic source configuration and monitoring (such as automatic generation of parsing rules by DOM tree comparison) and dynamic expansion of the target knowledge base; when the platform page structure changes, the system automatically detects differences, generates new parsing rules (verification accuracy > 95%), and supplements emerging cultural terms (such as "national tide" and "metaverse") based on user search behavior; combined with anti-crawling strategies (IP proxy pool rotation, verification code cracking), it realizes adaptive capture of multi-source data across social media, video platforms, forums, etc., and effectively copes with the problem of dynamic changes in data sources.

[0132] In addition, based on the cultural event life cycle model (warm-up period / outburst period / decline period) and the popularity prediction algorithm (LSTM+public opinion dissemination SIR model), the collection frequency is dynamically adjusted; a double-layer priority queue is used to respond to hot search mutation events in real time; combined with the spatiotemporal clustering algorithm (DBSCAN+dynamic time warping DTW), cross-platform event clusters are identified (such as a certain IP controversy spreading from Weibo to Douyin), and an event map is constructed with cultural entities as nodes and dissemination behaviors as edges; the edge weight integrates voice volume, emotional polarity and user influence factors (such as official media weight = 1.0) to accurately quantify the public opinion dissemination path and solve the problem of tracking sudden cultural events.

[0133] In addition, through multimodal data fusion processing and cultural field pre-training models, deep semantic annotation is achieved: precise sentiment analysis: expanding the scene vocabulary such as "cultural confidence" and "excellent costumes and props", distinguishing basic emotions, aesthetic evaluations and values ​​(such as national / historical issues); (1) semantic disambiguation and entity linking: identifying metaphorical expressions (such as "Fa Hai does not understand love" pointing to "The Legend of White Snake") and dialects / network terms (Cantonese translation, dynamic update of "nbcs" and other terms); (2) trusted data filtering: eliminating network gunmen based on release frequency / device ID clustering, and combining authority weighting to improve data reliability; (3) avoiding the misjudgment of cultural terms by general public opinion systems (such as mislabeling "50-cent special effects" as politically sensitive words), and significantly improving analysis accuracy.

[0134] In addition, by finally outputting a dynamic public opinion dashboard, it integrates multi-dimensional indicators (voice trend, emotional cloud map, platform penetration) and high-influence nodes (TOP-K opinion leaders); through crisis warning signals (such as negative emotional clustering exceeding the threshold), it marks the risk of cross-circle diffusion of controversial events (such as the two-dimensional circle → mainstream media), helping users to quickly locate the turning point of public opinion (such as the detonation period when positive reviews turn into negative reviews); combined with historical data backtracking to supplement the associated edges, a closed-loop analysis system is formed to provide data support for cultural communication strategies.

[0135] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0136] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for dynamically capturing cultural communication public opinion data, characterized in that: The following steps are involved: Step 1: Dynamic source configuration and monitoring: Used to build a target knowledge base for cultural communication, including keywords and synonyms for IP names, artists, work identifiers, and cultural symbols; Configure the three-layer resolution protocol for each target platform (social media / video platform / forum / news site): (1) API interface authentication and field mapping rules; (2) Operation path script of dynamic page rendering engine (HeadlessBrowser); (3) XPath / CSS positioning rule tree for static HTML parser; Real-time monitoring of platform page structure changes, triggering rule failure alarms; Step 2: Intelligent incremental crawling scheduling: Based on the cultural event life cycle model, data collection intensity levels are divided (warm-up period / explosion period / decline period); Dynamically adjust the collection frequency through the popularity prediction algorithm (LSTM neural network + public opinion dissemination SIR model); Build a two-layer priority queue: (1) Level 1 queue: responds to real-time external event signals (sudden changes in the hot search list, third-party public opinion warnings); (2) Secondary cohort: executed according to the preset cultural target collection plan; Step 3: Multimodal data fusion processing: Perform cross-modal unified encoding on the original data stream: (1) Text comments: remove advertisements, garbled text, and non-target language content; (2) Short video / live broadcast: extract ASR speech-to-text and barrage stream; (3) Image and text content: OCR recognition and overlay image and text correlation analysis; Deep semantic annotation based on pre-trained models in cultural fields; Step 4: Event-based data aggregation and tracing: Construct a cultural communication event graph: cultural entities are nodes and communication behaviors (forwarding / secondary creation / controversy) are edges; Identify cross-platform event clusters using spatiotemporal clustering algorithms (DBSCAN + Dynamic Time Warping DTW); Generate a weighted propagation path graph, where edge weights include voice volume, sentiment polarity, and user influence factor; Step 5: Data credibility filtering: Eliminate online gunman accounts: Based on posting frequency / content duplication / device ID cluster analysis; Authority weighting: official media weight = 1.0, individual user weight = 0.2-0.8 (based on number of followers); Step 6: Retrospective analysis of historical data: Periodically rerun historical data and supplement associated edges based on new events.

2. A method for dynamically capturing cultural communication public opinion data according to claim 1, characterized in that: In step 1: The rule self-learning mechanism includes: (1) When the parsing failure rate is greater than the threshold, the DOM tree comparison module is started to locate the differences in the page structure between the new and old versions; (2) Automatically generate candidate XPaths based on the difference regions and update the rule base after verifying the accuracy rate >95% through historical samples; Dynamic expansion of the target knowledge base: Automatically supplement emerging cultural terms through user search behavior analysis and industry report mining.

3. A method for dynamically capturing cultural communication public opinion data according to claim 1, characterized in that: In step 2: Anti-climbing strategies include: (1) Traffic control: simulate human operation interval (2-10 seconds random delay) + IP proxy pool rotation; (2) Behavioral camouflage: Automatically generate legitimate HTTP headers (including Referer / Cookie updates); (3) Captcha cracking: integrated image and text captcha recognition model and slider trajectory simulator; Intermittent acquisition mechanism: When the acquisition is interrupted, record the last data fingerprint (such as the maximum ID of Weibo), and perform incremental synchronization based on the fingerprint after recovery.

4. A method for dynamically capturing cultural communication public opinion data according to claim 1, characterized in that: The deep semantic annotation in step 3 includes: Multi-dimensional analysis of cultural emotions: (1) Basic emotions: Based on the domain emotion dictionary (expanding cultural scene words such as "feelings" and "cultural confidence"); (2) Aesthetic evaluation: Identify evaluations in dimensions such as vision / narrative / performance (such as "excellent costumes, makeup, and props" and "拖沓plot"); (3) Value analysis: Detect sensitive expressions related to ethnic / historical / gender issues; Cultural entity linking: (1) Link the character names and IP terms in the comments to the knowledge base entities; (2) Identify metaphorical expressions (such as "Fa Hai doesn't understand love" pointing to the cultural symbol of "The Legend of White Snake"); New language and dialect processing: (1) Integrate minority language tokenizers (Tibetan / Uyghur); (2) Build a dialect conversion model (semantic alignment between Cantonese and Mandarin); Real-time update of internet buzzwords: (1) Discover new abbreviations through bullet screen frequency statistics (such as "nbcs" = nobody cares), and dynamically update the term library; (2) Disambiguate polysemous words in context (such as "虐" referring to a heart-wrenching plot or an actor's injury).

5. A method for dynamically capturing cultural communication public opinion data according to claim 4, characterized in that: The event graph construction in step 4 includes: Propagation influence quantification algorithm: Edge weight W = α*(retweet volume) + β*(core user influence value) + γ*(emotional intensity), where α, β, γ are industry adjustment coefficients, and the core user influence value is calculated based on the number of fans, authentication level, and historical propagation power; Derivative event detection: [[ID=,18]](1) Identify sub-topic branches through the topic model (LDA) (such as "casting controversy" deriving "digging into an actor's past remarks"); (2) Locate the turning point of public opinion through emotional migration analysis (such as the detonation period when positive reviews turn into negative reviews).

6. A method for dynamically capturing cultural communication public opinion data according to claims 1-5, characterized in that: Final output: Dynamic public opinion dashboard: (1) Multi-dimensional indicators of cultural entities (volume trend, emotional cloud map, platform penetration rate); (2) List of high-propagation power nodes (TOP-K opinion leaders, key communication manuscripts); Crisis warning signal: (1) Generate a warning when the clustering of sudden negative emotions exceeds the threshold; (2) Mark the cross-circle diffusion risk index of controversial events (such as from the ACG circle to mainstream media).

Images