A cultural service live broadcast platform operation process management method and system

Abstract

This invention relates to the technical field of the operation process of a cultural service live streaming platform, and discloses a method and system for managing the operation process of a cultural service live streaming platform. The core of the method lies in the introduction of a cultural context rationality assessment mechanism. By conducting multi-dimensional analysis of live streaming activity applications and combining them with a cultural knowledge base for intelligent judgment, the method can achieve more accurate and efficient review and scheduling of live streaming content.

Description

Technical Field

[0001] This invention relates to the field of operation process management for cultural service live streaming platforms, and more specifically, to a method and system for operation process management of cultural service live streaming platforms. Background Technology

[0002] In digital cultural and creative service scenarios, live-streaming platforms for cultural services often encounter inefficiencies in their operational process management. Traditional methods relying on manual scheduling struggle to handle the large influx of simultaneous live-streaming applications, allocate resources effectively, and conduct content review. This frequently results in live-streaming events failing to start on time, resources being wasted, and even non-compliant situations arising. Existing systems also often lack the ability to analyze all data from the beginning to the end of a live-streaming event in real time, making it difficult to flexibly adjust each stage of the process. Summary of the Invention

[0003] This invention provides a method and system for managing the operation process of a cultural service live streaming platform. It aims to solve the problems in the operation process management of existing cultural service live streaming platforms, such as the large number of "false reports" caused by the insufficient understanding of cultural content by the automated review system, which in turn leads to huge pressure on manual review, extended review cycle, waste of platform resources, chaotic live streaming schedules, and damage to the platform's reputation.

[0004] The technical solution of this application is as follows:

[0005] Firstly, this application discloses a method for managing the operation process of a cultural service live streaming platform, the method comprising:

[0006] Receive applications for live streaming events, perform text parsing and image recognition on the content of the applications, extract sensitive words from a preset sensitive word library and sensitive image features from a preset sensitive image library, and use the extracted sensitive words and sensitive image features as risk indication information.

[0007] In response to the extracted risk indication information, cultural theme keywords are extracted from the content of the live streaming event application.

[0008] Based on the extracted cultural theme keywords, matching cultural background records are retrieved from the cultural knowledge base. Each cultural background record in the cultural knowledge base includes a cultural theme field, a risk indication field, and a corresponding cultural context rationality benchmark value.

[0009] Calculate the string similarity between the risk indication information and the risk indication field in the retrieved cultural background records, and use the calculated string similarity as the matching score;

[0010] The matching degree is multiplied by the cultural context rationality benchmark value in the cultural background record to generate a rationality score for the risk indication information in the cultural context.

[0011] Based on the reasonableness score, the ranking weight of live streaming event applications in the review queue is dynamically adjusted, and the ranking weight is positively correlated with the reasonableness score;

[0012] When the adjusted sorting weight is greater than or equal to the preset acceleration threshold, a cultural exemption flag is assigned to the live event application, and the live event application is routed to the preset acceleration processing thread pool.

[0013] Furthermore, this application proposes that the cultural knowledge base be pre-constructed in the following ways:

[0014] Get multiple approved historical live streaming event applications, each with its final review result marked;

[0015] For each historical live-streaming event application, extract the cultural theme keywords and risk indication information from the application.

[0016] For historical live-streaming event applications that have passed the review and contain risk indication information, the combination of cultural theme keywords and risk indication information in the historical live-streaming event application will be treated as a cultural background record, and the cultural context rationality benchmark value will be initially set to 1.

[0017] For historical live-streaming event applications that are rejected but whose risk indication information is marked as valid, the combination of cultural theme keywords and risk indication information in the historical live-streaming event application will be used as a cultural background record, and the cultural context rationality benchmark value will be initially set to 0.

[0018] All completed cultural background records are stored in a cultural knowledge base.

[0019] Through this technical solution, this application can use historical review data to build and initialize a cultural knowledge base, providing basic data for subsequent cultural context rationality assessment, ensuring the accuracy and reliability of the assessment, avoiding the complexity of building a knowledge base from scratch, and providing the system with the ability to learn and evolve.

[0020] In some preferred implementations, string similarity is calculated using an edit distance algorithm:

[0021] Similarity = 1 - (edit distance / max(length of first string, length of second string));

[0022] The edit distance is the minimum number of single-character edit operations required to convert risk indication information into a risk indication field. Edit operations include insertion, deletion, and replacement.

[0023] As an optional approach, string similarity can be calculated using the cosine similarity algorithm, including:

[0024] The risk indication information and the risk indication field are segmented into words to obtain the first term set and the second term set.

[0025] Construct a vector space with all terms as the dimension, and map the first term set and the second term set into the first vector and the second vector, respectively;

[0026] Similarity = (first vector · second vector) / (|first vector| × |second vector|).

[0027] Through this technical solution, this application can use the cosine similarity algorithm to evaluate the similarity between risk indication information and risk indication fields in cultural background records from a semantic level. It is particularly suitable for handling text matching scenarios where word order is not sensitive, further improving the accuracy and robustness of matching.

[0028] Furthermore, the dynamic adjustment method for the ranking weights is as follows:

[0029] Set the baseline weight W0;

[0030] Calculate the weight increment ΔW = k × S, where S is the rationality score and k is the preset proportional coefficient, which ranges from 0 to 1.

[0031] The sorting weight W = W0 + ΔW;

[0032] When the ranking weight W exceeds the preset maximum weight threshold, W is truncated to the maximum weight threshold.

[0033] Building upon this, this application further proposes steps for the automatic updating of the cultural knowledge base:

[0034] Record the actual review results of each live event application after the review is completed;

[0035] If the actual review result is inconsistent with the expected review result corresponding to the reasonableness score, the cultural knowledge base will be updated.

[0036] The update method is to lower the confidence level of the cultural context rationality baseline in the cultural background records and accumulate the number of inconsistencies;

[0037] When the number of inconsistencies exceeds the preset correction threshold, the cultural context rationality benchmark value will be adjusted by one step size in the direction corresponding to the actual audit result, with a step size of 0.1.

[0038] To enhance functionality, the method in this application also includes an audit efficiency monitoring step:

[0039] Record the average processing time T_acc for each live event application that has been added to the accelerated processing thread pool, from receipt to completion of review;

[0040] Record the average processing time T_normal from receipt to completion of review for each live event application that is not included in the accelerated processing thread pool;

[0041] Calculate the speedup efficiency ratio R = T_normal / T_acc;

[0042] When the acceleration efficiency ratio R is lower than the preset efficiency maintenance threshold, the acceleration threshold is lowered by 5% of the current acceleration threshold.

[0043] Building upon the above, this application further proposes that, before performing text parsing and image recognition on the content of the live-streaming event application, the following steps are also included:

[0044] The format of the live streaming event application is validated to determine whether the application contains a complete title, description, cover image, and appointment time information.

[0045] If any of the required information is missing, the live event application will be routed to the information completion queue, and an information completion request will be returned to the initiator.

[0046] Furthermore, this application also proposes that after routing live streaming event requests to a pre-defined accelerated processing thread pool, it further includes:

[0047] Monitor the current load rate of the accelerated processing thread pool;

[0048] When the current load rate exceeds the preset load threshold, the value of the acceleration threshold is temporarily increased to reduce the number of live event applications newly added to the acceleration processing thread pool.

[0049] When the current load rate is lower than the load threshold and continues to exceed the preset duration threshold, the acceleration threshold is restored to its original value.

[0050] Secondly, this application also discloses a management system for the operation process of a cultural service live streaming platform, including:

[0051] The risk identification module is used to receive applications for live streaming events, and to perform text parsing and image recognition on the content of the applications. It extracts sensitive words from a preset sensitive word library and sensitive image features from a preset sensitive image library, and uses the extracted sensitive words and sensitive image features as risk indication information.

[0052] The theme extraction module is used to extract cultural theme keywords from the content of the live event application in response to the risk indication information extracted by the risk identification module.

[0053] The retrieval module is used to retrieve matching cultural background records from the cultural knowledge base based on the cultural theme keywords extracted by the theme extraction module. Each cultural background record in the cultural knowledge base includes a cultural theme field, a risk indication field, and a corresponding cultural context rationality benchmark value.

[0054] The similarity calculation module is used to calculate the string similarity between the risk indication information and the risk indication field in the cultural background records retrieved by the retrieval module, and the calculated string similarity is used as the matching degree.

[0055] The rating generation module is used to multiply the matching degree calculated by the similarity calculation module with the cultural context rationality benchmark value in the cultural background record to generate a rationality rating of the risk indication information in the cultural context.

[0056] The weight adjustment module is used to dynamically adjust the ranking weight of live event applications in the review queue based on the rationality score generated by the scoring generation module. The ranking weight is positively correlated with the rationality score.

[0057] The routing module is used to assign a cultural exemption flag to the live event application and route the live event application to the preset acceleration processing thread pool when the sorting weight adjusted by the weight adjustment module is greater than or equal to the preset acceleration threshold.

[0058] Beneficial effects

[0059] The cultural service live streaming platform operation process management method and system disclosed in this application receive live streaming activity applications and perform text parsing and image recognition to extract potential risk indication information. Based on this, the system responds to the risk indication information by extracting cultural theme keywords from the application content and retrieving matching cultural background records from a pre-built cultural knowledge base based on these keywords. Each record in this cultural knowledge base contains not only cultural theme and risk indication fields but also a corresponding cultural context rationality benchmark value. Subsequently, the system calculates the string similarity between the risk indication information and the risk indication fields in the retrieved cultural background records as a matching degree, and multiplies this matching degree by the cultural context rationality benchmark value to generate a rationality score for the risk indication information in a specific cultural context. Based on this rationality score, the system dynamically adjusts the ranking weight of the live streaming activity application in the review queue, where the ranking weight is positively correlated with the rationality score. When the adjusted ranking weight is greater than or equal to a preset acceleration threshold, the system assigns a cultural exemption mark to the live streaming activity application and routes it to a preset accelerated processing thread pool.

[0060] Through the above technical solution, this application effectively solves the problem of "false alarms" caused by insufficient understanding of cultural content in existing automated review systems. Attached Figure Description

[0061] Figure 1 This is a flowchart illustrating a method for managing the operation process of a cultural service live streaming platform, as provided in an embodiment of the present invention.

[0062] Figure 2 This is a schematic diagram of the operation process management system for a cultural service live streaming platform provided in an embodiment of the present invention. Detailed Implementation

[0063] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] Reference Figure 1 , Figure 1 This is a flowchart illustrating a method for managing the operation process of a cultural service live streaming platform, as provided in an embodiment of the present invention, including:

[0065] Step S11: Receive a live event application, and perform text parsing and image recognition on the content of the live event application. Extract sensitive words from a preset sensitive word library and sensitive image features from a preset sensitive image library contained in the content, and use the extracted sensitive words and sensitive image features as risk indication information.

[0066] Step S12: In response to the extracted risk indication information, extract cultural theme keywords from the content of the live broadcast event application;

[0067] Step S13: Based on the extracted cultural theme keywords, retrieve matching cultural background records from the cultural knowledge base. Each cultural background record in the cultural knowledge base includes a cultural theme field, a risk indication field, and a corresponding cultural context rationality benchmark value.

[0068] Step S14: Calculate the string similarity between the risk indication information and the risk indication field in the retrieved cultural background record, and use the calculated string similarity as the matching degree;

[0069] Step S15: Multiply the matching degree by the cultural context rationality benchmark value in the cultural background record to generate a rationality score of the risk indication information in the cultural context;

[0070] Step S16: Based on the rationality score, dynamically adjust the ranking weight of the live streaming activity application in the review queue, wherein the ranking weight is positively correlated with the rationality score;

[0071] Step S17: When the adjusted sorting weight is greater than or equal to the preset acceleration threshold, assign a cultural exemption flag to the live streaming activity application and route the live streaming activity application to the preset acceleration processing thread pool.

[0072] This application, by introducing a cultural context rationality assessment mechanism, can effectively identify and address potential risks in live-streaming content related to cultural services, while avoiding excessive censorship of content with cultural value. This significantly improves review efficiency, optimizes resource allocation, and ultimately enhances user experience and platform reputation.

[0073] This application provides a method for managing the operation process of a cultural service live streaming platform, aiming to solve the problems of low review efficiency, resource waste, and poor user experience caused by the lack of understanding of the cultural context when handling cultural content in existing live streaming platforms. The core of this method lies in introducing a cultural context rationality assessment mechanism. By conducting multi-dimensional analysis of live streaming activity applications and combining it with a cultural knowledge base for intelligent judgment, it achieves more accurate and efficient review and scheduling of live streaming content.

[0074] Specifically, this method first receives a live streaming event application and then performs text parsing and image recognition on its content. Text parsing involves using natural language processing on the text descriptions, titles, and other information in the live streaming application to identify potentially sensitive words. Image recognition analyzes visual content such as the live streaming cover and promotional images to detect the presence of pre-defined sensitive image features. These identified sensitive words and image features are collectively referred to as risk indication information, serving as the initial basis for judging whether the live streaming content poses a risk.

[0075] After identifying risk indication information, the system further extracts cultural theme keywords from the content of the live streaming event application. These keywords can be categories of live streaming content (such as "intangible cultural heritage," "traditional opera," "ethnic dance," etc.) or specific cultural elements (such as "Peking Opera," "paper cutting," "Miao embroidery," etc.). The extraction of cultural theme keywords can be achieved in various ways. For example, rule-based keyword matching can be used, where a pre-defined database of cultural theme keywords is established, and the system compares the live streaming content against this database; alternatively, machine learning-based methods can be used to automatically identify and extract cultural themes from the text through model training.

[0076] After extracting cultural theme keywords, the system retrieves matching cultural background records from the cultural knowledge base based on these keywords. The cultural knowledge base is a pre-built database storing a large amount of cultural background information. Each cultural background record includes a cultural theme field, a risk indicator field, and a corresponding cultural context rationality benchmark value. The cultural theme field is used to match the cultural theme keywords in the live stream application; the risk indicator field records potentially "sensitive" content that is considered reasonable under that cultural theme; and the cultural context rationality benchmark value quantifies the reasonableness of specific risk indicator information within that cultural context. For example, for a live stream about "dance," its cultural background record might include "sensitive" as a cultural theme field, while features such as "specific clothing" and "specific movements" might be initially identified as sensitive by the system. These features would be marked as reasonable within the specific cultural context in the risk indicator field and assigned a higher cultural context rationality benchmark value.

[0077] Next, the system calculates the string similarity between the risk indication information in the live-streaming event application and the risk indication field in the retrieved cultural background record. String similarity measures the degree of similarity between two text strings, and there are various methods for calculating it. For example, the edit distance algorithm can be used, which measures similarity by calculating the minimum number of single-character editing operations (including insertion, deletion, and replacement) required to transform one string into another. The higher the similarity, the closer the risk indication information is to the reasonably sensitive content described in the cultural background record.

[0078] After obtaining the string similarity (i.e., matching degree), the system multiplies it by the cultural context reasonableness benchmark value in the cultural background record, thereby generating a reasonableness score for the risk indication information within the cultural context. This reasonableness score is a comprehensive indicator that considers not only the degree of matching between the risk indication information and the known cultural background, but also the degree of reasonableness of a specific risk within that cultural context. The higher the score, the more reasonable the risk indication information in the live streaming event application is within the specific cultural context, and the lower the likelihood of it being misjudged as non-compliant.

[0079] Based on the generated reasonableness score, the system dynamically adjusts the ranking weight of live streaming applications in the review queue. The ranking weight is positively correlated with the reasonableness score; that is, a higher reasonableness score results in a higher ranking weight. This means that live streaming applications assessed as having higher reasonableness within the cultural context will be prioritized, thus shortening their review waiting time. The ranking weight can be adjusted in several ways. For example, a baseline weight can be set, and then the weight increment can be calculated by multiplying the reasonableness score by a proportional coefficient. The final ranking weight equals the baseline weight plus the weight increment.

[0080] Finally, when the adjusted sorting weight is greater than or equal to the preset acceleration threshold, the system will assign a cultural exemption tag to the live stream application and route it to the preset accelerated processing thread pool. The cultural exemption tag indicates that the live stream application is considered highly reasonable within the cultural context and can enjoy a faster review process. The accelerated processing thread pool is a resource pool specifically designed to process high-priority live stream applications or those that have already received a cultural exemption tag. Its processing speed is typically faster than the regular review queue, thereby further improving review efficiency.

[0081] The cultural service live streaming platform operation process management method proposed in this application effectively solves the "false alarm" problem in traditional review systems when processing cultural content by introducing a cultural context rationality assessment mechanism. Traditional systems often rely solely on the recognition of sensitive words and images, lacking a deep understanding of the cultural background. This leads to a large amount of culturally valuable live streaming content being incorrectly marked as risky, thereby increasing the burden of manual review, lengthening the review cycle, and wasting resources.

[0082] The core innovation of this application lies in its ability to go beyond simply identifying risk indications; it evaluates these indications within a specific cultural context. By extracting cultural keywords and combining them with a pre-built cultural knowledge base, the system can determine the appropriateness of specific sensitive content within a particular cultural context. For example, in a live stream involving traditional dance, certain costumes or movements might be considered sensitive in a general context, but perfectly reasonable and necessary in its specific cultural context. This application achieves intelligent understanding and quantitative evaluation of such cultural contexts by calculating the similarity between the risk indication information and risk indication fields in the cultural background record, and by combining this with a cultural context rationality benchmark value to generate a rationality score.

[0083] Compared with existing technologies, this application has the following advantages: First, it significantly improves review efficiency. By assigning cultural exemption tags to live streaming applications with high reasonableness scores and routing them to an accelerated processing thread pool, the review time for these applications can be greatly shortened, reducing manual intervention and alleviating the workload of the professional review team. Second, it optimizes resource allocation. Due to the improved review efficiency, live streaming applications can be approved and launched more quickly, avoiding long-term idleness and waste of computing, storage, and network bandwidth resources caused by review delays. Third, it improves user experience and platform reputation. Live streaming applications from cultural inheritors and other content creators can be processed more quickly and fairly, enhancing their trust and satisfaction with the platform, helping to attract more high-quality cultural content and promoting the prosperity and development of the cultural ecosystem. Finally, this application provides a smarter and more refined operation and management approach, which can optimize processes and allocate resources more effectively from a global perspective, combining the characteristics of cultural content and the actual workload of the review team, thereby overcoming the limitations of traditional systems when facing the special field of cultural services.

[0084] In some of the embodiments described above in this application, the accuracy and effectiveness of the cultural knowledge base are crucial for assessing the cultural contextual rationality of live-streaming event applications. However, if the construction of the cultural knowledge base lacks systematicity and data support, it may lead to inaccurate cultural contextual rationality scores for risk indication information, thereby affecting the adjustment of the ranking weights in the review queue and the assignment of cultural exemption tags, reducing the efficiency and accuracy of the entire operational process. Therefore, this application further proposes a pre-construction method for the cultural knowledge base to ensure the reliability and usability of its content.

[0085] The aforementioned cultural knowledge base is pre-constructed in the following ways:

[0086] Retrieve multiple approved historical live streaming event applications, each of which is marked with its final review result;

[0087] For each historical live-streaming event application, extract the cultural theme keywords and risk indication information of the historical live-streaming event application;

[0088] For historical live streaming activity applications that have passed the review and contain risk indication information, the combination of cultural theme keywords and risk indication information in the historical live streaming activity application will be used as a cultural background record, and the rationality benchmark value of the cultural context will be initially set to 1.

[0089] For historical live streaming event applications that are rejected but whose risk indication information is marked as valid, the combination of cultural theme keywords and risk indication information in the historical live streaming event application is taken as a cultural background record, and the rationality benchmark value of the cultural context is initially set to 0.

[0090] All the constructed cultural background records are stored in the cultural knowledge base.

[0091] Specifically, the construction of the cultural knowledge base first requires acquiring a large number of approved historical live-streaming event applications. These applications have all undergone final review by humans or the system, and their review results are clearly marked as "approved" or "not approved." This historical data with clear review results forms the foundation for building the cultural knowledge base, providing a basis for judging real-world cultural contexts.

[0092] Subsequently, for each historical live-streaming event application received, the system processes its content to extract cultural theme keywords and risk indication information. Cultural theme keywords identify the cultural field or content involved in the live-streaming event, such as "traditional opera," "ethnic dance," or "historical lecture." Risk indication information refers to sensitive words or image features detected in the live-streaming event content. This information may constitute a risk in some contexts, but may be considered reasonable within a specific cultural context.

[0093] Furthermore, based on the review results of historical live-streaming event applications, the extracted cultural theme keywords and risk indication information are categorized and processed to generate cultural background records. Specifically, if a historical live-streaming event application is approved and contains risk indication information, it indicates that the risk indication information is acceptable within that specific cultural theme context. In this case, the combination of the cultural theme keywords and risk indication information of the historical live-streaming event application will be treated as a cultural background record, and its corresponding cultural context rationality benchmark value will be initially set to 1, indicating that the risk indication information is highly reasonable within that cultural context. Conversely, if a historical live-streaming event application is rejected, and the risk indication information contained therein is deemed a valid reason for the rejection, it indicates that the risk indication information is unacceptable within that cultural theme context. In this case, the combination of the cultural theme keywords and risk indication information of the historical live-streaming event application will also be treated as a cultural background record, but its corresponding cultural context rationality benchmark value will be initially set to 0, indicating that the risk indication information is not reasonable within that cultural context.

[0094] Finally, all cultural background records constructed in the above manner are stored uniformly in a cultural knowledge base. Each record in this cultural knowledge base includes a cultural theme field, a risk indication field, and a corresponding cultural context rationality benchmark value, providing data support for subsequent live-streaming event application review.

[0095] This application's solution utilizes a large amount of historical review data to pre-build a cultural knowledge base, enabling it to automatically learn and accumulate the rationality judgment criteria for risk indication information in different cultural contexts. By setting the cultural context rationality benchmark value of approved historical applications containing risk information to 1, and setting the cultural context rationality benchmark value of unapproved historical applications with valid risk information to 0, the cultural knowledge base can effectively encode and store actual cultural judgment standards. Therefore, in subsequent live-streaming event application reviews, when the system retrieves a matching cultural background record, it can directly use the pre-set cultural context rationality benchmark value, combined with the matching degree of the risk indication information, to generate a more accurate rationality score, avoiding the subjectivity and inaccuracy that may result from relying entirely on human experience or pre-set rules.

[0096] Through the aforementioned technical solutions, the construction process of the cultural knowledge base is automated and data-driven, significantly improving the accuracy and objectivity of the assessment of the rationality of cultural contexts. This method effectively captures and reflects the actual acceptance of sensitive content under different cultural themes, enabling the system to more intelligently and accurately identify and assess potential risks when processing applications for live-streaming events with cultural backgrounds. This not only improves review efficiency and reduces human intervention but also ensures the fairness and consistency of review results, thereby optimizing the overall operational process management of the cultural service live-streaming platform.

[0097] In some preferred embodiments, suppose the platform receives a live streaming application for a "Peking Opera Mask Art Performance". After text parsing and image recognition of the application, the system detects that it contains sensitive words such as "ghosts" and sensitive image features such as "fierce masks", which are identified as risk indication information. At this time, the system will extract "Peking Opera Mask Art Performance" as a cultural theme keyword. During the construction phase of the cultural knowledge base, if there are multiple approved live streaming applications related to "Peking Opera Masks" or "traditional opera" in the historical data, and these applications have also contained risk indication information such as "ghosts" or "fierce masks", then in the cultural knowledge base, the reasonableness benchmark value of the cultural background records of "ghosts" or "fierce masks" related to the cultural theme of "Peking Opera Masks" will be set to 1. Conversely, if a historical application for a live stream of a "Children's Storytelling Session" contains sensitive words such as "ghosts," but is ultimately rejected due to inappropriate content for children, then the baseline value for the reasonableness of the cultural background record of "ghosts" related to the "Children's Storytelling Session" cultural theme will be set to 0 in the cultural knowledge base. In this way, the cultural knowledge base can automatically learn and distinguish the reasonableness of "ghosts" in the context of a "Peking Opera Mask Art Performance" from the unreasonableness of "ghosts" in the context of a "Children's Storytelling Session," thus providing an accurate basis for subsequent reasonableness scoring.

[0098] Specifically, in the above method, the calculation of string similarity can be performed using the edit distance algorithm.

[0099] According to the above method, the string similarity is calculated using the edit distance algorithm:

[0100] Similarity = 1 - (edit distance / max(length of first string, length of second string));

[0101] The edit distance is the minimum number of single-character edit operations required to convert the risk indication information into the risk indication field, and the edit operations include insertion, deletion, and replacement.

[0102] Specifically, the edit distance algorithm is a method for measuring the difference between two strings. Its core idea is to calculate the minimum number of single-character edit operations required to transform one string into another. These edit operations typically include inserting a character, deleting a character, or replacing a character. For example, transforming the string "kitten" into "sitting" requires three edit operations (k->s, e->i, n->g), therefore the edit distance is 3. In the similarity calculation formula above, "first string length" refers to the length of the risk indication information, and "second string length" refers to the length of the risk indication field in the cultural background record. By dividing the edit distance by the length of the longer of the two strings, a normalized distance value can be obtained. Subtracting this value from 1 yields a similarity between 0 and 1, where 1 represents identical strings and 0 represents completely different strings.

[0103] This application employs an edit distance algorithm to calculate the string similarity between risk indication information and risk indication fields in cultural background records, accurately quantifying the degree of difference between the two strings. The edit distance algorithm not only identifies perfectly matching words but also effectively handles minor differences caused by spelling errors, added or deleted characters, or replaced characters, thus reflecting to some extent the semantic similarity of sensitive words or image features in different cultural contexts. This calculation method enables the system to more accurately assess the correlation between risk indication information and known cultural background, providing reliable matching data for subsequent reasonableness scoring.

[0104] By employing the edit distance algorithm to calculate string similarity using the aforementioned technical solution, the accuracy and robustness of matching risk indication information with risk indication fields in cultural background records can be effectively improved. Compared to simple exact matching, the edit distance algorithm can tolerate a certain degree of character differences, allowing the system to still identify potential cultural relevance even when faced with variations, typos, or minor modifications to sensitive words or image features, avoiding misjudgments due to subtle differences. Consequently, the generated reasonableness score will be more refined and objective, helping the platform more accurately determine the cultural contextual reasonableness of potential risks in live-streaming event applications, thereby improving review efficiency and decision-making quality.

[0105] This application further proposes a method for calculating string similarity, which uses a cosine similarity algorithm for calculation, specifically including:

[0106] The risk indication information and the risk indication field are respectively processed by word segmentation to obtain a first term set and a second term set;

[0107] Construct a vector space with all terms as the dimension, and map the first term set and the second term set into a first vector and a second vector, respectively;

[0108] Similarity = (first vector · second vector) / (|first vector| × |second vector|).

[0109] Specifically, word segmentation refers to dividing a continuous text sequence into word units with independent semantic meaning. For example, for Chinese text, word segmentation tools based on dictionaries, statistics, or deep learning can be used to process risk indication information and risk indication fields to obtain their respective term sets. The first term set corresponds to the segmentation result of the risk indication information, and the second term set corresponds to the segmentation result of the risk indication field.

[0110] Furthermore, constructing a vector space with all terms as its dimension means using all independent terms appearing in the first and second term sets as the dimension of the vector space. The value of each term in the vector can be determined based on its frequency of occurrence in the corresponding text (e.g., term frequency TF) or its importance in the entire corpus (e.g., inverse document frequency IDF), forming a bag-of-words model or a TF-IDF model. Thus, the first term set and the second term set are mapped to the first and second vectors in the vector space, respectively.

[0111] The similarity score, calculated as (first vector × second vector) / (|first vector| × |second vector|), is obtained by calculating the dot product of the two vectors and dividing by the product of their respective magnitudes. The dot product measures the similarity of the two vectors in a direction, while the magnitude represents the length or strength of the vectors. In this way, cosine similarity quantifies the semantic closeness of two texts, with a value between -1 and 1. A value closer to 1 indicates greater semantic similarity, closer to -1 indicates less semantic similarity, and a value close to 0 indicates independence.

[0112] This application's solution utilizes a cosine similarity algorithm to calculate string similarity, effectively addressing the shortcomings of traditional simple string matching in handling semantic complexity. Through this technical solution, the application significantly improves the accuracy and robustness of similarity calculation between risk indication information and risk indication fields in cultural background records. Compared to methods relying solely on character matching, the cosine similarity algorithm better captures deep semantic relationships within the text, effectively handling semantic consistency issues related to synonyms, near-synonyms, and different expressions. Consequently, the rationality score of the generated risk indication information within the cultural context becomes more accurate, leading to a more reasonable adjustment of the ranking weight in the review queue for live-streaming event applications. Ultimately, this improves the accuracy of granting cultural exemption markers and optimizes the intelligence level and review efficiency of the entire cultural service live-streaming platform's operational process management.

[0113] In some of the embodiments described above in this application, the ranking weight of live event applications in the review queue is dynamically adjusted based on the rationality score. However, if the adjustment method of the ranking weight is not carefully designed during its implementation, the weight adjustment may lack stability and predictability, or even result in the weight growing without limit, thereby affecting the fairness and efficiency of the review queue.

[0114] In response, this application further proposes a dynamic adjustment method for the ranking weights as follows:

[0115] Set the baseline weight W0;

[0116] Calculate the weight increment ΔW = k × S, where S is the rationality score and k is a preset proportional coefficient, the value of which is between 0 and 1;

[0117] The sorting weight W = W0 + ΔW;

[0118] When the sorting weight W exceeds the preset maximum weight threshold, W is truncated to the maximum weight threshold.

[0119] Specifically, the baseline weight W0 refers to an initial or minimum ranking weight value set for all live event applications. Its purpose is to ensure that even with a low reasonableness score, an application receives a basic priority, preventing it from being indefinitely shelved. The weight increment ΔW is calculated by multiplying the reasonableness score S by a preset proportional coefficient k. The reasonableness score S is generated based on the reasonableness of the risk indication information within the cultural context. The proportional coefficient k can be understood as a factor used to adjust the degree of influence of the reasonableness score on the ranking weight, with a value ranging from 0 to 1, allowing the system to flexibly control the contribution of cultural context reasonableness to priority based on actual operational needs. Therefore, the final ranking weight W is determined as the sum of the baseline weight W0 and the weight increment ΔW. Furthermore, to prevent the ranking weight W from growing indefinitely or exceeding a reasonable range, this application sets a maximum weight threshold. When the calculated ranking weight W exceeds this threshold, its value will be truncated to the maximum weight threshold. This aims to maintain the fairness of the review queue, prevent a few high-scoring applications from occupying the top of the queue for extended periods, and ensure that all applications have a chance to be processed.

[0120] This application's solution constructs a structured and controlled dynamic adjustment mechanism for ranking weights by introducing a baseline weight, incremental calculation based on reasonableness scores, and a maximum weight threshold. The baseline weight W0 provides a fair starting point for all applications, ensuring basic processing opportunities. The weight increment ΔW allows applications for live-streaming events with higher cultural context reasonableness scores to receive corresponding priority increases, thereby accelerating their review process. The introduction of a proportional coefficient k allows the system to flexibly adjust the degree of influence of reasonableness scores on the final ranking weights to adapt to different operational strategies and review pressures. More importantly, by setting a maximum weight threshold and implementing truncation, the problem of unlimited growth in ranking weights is effectively avoided, ensuring the stability and fairness of the review queue, preventing a few high-scoring applications from excessively crowding out review resources, and thus optimizing the overall review process management.

[0121] The above technical solution provides a more refined, stable, and fair method for adjusting the ranking weights of live-streaming event application review queues. This method not only effectively utilizes cultural context rationality scoring to accelerate applications that align with the cultural background, but also avoids the uncertainty and imbalance that may arise from weight adjustments by setting baseline weights and maximum weight thresholds, significantly improving the efficiency and fairness of the review process.

[0122] In some preferred embodiments, it is assumed that the baseline weight W0 is set to 50, the scaling factor k is set to 0.5, and the maximum weight threshold is set to 100. When a live event application has a cultural context rationality score S of 0.8, its weight increment ΔW will be calculated as 0.5 × 0.8 = 0.4. At this time, the ranking weight W of the application will be 50 + 0.4 = 50.4. If another live event application has a cultural context rationality score S of 0.2, its weight increment ΔW is 0.5 × 0.2 = 0.1, and the ranking weight W is 50 + 0.1 = 50.1. This shows that applications with higher rationality scores have higher ranking weights, but the increment is controlled. Further, if an application has a baseline weight W0 of 99.8 and a rationality score S of 0.8, the weight increment ΔW remains 0.4, and the calculated ranking weight W will be 99.8 + 0.4 = 100.2. Since the value exceeds the preset maximum weight threshold of 100, the final ranking weight of this application will be truncated to 100. In this way, the system can ensure that applications with high reasonableness scores are given priority processing, while effectively preventing the weight of any application from growing indefinitely, thereby maintaining the dynamic balance and fairness of the review queue.

[0123] In some of the embodiments described above in this application, a method is proposed to score the reasonableness of live-streaming event applications based on a cultural knowledge base and dynamically adjust the ranking weight of the review queue. However, in actual operation, the cultural context reasonableness benchmark value in the cultural knowledge base may gradually lose accuracy due to changes in the cultural context or initial setting deviations, leading to biases in the system's judgment of the cultural context of risky content, thereby affecting review efficiency and accuracy. If the above problems are not addressed, the long-term stability and adaptability of the system will be limited.

[0124] In this regard, this application further proposes that the above method also includes an automatic updating step for the cultural knowledge base:

[0125] Record the actual review results of each live streaming event application after the review is completed;

[0126] If the actual review result is inconsistent with the expected review result corresponding to the reasonableness score, the cultural knowledge base will be updated.

[0127] The update method is to reduce the confidence level of the cultural context rationality benchmark value in the cultural background record and accumulate the number of inconsistencies;

[0128] When the number of inconsistencies exceeds a preset correction threshold, the cultural context rationality benchmark value is adjusted by a step size in the direction corresponding to the actual audit result, where the step size is 0.1.

[0129] Specifically, "recording the actual review result of the live streaming activity application after each review" means that after the system completes the manual review or further automated review process, it associates and stores the final review conclusion (e.g., "passed" or "failed") with the live streaming activity application. The actual review result can be a Boolean value (e.g., true for pass, false for fail) or a multi-level classification result.

[0130] The "expected review result corresponding to the reasonableness score" refers to the review result predicted by the system based on the reasonableness score of the live streaming event application. For example, when the reasonableness score is higher than a certain preset threshold, the expected review result is judged as "passed"; otherwise, it is judged as "failed".

[0131] In practical applications, "inconsistency" refers to a difference between the actual audit result and the expected audit result. For example, the system might expect a "pass" based on the reasonableness score, but the actual audit result might be "fail," or the system might expect a "fail," but the actual audit result might be "pass."

[0132] Furthermore, "triggering the update of the cultural knowledge base" means that when an inconsistency is detected, the system starts a background process or calls a specific update function to adjust the relevant cultural background records.

[0133] Specifically, "reducing the confidence level of the cultural context rationality benchmark value in the cultural background record" means that in the cultural knowledge base, each cultural background record, in addition to containing the cultural context rationality benchmark value, can also store an additional confidence level parameter. When inconsistency occurs, this confidence level parameter will be appropriately reduced, indicating that the reliability of the current benchmark value has decreased.

[0134] The "cumulative number of inconsistencies" refers to the number of times the actual review result of the live streaming application related to that record is inconsistent with the expected review result.

[0135] In a preferred implementation, the "correction threshold" is a preset integer value used to control the sensitivity of adjusting the cultural context rationality benchmark. When the number of inconsistencies reaches or exceeds this threshold, it indicates that the current benchmark may have a persistent deviation and requires substantial adjustment.

[0136] Specifically, "adjusting the cultural context rationality benchmark value by one step in the direction corresponding to the actual review result" means that if the actual review result is more stringent than expected (e.g., expected to pass but actually failed), the benchmark value will be adjusted to lower the rationality score; if the actual review result is more lenient than expected (e.g., expected to fail but actually passed), the benchmark value will be adjusted to increase the rationality score. The "step" is a preset fixed value, such as 0.1, used to control the magnitude of each adjustment, ensuring the stability and gradualness of the adjustment process.

[0137] This application's solution effectively addresses the issue of cultural context rationality benchmarks potentially losing accuracy over time or due to cultural evolution by introducing an automatic update mechanism for the cultural knowledge base. Through this technical solution, the application achieves dynamic adaptive updates to the cultural knowledge base, significantly improving the long-term accuracy and robustness of the operational process management methods for cultural service live-streaming platforms.

[0138] This application further proposes review efficiency monitoring steps to ensure the continued effectiveness of the accelerated review mechanism.

[0139] The above method also includes a review efficiency monitoring step:

[0140] Record the average processing time T_acc for each live event application from receipt to completion of review when it is added to the accelerated processing thread pool;

[0141] Record the average processing time T_normal from receipt to completion of review for each live event application that is not included in the accelerated processing thread pool;

[0142] Calculate the speedup efficiency ratio R = T_normal / T_acc;

[0143] When the acceleration efficiency ratio R is lower than the preset efficiency maintenance threshold, the acceleration threshold is lowered by 5% of the current acceleration threshold.

[0144] Specifically, the review efficiency monitoring step refers to the system continuously tracking and quantifying the actual performance of the accelerated processing mechanism. The average processing time T_acc refers to the average time taken from receiving a live event application with a cultural exemption tag and routed to the accelerated processing thread pool to the completion of its review. The average processing time T_normal refers to the average time taken from receiving a live event application that was not routed to the accelerated processing thread pool, i.e., reviewed according to the regular process, to the completion of its review. These times can be automatically calculated by the system by recording the application's reception timestamp and review completion timestamp.

[0145] Furthermore, the acceleration efficiency ratio R is defined as the ratio of T_normal to T_acc, i.e., R = T_normal / T_acc. This ratio directly reflects the efficiency improvement brought by the accelerated processing thread pool compared to the regular review process. For example, if T_normal is 10 hours and T_acc is 2 hours, then R is 5, indicating that accelerated processing reduces the review time by 5 times.

[0146] When the calculated acceleration efficiency ratio R is lower than the preset efficiency maintenance threshold, it indicates that the efficiency improvement effect of the accelerated review mechanism has not met expectations or has declined. In this case, the system will trigger a reduction in the acceleration threshold. The reduction amount is 5% of the current acceleration threshold. For example, if the current acceleration threshold is 0.8, it will be reduced to 0.8 * (1 - 0.05) = 0.76. By lowering the acceleration threshold, more live streaming activity applications will be able to meet the acceleration conditions and be routed to the accelerated processing thread pool, aiming to improve overall review efficiency again. The efficiency maintenance threshold can be set according to actual operational needs and historical data. For example, it can be set to 2, indicating that the accelerated processing is expected to at least halve the review time.

[0147] This application's solution addresses the efficiency degradation issue that can arise from a fixed acceleration threshold in the basic solution by introducing a review efficiency monitoring step. Through this technical solution, the application enables dynamic optimization and adaptive adjustment of the accelerated review mechanism within the operational process management method of a cultural service live streaming platform. This solution overcomes the efficiency bottlenecks that traditional fixed threshold mechanisms may encounter when facing dynamically changing workloads and review resources, ensuring the accelerated review mechanism maintains high efficiency over the long term. Specifically, by monitoring the acceleration efficiency ratio in real time and judging based on preset thresholds, the system can promptly detect and correct the degradation of acceleration effects, avoiding resource waste or review backlogs caused by inappropriate acceleration thresholds. This adaptive adjustment capability significantly improves the intelligence level and response speed of the entire live streaming platform's review process, enabling the platform to more flexibly respond to various operational challenges and continuously optimize user experience and content compliance.

[0148] In some preferred embodiments, a specific example is given below. Suppose a cultural service live streaming platform initially sets the acceleration threshold to 0.7 and the efficiency maintenance threshold to 2.5. After the system has been running for a period of time, it begins recording review data.

[0149] Within a monitoring period, the system statistics show that:

[0150] The average processing time T_acc for live event applications that are added to the accelerated processing thread pool is 2 hours.

[0151] The average processing time T_normal for live event requests that are not included in the accelerated processing thread pool is 6 hours.

[0152] At this point, the calculated acceleration efficiency ratio R = T_normal / T_acc = 6 hours / 2 hours = 3.

[0153] Since R(3) is greater than the preset efficiency maintenance threshold (2.5), it indicates that the accelerated review mechanism is working well and the efficiency improvement effect is significant. The system will maintain the current acceleration threshold unchanged.

[0154] However, during another monitoring period, with the increase in applications for live streaming events and changes in the number of reviewers, the system statistics showed that:

[0155] The average processing time T_acc for live event applications that are added to the accelerated processing thread pool is 3 hours.

[0156] The average processing time T_normal for live event requests that are not included in the accelerated processing thread pool is 6 hours.

[0157] At this point, the calculated acceleration efficiency ratio R = T_normal / T_acc = 6 hours / 3 hours = 2.

[0158] Since R(2) is lower than the preset efficiency maintenance threshold (2.5), the system determines that the accelerated review efficiency has decreased and has failed to achieve the expected effect. Therefore, the system will trigger a reduction in the acceleration threshold. If the current acceleration threshold is 0.7 and the reduction is 5% of the current acceleration threshold, the new acceleration threshold will be adjusted to 0.7 * (1 - 0.05) = 0.665. By reducing the acceleration threshold, more live streaming applications will be able to meet the acceleration conditions and be routed to the accelerated processing thread pool, in order to improve the overall review efficiency and make the acceleration efficiency ratio R rise back above the efficiency maintenance threshold.

[0159] In some of the embodiments described above in this application, content parsing and image recognition are performed directly upon receiving a live streaming event application. However, in actual operation, some submitted live streaming event applications may contain incomplete information, such as missing titles, descriptions, cover images, or appointment time information. Directly performing subsequent risk identification and cultural theme extraction on these incomplete applications would not only increase the system's processing burden but may also lead to interruptions in the review process or inaccurate results, thereby affecting overall review efficiency and user experience.

[0160] In response, this application further proposes that before performing text parsing and image recognition on the content of the aforementioned live event application, the following steps are also included: performing format verification on the live event application to determine whether the live event application contains complete title, description, cover image, and appointment time information; if any of the necessary information is missing, the live event application is routed to the information completion queue, and an information completion request is returned to the initiator.

[0161] Specifically, format verification refers to a structural check of the key metadata included in the submitted live event application. This key metadata typically includes, but is not limited to, the live event title, detailed description, cover image, and scheduled live event time. This information is fundamental to ensuring that the live event application can be effectively understood and processed. The title summarizes the live event content, the description provides detailed background and flow, the cover image is used for visual appeal and initial content assessment, and the scheduled time information relates to the live event scheduling and resource allocation. If, during format verification, any of the above necessary information is found to be missing from the live event application—for example, an empty title, an overly short description, or no cover image—the application will not immediately proceed to the subsequent risk identification process. Instead, the application will be routed to a dedicated information completion queue. Simultaneously, the system will return an information completion request to the user or organization that initiated the live event application, clearly indicating the missing information and guiding them to submit the missing information. The information completion queue aims to centrally manage all applications suspended due to incomplete information for unified follow-up and processing.

[0162] This application's solution introduces a format verification step before live event applications enter the core processing flow, effectively identifying and blocking incomplete applications. When necessary information is detected as missing, the system no longer forces subsequent text parsing and image recognition; instead, it guides the application to an information completion queue and promptly notifies the initiator to supplement the information. This mechanism avoids invalid processing of incomplete data, thereby reducing the waste of computing resources and preventing subsequent processing errors or interruptions due to missing information. By separating incomplete applications from the main review process and providing clear completion guidelines, it ensures that only applications meeting basic requirements can enter the subsequent intelligent review stage, improving the smoothness and accuracy of the entire review process.

[0163] Through the aforementioned technical solution, this application significantly improves the preprocessing efficiency and accuracy of live-streaming event applications. Firstly, pre-processing format verification effectively filters out ineligible applications, avoiding the need for complex algorithms to process invalid data and thus saving system resources. Secondly, the introduction of an information completion queue and completion requests allows initiators to promptly identify and correct issues, reducing repeated communication and review delays caused by incomplete information and optimizing the user experience. Finally, it ensures that all applications entering the intelligent review process possess complete and standardized information, providing a high-quality data foundation for subsequent risk identification and cultural context rationality assessment, thereby improving the automation level and decision-making reliability of the overall review process.

[0164] In some embodiments of this application, when the ranking weight of a live event application is greater than or equal to a preset acceleration threshold, the application is given a cultural exemption tag and routed to a preset accelerated processing thread pool for rapid review. However, during implementation, if the load of the accelerated processing thread pool is not effectively managed, the thread pool may become saturated, preventing new applications from entering in a timely manner and even affecting the processing efficiency of existing tasks. If this problem is not addressed, the advantages of accelerated processing will not be fully realized, and may even lead to a decline in overall system performance. Therefore, this application further proposes a scheme to optimize system resource utilization by monitoring the thread pool load and dynamically adjusting the acceleration threshold after routing live event applications to the preset accelerated processing thread pool.

[0165] After routing live streaming event requests to a pre-defined accelerated processing thread pool, the process also includes:

[0166] Monitor the current load rate of the accelerated processing thread pool;

[0167] When the current load rate exceeds the preset load threshold, the value of the acceleration threshold is temporarily increased to reduce the number of live event applications newly added to the acceleration processing thread pool.

[0168] When the current load rate is lower than the load threshold and continues to exceed the preset duration threshold, the acceleration threshold is restored to its original value.

[0169] Specifically, monitoring the current load rate of the accelerated processing thread pool refers to the system continuously tracking and evaluating the running status of the accelerated processing thread pool. This can be done by statistically analyzing metrics such as the number of tasks currently being processed, CPU utilization, memory usage, or queue length to determine its workload. This load rate can be understood as the proportion of thread pool resources being utilized, and its purpose is to monitor the real-time health of the thread pool.

[0170] Specifically, when the current load rate exceeds a preset load threshold, the system temporarily increases the acceleration threshold. The preset load threshold is an upper limit, representing the maximum load level at which the thread pool can operate efficiently. Once the actual load exceeds this threshold, it means the thread pool may face overload risk. At this time, by temporarily increasing the acceleration threshold, only live streaming event applications with higher rationality scores (i.e., higher sorting weights) can be routed to the accelerated processing thread pool, thereby reducing the number of new applications entering the accelerated processing thread pool in the short term and alleviating its processing pressure.

[0171] In practical applications, when the current load rate is below the load threshold and continues to exceed the preset duration threshold, the system will restore the acceleration threshold to its original value. The duration threshold is used to ensure that the low load state of the thread pool is not a short-term fluctuation, but a relatively stable one. When the thread pool load recovers to an acceptable level and remains so for a period of time, the system determines that it has the capacity to process more acceleration requests, and will then adjust the acceleration threshold back to its initial setting to ensure that more eligible live streaming event applications can enjoy the convenience of accelerated review.

[0172] This application's solution effectively addresses the thread pool overload issue that may occur in the basic solution by introducing a real-time monitoring mechanism for the accelerated processing thread pool load. Through this technical solution, the application can intelligently adjust the accelerated review strategy based on the actual operating load of the accelerated processing thread pool. This not only effectively avoids the risk of performance degradation or system crashes due to thread pool overload, ensuring the review efficiency of high-priority applications, but also allows for timely restoration of the accelerated review scope when the load returns to normal, maximizing the utilization of system resources. Compared to the basic solution, this application, by introducing a load adaptive adjustment mechanism, significantly improves the robustness and efficiency of the cultural service live streaming platform's operational process management, ensuring stable operation and efficient service under different business pressures.

[0173] refer to Figure 2 , Figure 2 This is a schematic diagram of the operation process management system for a cultural service live streaming platform proposed in this application, including:

[0174] The risk identification module is used to receive live event applications, perform text parsing and image recognition on the content of the live event applications, extract sensitive words from a preset sensitive word library and sensitive image features from a preset sensitive image library contained in the content, and use the extracted sensitive words and sensitive image features as risk indication information.

[0175] The theme extraction module is used to extract cultural theme keywords from the content of the live broadcast event application in response to the risk indication information extracted by the risk identification module.

[0176] The retrieval module is used to retrieve matching cultural background records from the cultural knowledge base based on the cultural theme keywords extracted by the theme extraction module. Each cultural background record in the cultural knowledge base includes a cultural theme field, a risk indication field, and a corresponding cultural context rationality benchmark value.

[0177] The similarity calculation module is used to calculate the string similarity between the risk indication information and the risk indication field in the cultural background record retrieved by the retrieval module, and use the calculated string similarity as the matching degree.

[0178] The rating generation module is used to multiply the matching degree calculated by the similarity calculation module with the cultural context rationality benchmark value in the cultural background record to generate a rationality score of the risk indication information in the cultural context.

[0179] The weight adjustment module is used to dynamically adjust the ranking weight of the live event application in the review queue based on the rationality score generated by the scoring generation module, wherein the ranking weight is positively correlated with the rationality score;

[0180] The routing module is used to assign a cultural exemption flag to the live streaming activity application and route the live streaming activity application to a preset acceleration processing thread pool when the sorting weight adjusted by the weight adjustment module is greater than or equal to a preset acceleration threshold.

[0181] This system aims to address the problems of low review efficiency, resource waste, and poor user experience in the operational process management of existing cultural service live streaming platforms due to a lack of in-depth understanding of the cultural context. By modularizing functions such as risk identification of live streaming application, cultural theme extraction, cultural background retrieval, similarity calculation, reasonableness score generation, review weight adjustment, and accelerated routing, this system enables intelligent and refined management of cultural service live streaming content. The various modules work collaboratively to form a complete processing chain, ensuring efficient and accurate operation of the entire process from application receipt to final review and routing. Especially when handling live streaming content with specific cultural backgrounds, it effectively avoids misjudgments, improving overall operational efficiency and user satisfaction.

[0182] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A cultural service live broadcast platform operation process management method, characterized in that, The method comprises the following steps: receiving a live broadcast activity application, and performing text analysis and image recognition on the content of the live broadcast activity application, extracting sensitive words in a preset sensitive word library and sensitive image features in a preset sensitive image library contained in the content, and taking the extracted sensitive words and sensitive image features as risk indication information; extracting cultural theme keywords from the content of the live broadcast activity application in response to the extracted risk indication information; retrieving a matched cultural background record from a cultural knowledge base according to the extracted cultural theme keywords, each cultural background record in the cultural knowledge base containing a cultural theme field, a risk indication field and a corresponding cultural context rationality benchmark value; calculating the string similarity between the risk indication information and the risk indication field in the retrieved cultural background record, and taking the calculated string similarity as a matching degree; multiplying the matching degree by the cultural context rationality benchmark value in the cultural background record to generate a rationality score of the risk indication information in a cultural context; dynamically adjusting the sorting weight of the live broadcast activity application in an audit queue according to the rationality score, wherein the sorting weight is positively correlated with the rationality score; when the adjusted sorting weight is greater than or equal to a preset acceleration threshold, assigning a cultural exemption mark to the live broadcast activity application, and routing the live broadcast activity application to a preset acceleration processing thread pool.

2. The method of claim 1, wherein, The cultural knowledge base is pre-constructed in the following manner: obtaining a plurality of historical live broadcast activity applications that have been audited and passed, each of the historical live broadcast activity applications being marked with a final audit result; for each of the historical live broadcast activity applications, extracting cultural theme keywords and risk indication information of the historical live broadcast activity application; for the historical live broadcast activity application whose audit result is pass and which contains risk indication information, taking the combination of the cultural theme keywords and risk indication information of the historical live broadcast activity application as a cultural background record, and initially setting the cultural context rationality benchmark value as 1; for the historical live broadcast activity application whose audit result is not pass and whose risk indication information is marked as valid, taking the combination of the cultural theme keywords and risk indication information of the historical live broadcast activity application as a cultural background record, and initially setting the cultural context rationality benchmark value as 0; storing all the constructed cultural background records in the cultural knowledge base.

3. The method of claim 1, wherein, The string similarity is calculated by using an edit distance algorithm: similarity = 1 - (edit distance / max (length of the first string, length of the second string)); wherein the edit distance is the minimum number of single-character edit operations required to convert the risk indication information into the risk indication field, and the edit operations include insertion, deletion and replacement.

4. The method of claim 1, wherein, The string similarity is calculated by using a cosine similarity algorithm, comprising: performing word segmentation on the risk indication information and the risk indication field respectively to obtain a first term set and a second term set; Construct a vector space with all terms as the dimension, and map the first term set and the second term set into a first vector and a second vector, respectively; Similarity = (first vector · second vector) / (|first vector| × |second vector|).

5. The method of claim 1, wherein, The dynamic adjustment method for the ranking weights is as follows: Set the baseline weight W0; Calculate the weight increment ΔW = k × S, where S is the rationality score and k is a preset proportional coefficient, the value of which is between 0 and 1; The sorting weight W = W0 + ΔW; When the sorting weight W exceeds the preset maximum weight threshold, W is truncated to the maximum weight threshold.

6. The method of claim 1, wherein, The method also includes an automatic update step for the cultural knowledge base: Record the actual review results of each live streaming event application after the review is completed; If the actual review result is inconsistent with the expected review result corresponding to the reasonableness score, the cultural knowledge base will be updated. The update method is to reduce the confidence level of the cultural context rationality benchmark value in the cultural background record and accumulate the number of inconsistencies; When the number of inconsistencies exceeds a preset correction threshold, the cultural context rationality benchmark value is adjusted by a step size in the direction corresponding to the actual audit result, where the step size is 0.

1.

7. The method of claim 1, wherein, The method also includes an audit efficiency monitoring step: Record the average processing time T_acc for each live event application from receipt to completion of review when it is added to the accelerated processing thread pool; Record the average processing time T_normal from receipt to completion of review for each live event application that is not included in the accelerated processing thread pool; Calculate the speedup efficiency ratio R = T_normal / T_acc; When the acceleration efficiency ratio R is lower than the preset efficiency maintenance threshold, the acceleration threshold is lowered by 5% of the current acceleration threshold.

8. The method of claim 1, wherein, Before performing text parsing and image recognition on the content of the live streaming event application, the method further includes: The format of the live streaming event application is validated to determine whether the application contains a complete title, description, cover image, and appointment time information. If any necessary information is missing, the live event application will be routed to the information completion queue, and an information completion request will be returned to the initiator.

9. The method of claim 1, wherein, After routing the live streaming activity request to a preset accelerated processing thread pool, the method further includes: Monitor the current load rate of the accelerated processing thread pool; When the current load rate exceeds the preset load threshold, the value of the acceleration threshold is temporarily increased to reduce the number of live event applications newly added to the acceleration processing thread pool. When the current load rate is lower than the load threshold and continues to exceed the preset duration threshold, the acceleration threshold is restored to its original value.

10. A cultural service live broadcast platform operation process management system, characterized in that, include: The risk identification module is used to receive live event applications, perform text parsing and image recognition on the content of the live event applications, extract sensitive words from a preset sensitive word library and sensitive image features from a preset sensitive image library contained in the content, and use the extracted sensitive words and sensitive image features as risk indication information. The theme extraction module is used to extract cultural theme keywords from the content of the live broadcast event application in response to the risk indication information extracted by the risk identification module. The retrieval module is used to retrieve matching cultural background records from the cultural knowledge base based on the cultural theme keywords extracted by the theme extraction module. Each cultural background record in the cultural knowledge base includes a cultural theme field, a risk indication field, and a corresponding cultural context rationality benchmark value. The similarity calculation module is used to calculate the string similarity between the risk indication information and the risk indication field in the cultural background record retrieved by the retrieval module, and use the calculated string similarity as the matching degree. The rating generation module is used to multiply the matching degree calculated by the similarity calculation module with the cultural context rationality benchmark value in the cultural background record to generate a rationality score of the risk indication information in the cultural context. The weight adjustment module is used to dynamically adjust the ranking weight of the live event application in the review queue based on the rationality score generated by the scoring generation module, wherein the ranking weight is positively correlated with the rationality score; The routing module is used to assign a cultural exemption flag to the live streaming activity application and route the live streaming activity application to a preset acceleration processing thread pool when the sorting weight adjusted by the weight adjustment module is greater than or equal to a preset acceleration threshold.

Images