Anchor management method and system based on data analysis

Abstract

The invention discloses an anchor management method and system based on data analysis, and relates to the field of intelligent management, and the method comprises the steps: extracting user interaction information from a live broadcast platform to obtain a set of user interaction information, carrying out the semantic coding of each piece of user interaction information through employing a data analysis and processing technology based on artificial intelligence, and carrying out the semantic coding of all pieces of user interaction information; and then performing medium-granularity semantic enhancement on the semantic coding features of the user interaction information, thereby intelligently generating anchor content optimization suggestions according to graph migration semantic feature aggregation representation among the semantic coding features of the medium-granularity semantic enhancement user interaction information. By collecting and analyzing interaction information from different audience groups, the application can more widely and deeply understand the demands of the audience. And moreover, interactive information can be deeply mined, so that diversified requirements can be better met.

Description

Technical Field

[0001] The present application relates to the field of intelligent management, and more specifically, to an anchor management method and system based on data analysis. Background Art

[0002] In recent years, with the rapid development of Internet technology, the live broadcast industry has become an emerging and dynamic way of content dissemination. Live broadcasters interact with viewers in real time through live broadcast platforms and share various content, including but not limited to entertainment, education, shopping, sports events, etc. As the live broadcast market expands, how to effectively manage and optimize live broadcast content has become an urgent problem to be solved. Live broadcasters need to constantly adjust and improve their content strategies to maintain audience interest and improve the level of interaction.

[0003] In traditional live broadcast management, anchors mainly rely on personal experience to plan content and design interactions, which has great limitations. Specifically, personal experience is often limited to the anchor's own background and experience, and it is difficult to cover the diverse needs of the audience. In addition, the lack of systematic and data-supported empirical methods cannot accurately quantify audience preferences and behavior patterns, and it is difficult to deeply explore the potential value of interactive information, which limits content innovation and real-time optimization, thereby affecting the quality of live broadcast content and the long-term engagement of the audience.

[0004] Therefore, an anchor management solution based on data analysis is desired. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed.

[0006] According to one aspect of the present application, a method for managing anchors based on data analysis is provided, which includes: Extracting user interaction information from the live broadcast platform to obtain a collection of user interaction information; Performing semantic coding on each piece of user interaction information in the set of user interaction information to obtain a set of semantic coding features of the user interaction information; The set of user interaction information semantic coding features is subjected to text set local semantic enhancement to obtain a set of medium-granularity semantically enhanced user interaction information semantic coding features, including: calculating the grammatical depth implicit representation value of each user interaction information semantic coding feature in the set of user interaction information semantic coding features to obtain a set distribution of user interaction information semantic grammatical depth implicit representation values; based on the set distribution of user interaction information semantic grammatical depth implicit representation values, medium-granularity contextual semantic enhancement is performed on each user interaction information semantic coding feature in the set of user interaction information semantic coding features to obtain the set of medium-granularity semantically enhanced user interaction information semantic coding features; The set of medium-granularity semantically enhanced user interaction information semantic coding features is subjected to graph walk adaptive semantic feature aggregation to obtain user interaction information semantic aggregation features, including: calculating graph walk topological features of the set of medium-granularity semantically enhanced user interaction information semantic coding features; based on the graph walk topological features, the set of medium-granularity semantically enhanced user interaction information semantic coding features is subjected to autocorrelation significant aggregation to obtain the user interaction information semantic aggregation features; Based on the semantic aggregation features of the user interaction information, anchor content optimization suggestions are obtained.

[0007] According to another aspect of the present application, a host management system based on data analysis is provided, which includes: A user interaction information extraction module, used to extract user interaction information from the live broadcast platform to obtain a collection of user interaction information; A user interaction information semantic coding module, used for semantically coding each user interaction information in the set of user interaction information to obtain a set of user interaction information semantic coding features; A user interaction information enhancement module, used for performing text set local semantic enhancement on the set of user interaction information semantic coding features to obtain a set of medium-granularity semantically enhanced user interaction information semantic coding features, wherein the user interaction information enhancement module comprises: a deep implicit representation value calculation unit, used for calculating the grammatical depth implicit representation value of each user interaction information semantic coding feature in the set of user interaction information semantic coding features to obtain a set distribution of user interaction information semantic grammatical depth implicit representation values; a medium-granularity semantic enhancement unit, used for performing medium-granularity contextual semantic enhancement on each user interaction information semantic coding feature in the set of user interaction information semantic coding features based on the set distribution of the user interaction information semantic grammatical depth implicit representation values ​​to obtain the set of medium-granularity semantically enhanced user interaction information semantic coding features; A user interaction information semantic aggregation module, used for performing graph walk adaptive semantic feature aggregation on the set of medium-granularity semantically enhanced user interaction information semantic coding features to obtain user interaction information semantic aggregation features, wherein the user interaction information semantic aggregation module includes: a graph walk topology calculation unit, used for calculating the graph walk topology features of the set of medium-granularity semantically enhanced user interaction information semantic coding features; an autocorrelation aggregation unit, used for performing autocorrelation significant aggregation on the set of medium-granularity semantically enhanced user interaction information semantic coding features based on the graph walk topology features to obtain the user interaction information semantic aggregation features; The anchor content optimization suggestion generation module is used to obtain anchor content optimization suggestions based on the semantic aggregation features of the user interaction information.

[0008] Compared with the prior art, the present application provides a method and system for anchor management based on data analysis, which extracts user interaction information from the live broadcast platform to obtain a collection of user interaction information, uses artificial intelligence-based data analysis and processing technology to semantically encode each user interaction information, and then performs medium-granular semantic enhancement on the semantic encoding features of each user interaction information, thereby intelligently generating anchor content optimization suggestions based on the graph walk semantic feature aggregation representation between the semantic encoding features of each medium-granular semantic enhancement user interaction information. By collecting and analyzing interactive information from different audience groups, the present application can transcend the limitations of the anchor's personal background and experience, and understand the audience's needs more broadly and deeply. In addition, it can deeply mine interactive information, identify the specific preferences of different audience groups, and provide personalized optimization suggestions and innovative ideas for anchors, so as to better meet diverse needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 The present invention is a flowchart of an anchor management method based on data analysis according to an embodiment of the present application.

[0011] Figure 2 This is a flowchart of step S130 in the anchor management method based on data analysis according to an embodiment of the present application.

[0012] Figure 3 This is a flowchart of step S140 in the anchor management method based on data analysis according to an embodiment of the present application.

[0013] Figure 4 It is a block diagram of an anchor management system based on data analysis according to an embodiment of the present application. DETAILED DESCRIPTION

[0014] The embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, but rather these embodiments are provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0015] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0016] It is worth noting that in this application, all actions to obtain signals, information or data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0017] In recent years, with the rapid development of Internet technology, the live broadcast industry has become an emerging and dynamic way of content dissemination. Live broadcasters interact with viewers in real time through live broadcast platforms and share various content, including but not limited to entertainment, education, shopping, sports events, etc. As the live broadcast market expands, how to effectively manage and optimize live broadcast content has become an urgent problem to be solved. Live broadcasters need to constantly adjust and improve their content strategies to maintain audience interest and improve the level of interaction.

[0018] In traditional live broadcast management, anchors mainly rely on personal experience to plan content and design interactions, which has great limitations. Specifically, personal experience is often limited to the anchor's own background and experience, and it is difficult to cover the diverse needs of the audience. In addition, the lack of systematic and data-supported empirical methods cannot accurately quantify audience preferences and behavior patterns, and it is difficult to deeply explore the potential value of interactive information, which limits content innovation and real-time optimization, thereby affecting the quality of live broadcast content and the long-term engagement of the audience.

[0019] Therefore, in response to the above technical problems, the technical concept of the present application is to extract user interaction information from the live broadcast platform to obtain a collection of user interaction information, use artificial intelligence-based data analysis and processing technology to semantically encode each user interaction information, and then perform medium-granular semantic enhancement on the semantic encoding features of each user interaction information, so as to intelligently generate anchor content optimization suggestions based on the graph walk semantic feature aggregation representation between the semantic encoding features of each medium-granular semantic enhancement user interaction information. By collecting and analyzing interactive information from different audience groups, the present application can transcend the limitations of the anchor's personal background and experience, and understand the audience's needs more broadly and deeply. Moreover, it can deeply mine interactive information, identify the specific preferences of different audience groups, and provide anchors with personalized optimization suggestions and innovative ideas, so as to better meet diverse needs.

[0020] Figure 1 FIG. 1 is a flow chart of a method for managing anchors based on data analysis according to an embodiment of the present application. Figure 1As shown, according to the anchor management method based on data analysis of the embodiment of the present application, the method includes: S110, extracting user interaction information from the live broadcast platform to obtain a set of user interaction information; S120, semantically encoding each user interaction information in the set of user interaction information to obtain a set of semantic encoding features of user interaction information; S130, performing local semantic enhancement of the text set on the set of semantic encoding features of user interaction information to obtain a set of medium-granularity semantically enhanced user interaction information semantic encoding features; S140, performing graph walking adaptive semantic feature aggregation on the set of medium-granularity semantically enhanced user interaction information semantic encoding features to obtain user interaction information semantic aggregation features; S150, obtaining anchor content optimization suggestions based on the semantic aggregation features of user interaction information.

[0021] In step S110, user interaction information is extracted from the live broadcast platform to obtain a collection of user interaction information. It should be understood that the collection of user interaction information refers to a collection of information collected from the live broadcast platform that covers the interactive behaviors between many users and anchors, such as comment content, barrage information, viewing time, etc. Therefore, the audience's interests and preferences can be better understood, thereby helping the anchor to adjust the content strategy to better meet the audience's needs. In the technical solution of the present application, user interaction information is extracted from the live broadcast platform to obtain a collection of user interaction information.

[0022] First of all, as a gathering place for information, live broadcast platforms contain a rich variety of user interaction data. In order to achieve effective extraction, it is necessary to connect with the application programming interface (API) of the live broadcast platform. Live broadcast platforms usually provide a series of interface specifications and permission settings, and developers need to perform legal access operations based on these requirements. During the access process, it is necessary to ensure the legality and security of data acquisition to prevent data leakage or illegal acquisition.

[0023] The extracted user interaction information covers multiple aspects. Among them, the comment content is a key part. When watching live broadcasts, viewers will express their thoughts, feelings, questions, and feedback on the host's content in text. These comments may involve evaluations of the products explained by the host. For example, in shopping live broadcasts, viewers will comment on the quality of the products, whether the price is reasonable, and how effective they are. In educational live broadcasts, they may give feedback on the difficulty and clarity of the explanation of knowledge points. In entertainment live broadcasts, it may be a liking or suggestion for the performance content. Barrage information is equally important. It has the characteristics of real-time and simplicity, and can quickly reflect the audience's immediate reaction during the viewing process. For example, in live broadcasts of sports events, viewers will express their excitement about a wonderful moment or their views on the situation of the game through barrages. Watching time is also an important indicator, which indirectly reflects the audience's interest in the live content. Longer viewing time usually means that the audience has a higher degree of attention and participation in the content, while shorter viewing time may imply that the content has failed to effectively attract the audience. This information is of great reference value for the host to understand the attractiveness of their own content.

[0024] During the extraction process, the timeliness of the data also needs to be considered. The information in the live broadcast industry is updated very quickly, and the interests and feedback of the audience are constantly changing. Therefore, it is necessary to obtain the latest interactive information in a timely manner so that the anchor can adjust the content strategy according to the real-time audience needs. At the same time, the analysis of historical data is also of certain value, which can help the anchor understand his own development trends and the evolution of audience preferences. In order to ensure the quality of the extracted user interaction information, it is also necessary to establish a data verification and error correction mechanism. Perform an integrity check on the extracted data to ensure that no important information fields are omitted. For some data that may contain errors, such as malformed comments, abnormal viewing time data, etc., it is necessary to identify and correct them, or mark them if they cannot be corrected, so that they can be specially handled in the subsequent analysis process.

[0025] In step S120, each user interaction information in the set of user interaction information is semantically encoded to obtain a set of semantic encoding features of the user interaction information. Specifically, in an embodiment of the present application, each user interaction information in the set of user interaction information is semantically encoded to obtain a set of semantic encoding features of the user interaction information, including: using a semantic encoder based on a bidirectional gated recurrent unit to semantically encode each user interaction information in the set of user interaction information to obtain a set of user interaction information semantic encoding vectors as the set of semantic encoding features of the user interaction information. Accordingly, considering that each user interaction information contains key feature information of user behavior, these key feature information represent the core meaning of user interaction. Based on this, in the technical solution of the present application, each user interaction information in the set of user interaction information is semantically encoded to better understand the specific content of each interactive information and the emotional color, intention, etc. behind it, to obtain a set of semantic encoding vectors of user interaction information, thereby more accurately capturing the audience's real feedback. In particular, in a specific embodiment of the present application, a semantic encoder based on a bidirectional gated recurrent unit is used to semantically encode each user interaction information in the set of user interaction information to integrate information in two directions, thereby obtaining a richer and more accurate semantic representation, that is, a set of semantic encoding vectors of user interaction information. In particular, the interactive information in a live broadcast room is usually represented as a time series data, that is, the interaction occurs in chronological order. The bidirectional gated recurrent unit, as a recurrent neural network (RNN), is particularly suitable for processing this type of data because it can remember past context and simultaneously consider future information through a bidirectional mechanism, thereby helping the anchor understand the audience's behavioral habits and preference changes.

[0026] In particular, the specific process of semantically encoding the user interaction information using a semantic encoder based on a bidirectional gated recurrent unit (BiGRU) is as follows: The input layer receives various text data from the user interaction information set. For example, for a comment message "This product is really powerful and easy to use", the input layer will digitally represent it in a certain way. Usually, word embedding technology, such as Word2Vec or GloVe, is used to convert each word into a vector of fixed dimension. Assuming that the word embedding dimension used is 100, then words such as "this", "product", and "function" will be represented as a 100-dimensional vector, and these vectors are input into BiGRU in order.

[0027] The hidden layer is the core part of BiGRU, which contains gated recurrent units for both forward and backward propagation. During the forward propagation process, the input vector of each time step enters the gated recurrent unit together with the hidden state of the previous time step. The gated recurrent unit is mainly composed of an update gate, a reset gate, and a memory unit. Taking the update gate as an example, it calculates a weight based on the current input and the previous hidden state to determine how much information of the previous hidden state to retain. For the above comment information, when processing the word "function", the update gate will comprehensively consider the hidden state information of "this product" and the hidden state information of the "product" after processing at the previous moment to determine how to update the current hidden state. The reset gate controls how to combine the new input information with the previous hidden state. The memory unit will store important information at the current moment based on the results of the update gate and the reset gate, and this information will be used in the subsequent propagation process.

[0028] The process of backpropagation is similar to forward propagation, but in the opposite direction. It starts from the end of the text and processes each word in turn. For the same review, backpropagation starts from "convenient" and processes forward step by step. Backpropagation can capture the impact of subsequent information in the text on the current word. For example, the subsequent information "convenient to use" may affect the semantic understanding of the word "function", and the backpropagation process can incorporate this impact into the encoding of "function".

[0029] At each time step, the hidden states of the forward and backward propagations are concatenated. Assume that the forward hidden state is , the backward hidden state is , the hidden state after splicing This splicing method enables BiGRU to use the context of the text at the same time, which is very helpful for understanding the semantic relationship in user interaction information. For example, in a comment "I didn't understand the knowledge point the host talked about at first, but it was explained clearly later", through two-way propagation, we can better understand the relationship between "I didn't understand it at first" and "It was explained clearly later", as well as their semantic impact on "this knowledge point".

[0030] The output layer will generate a semantic encoding vector of the user interaction information based on the final hidden state. This vector will be used as the semantic encoding feature of the user interaction information for subsequent processing steps.

[0031] In order to improve the performance of BiGRU, some technical means can also be used. For example, add a dropout layer to prevent overfitting. During the training process, the output of some neurons is randomly set to 0, so that the network will not overly rely on certain specific neurons during the training process, and enhance the generalization ability of the model. At the same time, the number of layers of the network can be increased to build a multi-layer BiGRU structure. Each layer can further extract more advanced semantic features, but it is necessary to avoid the problem of gradient vanishing or gradient exploding caused by too many layers. Appropriate initialization methods and gradient clipping techniques can be used to solve these problems.

[0032] In practical applications, BiGRU needs to be adjusted according to the characteristics of user interaction information and task requirements. If the user interaction information contains a large number of professional terms or domain-specific vocabulary, the word embedding layer may need to be optimized, such as using a domain-specific pre-trained word embedding model, or increasing the learning weight of these professional vocabulary during training. If the text length of the interactive information varies greatly, you can use a method to dynamically adjust the length of the input sequence, or appropriately truncate or segment the long text to ensure that BiGRU can effectively process text data of different lengths.

[0033] In step S130, the set of semantic coding features of user interaction information is subjected to local semantic enhancement of text collection to obtain a set of medium-granularity semantic enhancement semantic coding features of user interaction information. In particular, considering that the set of semantic coding features of user interaction information may contain a variety of semantic elements, there are both key information closely related to the optimization of anchor content and some relatively minor or redundant information, and there is semantic relevance between each semantic information. Therefore, in order to further refine and strengthen the semantic representation of each interactive information to capture deeper semantic information, such as implicit emotions, intentions or context dependencies, in the technical solution of the present application, the set of semantic coding features of user interaction information is subjected to local semantic enhancement of text collection to obtain a set of medium-granularity semantic enhancement semantic coding features of user interaction information, so that the local semantic structure can be better preserved while reducing noise interference. This is particularly useful for understanding the complex dialogue flow of the live broadcast room.

[0034] Figure 2 Flowchart of step S130 in the anchor management method based on data analysis according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 2As shown, the step S130, performs text set local semantic enhancement on the set of user interaction information semantic coding features to obtain a set of medium-granularity semantically enhanced user interaction information semantic coding features, including: S131, calculating the grammatical depth implicit representation value of each user interaction information semantic coding feature in the set of user interaction information semantic coding features to obtain the set distribution of user interaction information semantic grammatical depth implicit representation values; S132, based on the set distribution of user interaction information semantic grammatical depth implicit representation values, performs medium-granularity contextual semantic enhancement on each user interaction information semantic coding feature in the set of user interaction information semantic coding features to obtain the set of medium-granularity semantically enhanced user interaction information semantic coding features.

[0035] More specifically, in the embodiment of the present application, the step S131, calculating the grammatical depth implicit representation value of each user interaction information semantic encoding feature in the set of user interaction information semantic encoding features to obtain a set distribution of the user interaction information semantic grammatical depth implicit representation values, includes: Each user interaction information semantic coding vector in the set of user interaction information semantic coding vectors is mapped to the hyperbolic space to obtain a set of hyperbolic space user interaction information semantic coding vectors. The formula of this process can be expressed as: ; in, is a set of semantic coding vectors of the user interaction information, and are the first, second, and third vectors in the set of semantic encoding vectors of user interaction information. and user interaction information semantic encoding vector, and are the first weight matrix in hyperbolic space and the second weight matrix in hyperbolic space respectively, For Perform hyperbolic space mapping, for The semantic encoding vector of user interaction information in hyperbolic space after mapping; The syntax depth implicit representation value of each hyperbolic space user interaction information semantic encoding vector in the set of the hyperbolic space user interaction information semantic encoding vector is calculated to obtain the set distribution of the user interaction information semantic syntax depth implicit representation value. The formula of this process can be expressed as: ;in, for The hyperbolic space user interaction information semantic encoding vector after mapping, To calculate the square of the length of a vector, is the logarithmic function value with base 2, For calculation The syntax depth of implies the value, It is the first in the set distribution of the semantic grammar deep implicit representation value of user interaction information. The semantic grammar of user interaction information deeply implies the value.

[0036] It should be understood that each user interaction information semantic encoding vector is mapped to the hyperbolic space to obtain a set of hyperbolic space user interaction information semantic encoding vectors. That is, compared with the traditional Euclidean space, the hyperbolic space has unique advantages in representing data with hierarchical structures and complex semantic relationships. Mapping it to the hyperbolic space helps to better capture and represent these complex semantic hierarchical structures, and prepares for more accurate mining of semantic information in the future.

[0037] Correspondingly, the grammatical depth implicit representation value of each hyperbolic space user interaction information semantic encoding vector is calculated to further analyze the semantic information represented by the vector from a grammatical perspective, to mine the deep semantic features, and to obtain the set distribution of the semantic grammatical depth implicit representation values ​​of the user interaction information, thereby reflecting the semantic association strength and importance within the interactive information and between it and other information.

[0038] More specifically, in an embodiment of the present application, the step S132, based on the set distribution of the user interaction information semantic grammatical deep implicit representation values, performs medium-granularity contextual semantic enhancement on each user interaction information semantic coding feature in the set of user interaction information semantic coding features to obtain the set of medium-granularity semantically enhanced user interaction information semantic coding features, including: determining the user interaction information medium-granularity semantic association window of each user interaction information semantic coding vector in the set of user interaction information semantic coding vectors based on the set distribution of the user interaction information semantic grammatical deep implicit representation values; performing medium-granularity contextual semantic enhancement on each user interaction information semantic coding vector in the set of user interaction information semantic coding vectors based on the user interaction information medium-granularity semantic association window of each user interaction information semantic coding vector in the set of user interaction information semantic coding vectors to obtain a set of medium-granularity semantically enhanced user interaction information semantic coding vectors as the set of medium-granularity semantically enhanced user interaction information semantic coding features.

[0039] More specifically, in an embodiment of the present application, based on the set distribution of the semantic grammatical depth implicit representation values ​​of the user interaction information, determining the granular semantic association window in the user interaction information of each user interaction information semantic coding vector in the set of the user interaction information semantic coding vectors includes: extracting the semantic grammatical depth implicit representation value of the user interaction information of the first user interaction information semantic coding vector in the set of the user interaction information semantic coding vectors as the starting position of the granular semantic association window in the user interaction information; along the set distribution of the semantic grammatical depth implicit representation values ​​of the user interaction information, searching for the semantic grammatical depth implicit representation value of the user interaction information that is closest to the semantic grammatical depth implicit representation value of the first user interaction information semantic coding vector as the ending position of the granular semantic association window in the user interaction information; based on the starting position of the granular semantic association window in the user interaction information and the ending position of the granular semantic association window in the user interaction information, determining the granular semantic association window in the user interaction information of the first user interaction information semantic coding vector from the set of the user interaction information semantic coding vectors, and the formula of this process can be expressed as: ;in, and They are respectively the set distribution of the semantic grammar depth implicit representation value of user interaction information and The semantic grammar of user interaction information deeply implies the value, To find and The closest user interaction information semantic grammar depth implicit representation value corresponding position , for The size of the granular semantic association window in the user interaction information.

[0040] It should be understood that based on the set distribution of the implicit representation values ​​of the semantic grammar depth of user interaction information, the window size of each user interaction information semantic encoding vector is determined to define which interaction information should be regarded as closely related "neighbors", that is, the medium-granular semantic association window of the user interaction information. In this way, according to the characteristics of each interactive information itself, its appropriate medium-granular semantic association window can be adaptively determined to focus on those related information that are valuable for the current medium-granular semantic reinforcement, avoiding the association range being too large, resulting in too complicated information to be effectively processed, or the association range being too small, resulting in the omission of important semantic association content, so that subsequent semantic reinforcement operations can be carried out around the appropriate semantic association.

[0041] Specifically, in an embodiment of the present application, based on the medium-granularity semantic association window of the user interaction information of each user interaction information semantic encoding vector in the set of the user interaction information semantic encoding vector, medium-granularity contextual semantic enhancement is performed on each user interaction information semantic encoding vector in the set of the user interaction information semantic encoding vector to obtain a set of medium-granularity semantically enhanced user interaction information semantic encoding vectors as a set of medium-granularity semantically enhanced user interaction information semantic encoding features. The formula of this process can be expressed as: ;in, is the scoring reference vector, is the scoring reference weight matrix, is the rating bias vector, yes The first granular semantic association window in the user interaction information user interaction information semantic encoding vector, is matrix multiplication, represents the exponential function value with the natural constant e as the base, It is the first in the set of medium-grained semantically enhanced user interaction information semantic encoding vectors. The medium-granularity semantics enhances the semantic encoding vector of user interaction information.

[0042] Finally, based on the determined medium-granularity semantic association window of user interaction information, medium-granularity contextual semantic enhancement is performed on the semantic encoding vectors of each user interaction information. Through the contextual semantic enhancement operation, the useful semantics in other information closely related to the current user interaction information semantics can be integrated into the semantics represented by the current vector, highlighting the key semantic associations related to the live content optimization at the medium-granularity level, and obtaining a collection of medium-granularity semantically enhanced user interaction information semantic encoding vectors, which can better reflect the user's real demands and concerns for the live content at the medium-granularity level, and provide a high-quality data foundation for subsequent operations such as semantic feature aggregation analysis.

[0043] In step S140, the set of medium-granularity semantically enhanced user interaction information semantic coding features is subjected to graph walk adaptive semantic feature aggregation to obtain user interaction information semantic aggregation features. Furthermore, considering that each medium-granularity semantically enhanced user interaction information semantic coding feature has highlighted key semantics and enhanced relevance to a certain extent, it is still a plurality of relatively scattered vector representations, and there are implicit contextual deep semantic associations between each feature. Based on this, in the technical solution of the present application, the graph walk adaptive semantic features are aggregated through a graph walk adaptive semantic feature aggregation mechanism to obtain user interaction information semantic aggregation features.

[0044] Figure 3Flowchart of step S140 in the anchor management method based on data analysis according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 3 As shown, step S140, performs graph walk adaptive semantic feature aggregation on the set of medium-granularity semantically enhanced user interaction information semantic coding features to obtain user interaction information semantic aggregation features, including: S141, calculates the graph walk topological features of the set of medium-granularity semantically enhanced user interaction information semantic coding features; S142, performs autocorrelation significant aggregation on the set of medium-granularity semantically enhanced user interaction information semantic coding features based on the graph walk topological features to obtain the user interaction information semantic aggregation features.

[0045] Specifically, in step S141, the graph walk topology features of the set of semantic encoding features of the medium-granularity semantically enhanced user interaction information are calculated, including: Each medium-granularity semantically enhanced user interaction information semantic coding vector in the set of medium-granularity semantically enhanced user interaction information semantic coding vectors is subjected to hyperbolic space mapping to obtain a set of medium-granularity semantically enhanced user interaction information semantic coding vectors after hyperbolic space mapping. The formula of this process can be expressed as: ; in, is a set of semantic encoding vectors of the medium-granularity semantically enhanced user interaction information, and are the first, second, and third vectors in the set of medium-grained semantically enhanced user interaction information semantic encoding vectors. , and The medium-granularity semantics enhances the semantic encoding vector of user interaction information, and are respectively the first weight matrix in the hyperbolic space and the second weight matrix in the hyperbolic space, for A medium-granular semantically enhanced user interaction information semantic encoding vector after hyperbolic space mapping; The Poincare distance between any two semantic encoding vectors of medium-granularity semantically enhanced user interaction information after hyperbolic space mapping in the set of semantic encoding vectors of medium-granularity semantically enhanced user interaction information after hyperbolic space mapping is calculated to obtain the walk topology matrix of the medium-granularity semantically enhanced user interaction information graph. The formula of this process can be expressed as: ;in, for After the hyperbolic space mapping, the medium-granularity semantics enhances the semantic encoding vector of user interaction information. for After the hyperbolic space mapping, the medium-granularity semantics enhances the semantic encoding vector of user interaction information. is the square of the length of the vector, is the inverse hyperbolic cosine function, For calculation and The Poincare distance between Strengthening the user interaction information graph with medium-granular semantics in the topology matrix The eigenvalues ​​of the positions; The medium-granularity semantically enhanced user interaction information graph walk topology matrix is ​​subjected to dilated convolution coding to obtain a medium-granularity semantically enhanced user interaction information graph walk topology feature matrix as the graph walk topology feature. The formula of this process can be expressed as: ;in, is the medium-grained semantically enhanced user interaction information graph walk topology matrix, is the dilated convolutional coding, It is a medium-grained semantically enhanced user interaction information graph walk topology feature matrix.

[0046] It should be understood that each medium-granularity semantically enhanced user interaction information semantic encoding vector is respectively subjected to hyperbolic space mapping to better capture and retain the intrinsic semantic hierarchical relationship between these vectors, and the set of medium-granularity semantically enhanced user interaction information semantic encoding vectors after hyperbolic space mapping is obtained. Accordingly, the Poincare distance between any two medium-granularity semantically enhanced user interaction information semantic encoding vectors after hyperbolic space mapping is calculated to characterize the semantic distance between them, and the walk topology matrix of the medium-granularity semantically enhanced user interaction information graph is obtained to intuitively represent the connection relationship between each vector (the closeness of the connection is determined based on the distance).

[0047] It should be understandable that the medium-granularity semantically enhanced user interaction information graph walk topology matrix is ​​subjected to dilated convolutional encoding to capture the semantic association features at different scales in the graph walk topology matrix, thereby obtaining a medium-granularity semantically enhanced user interaction information graph walk topology feature matrix, which helps to discover deep and long-distance semantic connections and avoid focusing only on local neighbor relationships and missing important global semantic information.

[0048] Specifically, in step S142, based on the graph walk topology feature, the set of medium-granularity semantically enhanced user interaction information semantic encoding features is autocorrelatedly significantly aggregated to obtain the user interaction information semantic aggregation feature, including: The graph convolution global context walk is performed on the medium-granularity semantically enhanced user interaction information graph walk topological feature matrix and the set of medium-granularity semantically enhanced user interaction information semantic encoding vectors after hyperbolic space mapping to obtain a set of medium-granularity contextual semantically enhanced user interaction information feature vectors. The formula of this process can be expressed as: ;in, for After the hyperbolic space mapping, the medium-granularity semantics enhances the semantic encoding vector of user interaction information. It is the medium-grained semantically enhanced user interaction information graph walk topology feature matrix. is the graph convolutional coding, It is the first in the set of medium-grained contextual semantics-enhanced user interaction information feature vectors. The medium-granularity contextual semantics enhances the user interaction information feature vector; The set of medium-granularity contextual semantics-enhanced user interaction information feature vectors and the set of medium-granularity semantics-enhanced user interaction information semantic encoding vectors are subjected to autocorrelation gating processing for each corresponding set of medium-granularity contextual semantics-enhanced user interaction information feature vectors and medium-granularity semantics-enhanced user interaction information semantic encoding vectors to obtain a set of user interaction information autocorrelation gating significant confidence factors. The formula of this process can be expressed as: ;in, is the first in the set of medium-granularity semantically enhanced user interaction information semantic encoding vectors. The medium-granularity semantics enhances the semantic encoding vector of user interaction information, It is the first in the set of medium-grained contextual semantics-enhanced user interaction information feature vectors. The medium-granularity contextual semantics enhances the user interaction information feature vector, is the normalization function, To subtract by position point, It is calculated and The autocorrelation gated significance confidence factor between for and The user interaction information autocorrelation gated significant confidence factor between them; The set of user interaction information autocorrelation gated significant confidence factors is normalized based on the Softmax function to obtain a set of user interaction information autocorrelation gated significant confidence weight factors. The formula of this process can be expressed as: ;in, for and The user interaction information autocorrelation gated significant confidence factor between represents the exponential function value with the natural constant e as the base, is the number of user interaction information autocorrelation gated significant confidence factors in the set of user interaction information autocorrelation gated significant confidence factors, yes The corresponding user interaction information autocorrelation gated significant confidence weight factor; Based on the set of user interaction information autocorrelation gated significant confidence weight factors, the position-weighted sum of the set of medium-granularity semantically enhanced user interaction information semantic encoding vectors is calculated to obtain a user interaction information semantic aggregation feature vector as the user interaction information semantic aggregation feature. The formula of this process can be expressed as: ;in, is the first in the set of medium-granularity semantically enhanced user interaction information semantic encoding vectors. The medium-granularity semantics enhances the semantic encoding vector of user interaction information, yes The corresponding user interaction information autocorrelation gated significant confidence weight factor, is the number of vectors in the set of medium-granularity semantically enhanced user interaction information semantic encoding vectors, For point-by-position multiplication, is the semantic aggregation feature vector of the user interaction information.

[0049] Accordingly, the graph convolution global context walk is performed on the medium-granularity semantically enhanced user interaction information graph walk topological feature matrix and the set of medium-granularity semantically enhanced user interaction information semantic encoding vectors after hyperbolic space mapping to simulate the information flow in the entire semantic association graph. This can better capture the contextual semantic environment of each vector, further explore the semantic connections between different vectors based on the global context, and obtain a set of medium-granularity contextual semantically enhanced user interaction information feature vectors.

[0050] It should be understandable that each medium-granularity contextual semantically enhanced user interaction information feature vector and medium-granularity semantically enhanced user interaction information semantic encoding vector are subjected to autocorrelation gating processing to measure the correlation between each group of corresponding vectors, and the degree of correlation between them is used to determine which semantic information is more critical, more worthy of retention and highlighting, and higher attention is paid to the parts with high correlation, while those information with weak correlation or which may be noise are appropriately suppressed.

[0051] Accordingly, the user interaction information autocorrelation gated significant confidence factors obtained after each autocorrelation gating process are normalized based on the Softmax function and converted into relative weights to obtain a set of user interaction information autocorrelation gated significant confidence weight factors.

[0052] Finally, based on the set of weight factors, the position-weighted sum of the set of semantic encoding vectors of medium-granular semantically enhanced user interaction information is taken to integrate the semantic information contained in all vectors according to their respective importance weights, and a user interaction information semantic aggregation feature vector that can represent the overall semantic situation is obtained.

[0053] In step S150, based on the semantic aggregation features of the user interaction information, the anchor content optimization suggestion is obtained. Specifically, in the embodiment of the present application, based on the semantic aggregation features of the user interaction information, the anchor content optimization suggestion is obtained, including: after adding a prompt word at the end of the semantic aggregation feature vector of the user interaction information, it is input into the anchor content optimization suggestion generation module based on the large model to obtain the anchor content optimization suggestion. That is, after adding a prompt word at the end of the semantic aggregation feature vector of the user interaction information, it is input into the anchor content optimization suggestion generation module based on the large model to obtain the anchor content optimization suggestion. In particular, although the user interaction information semantic aggregation feature vector has integrated the semantics of many user interaction information, it is still a relatively general and objective semantic representation. Adding prompt words can provide clear direction guidance for the anchor content optimization suggestion generation module based on the large model, so that it can generate optimization suggestions according to specific needs and concerns. For example, if the prompt word "improve the professionalism of product explanation" is added, the model will focus on giving suggestions on how to introduce product knowledge in depth and use professional terms to accurately describe product characteristics, so as to meet the optimization needs of anchors in specific fields. In particular, in a specific embodiment of the present application, the large model may be GPT-4 or the like.

[0054] It should be understood that, here, in the case where each medium-granularity semantically enhanced user interaction information semantic coding vector in the set of medium-granularity semantically enhanced user interaction information semantic coding vectors represents the single-sample domain granularity enhanced coding semantic features of the corresponding user interaction information, when performing semantic feature aggregation analysis based on graph walk autocorrelation gating, the misalignment of the graph walk autocorrelation gating of the differentiated medium-granularity encoded semantic feature patterns will also cause the richness of the semantic aggregation distribution of the user interaction information semantic aggregation feature vector, resulting in repetition of unstructured probability mapping, affecting the accuracy of the generated results.

[0055] Therefore, preferably, after adding a prompt word to the tail of the user interaction information semantic aggregation feature vector, when it is input into the anchor content optimization suggestion generation module based on the large model to obtain the anchor content optimization suggestion, the user interaction information semantic aggregation feature vector is optimized. Specifically, the optimization includes the following steps: The length of the semantic aggregation feature vector of the user interaction information Subtract the zero norm of the user interaction information semantic aggregation feature vector to determine the number of zero eigenvalues ​​of the user interaction information semantic aggregation feature vector , and the number of zero eigenvalues Subtract one to get the zero differentiable value of the semantic aggregation of user interaction information ; The user interaction information semantically aggregates zero differentiable values Multiply and divide by the number of zero eigenvalues ​​respectively Get the first user interaction information semantic aggregation field fitting value Interaction information semantic aggregation field fitting value with the second user ; Calculate the square root of the sum of squares of all eigenvalues ​​of the user interaction information semantic aggregation feature vector to obtain the user interaction information semantic aggregation mode representation value ; Aggregate zero differentiable value with the user interaction information semantics Calculate each eigenvalue of the semantic aggregation feature vector of the user interaction information as an index The power function , and multiplied by the first user interaction information semantic aggregation field fitting value To obtain the semantic aggregation intermediate representation value of the first user interaction information ; Calculate each eigenvalue of the semantic aggregation feature vector of the user interaction information The semantic aggregation field fitting value of the interaction information with the second user and the user interaction information semantic aggregation mode representation value The product of gets the second user interaction information semantic aggregation intermediate representation value ; Aggregate the intermediate representation value of the first user interaction information semantically Interactive information semantic aggregation intermediate representation value with the second user The weighted difference Composition-optimized semantically aggregated feature vector of user interaction information.

[0056] That is, based on the vector field fitting law of high-dimensional manifolds, the isolated zeros in the semantic aggregation feature vector of the user interaction information are used as the differentiable dimension of the high-dimensional manifold, so as to fix the normal direction pointing to the distribution center near the local coordinates represented by the eigenvalues ​​of the semantic aggregation feature vector of the user interaction information, thereby achieving effective alignment of the characteristic distribution pattern of the semantic aggregation feature vector of the user interaction information to the probability density differentiable field to avoid invalid repetition in the probability mapping process caused by the unstructured characteristics of the interactively rich distribution of the semantic aggregation feature vector of the user interaction information, and improve the accuracy of the generation result obtained by the anchor content optimization suggestion generation module based on the large model.

[0057] In summary, the anchor management method based on data analysis based on the embodiment of the present application is explained, which extracts user interaction information from the live broadcast platform to obtain a collection of user interaction information, uses artificial intelligence-based data analysis and processing technology to semantically encode each user interaction information, and then performs medium-granular semantic enhancement on the semantic encoding features of each user interaction information, thereby intelligently generating anchor content optimization suggestions based on the graph walk semantic feature aggregation representation between the semantic encoding features of each medium-granular semantic enhancement user interaction information. By collecting and analyzing interactive information from different audience groups, the present application can transcend the limitations of the anchor's personal background and experience, and understand the audience's needs more broadly and deeply. In addition, it can deeply mine interactive information, identify the specific preferences of different audience groups, and provide personalized optimization suggestions and innovative ideas for anchors, so as to better meet diverse needs.

[0058] Figure 4 FIG. 1 is a block diagram of an anchor management system based on data analysis according to an embodiment of the present application. Figure 4As shown, according to the anchor management system 100 based on data analysis of the embodiment of the present application, it includes: a user interaction information extraction module 110, which is used to extract user interaction information from the live broadcast platform to obtain a set of user interaction information; a user interaction information semantic coding module 120, which is used to semantically encode each user interaction information in the set of user interaction information to obtain a set of user interaction information semantic coding features; a user interaction information enhancement module 130, which is used to perform text set local semantic enhancement on the set of user interaction information semantic coding features to obtain a set of medium-granularity semantically enhanced user interaction information semantic coding features, wherein the user interaction information enhancement module includes: a deep implicit representation value calculation unit, which is used to calculate the grammatical deep implicit representation value of each user interaction information semantic coding feature in the set of user interaction information semantic coding features to obtain a set distribution of user interaction information semantic grammar deep implicit representation values; a medium-granularity semantic enhancement unit, which is used to perform text set local semantic enhancement on the set of user interaction information semantic grammar deep implicit representation values ​​based on the user interaction information semantic grammar deep implicit representation values. A set distribution containing representation values, performing medium-granularity contextual semantic enhancement on each user interaction information semantic encoding feature in the set of user interaction information semantic encoding features to obtain the set of medium-granularity semantically enhanced user interaction information semantic encoding features; a user interaction information semantic aggregation module 140, used to perform graph walk adaptive semantic feature aggregation on the set of medium-granularity semantically enhanced user interaction information semantic encoding features to obtain user interaction information semantic aggregation features, wherein the user interaction information semantic aggregation module includes: a graph walk topology calculation unit, used to calculate the graph walk topology features of the set of medium-granularity semantically enhanced user interaction information semantic encoding features; an autocorrelation aggregation unit, used to perform autocorrelation significant aggregation on the set of medium-granularity semantically enhanced user interaction information semantic encoding features based on the graph walk topology features to obtain the user interaction information semantic aggregation features; an anchor content optimization suggestion generation module 150, used to obtain anchor content optimization suggestions based on the user interaction information semantic aggregation features.

[0059] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned anchor management system based on data analysis have been referred to above. Figures 1 to 3 The data analysis-based anchor management method has been introduced in detail, and therefore, its repeated description will be omitted.

[0060] As described above, the anchor management system 100 based on data analysis according to the embodiment of the present disclosure can be implemented in various wireless terminals, such as a server with an anchor management algorithm based on data analysis. In a possible implementation, the anchor management system 100 based on data analysis according to the embodiment of the present disclosure can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the anchor management system 100 based on data analysis can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the anchor management system 100 based on data analysis can also be one of the many hardware modules of the wireless terminal.

[0061] Alternatively, in another example, the data analysis-based anchor management system 100 and the wireless terminal may also be separate devices, and the data analysis-based anchor management system 100 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0062] The above descriptions of various implementations of the present disclosure are exemplary and not exhaustive. The disclosure is not limited to the disclosed implementations, and many modifications and variations are obvious to those of ordinary skill in the art without departing from the scope and spirit of the described implementations. The terms used herein are selected to best explain the principles of the implementations, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the various practical methods disclosed herein.

Claims

1. A method for anchor management based on data analysis, characterized in that: include: Extracting user interaction information from the live broadcast platform to obtain a collection of user interaction information; Performing semantic coding on each piece of user interaction information in the set of user interaction information to obtain a set of semantic coding features of the user interaction information; The set of user interaction information semantic coding features is subjected to text set local semantic enhancement to obtain a set of medium-granularity semantically enhanced user interaction information semantic coding features, including: calculating the grammatical depth implicit representation value of each user interaction information semantic coding feature in the set of user interaction information semantic coding features to obtain a set distribution of user interaction information semantic grammatical depth implicit representation values; based on the set distribution of user interaction information semantic grammatical depth implicit representation values, medium-granularity contextual semantic enhancement is performed on each user interaction information semantic coding feature in the set of user interaction information semantic coding features to obtain the set of medium-granularity semantically enhanced user interaction information semantic coding features; The set of medium-granularity semantically enhanced user interaction information semantic coding features is subjected to graph walk adaptive semantic feature aggregation to obtain user interaction information semantic aggregation features, including: calculating graph walk topological features of the set of medium-granularity semantically enhanced user interaction information semantic coding features; based on the graph walk topological features, the set of medium-granularity semantically enhanced user interaction information semantic coding features is subjected to autocorrelation significant aggregation to obtain the user interaction information semantic aggregation features; Based on the semantic aggregation features of the user interaction information, anchor content optimization suggestions are obtained.

2. The anchor management method based on data analysis according to claim 1 is characterized in that: Semantically encoding each piece of user interaction information in the set of user interaction information to obtain a set of semantic encoding features of the user interaction information, including: using a semantic encoder based on a bidirectional gated recurrent unit to semantically encode each piece of user interaction information in the set of user interaction information to obtain a set of user interaction information semantic encoding vectors as the set of semantic encoding features of the user interaction information.

3. The anchor management method based on data analysis according to claim 2 is characterized in that: Calculating the grammatical depth implicit representation value of each user interaction information semantic encoding feature in the set of user interaction information semantic encoding features to obtain a set distribution of the user interaction information semantic grammatical depth implicit representation values, including: Mapping each user interaction information semantic coding vector in the set of user interaction information semantic coding vectors to a hyperbolic space to obtain a set of hyperbolic space user interaction information semantic coding vectors; The grammatical depth implicit representation value of each hyperbolic space user interaction information semantic encoding vector in the set of the hyperbolic space user interaction information semantic encoding vector is calculated to obtain the set distribution of the user interaction information semantic grammatical depth implicit representation values.

4. The anchor management method based on data analysis according to claim 3 is characterized in that: Based on the set distribution of the user interaction information semantic grammar deep implicit representation value, each user interaction information semantic coding feature in the set of user interaction information semantic coding features is subjected to medium-granularity contextual semantic enhancement to obtain the set of medium-granularity semantically enhanced user interaction information semantic coding features, including: Determining a granular semantic association window in user interaction information of each user interaction information semantic encoding vector in the set of user interaction information semantic encoding vectors based on a set distribution of the user interaction information semantic grammar depth implicit representation values; Based on the medium-granularity semantic association window of the user interaction information of each user interaction information semantic coding vector in the set of user interaction information semantic coding vectors, medium-granularity contextual semantic enhancement is performed on each user interaction information semantic coding vector in the set of user interaction information semantic coding vectors to obtain a set of medium-granularity semantically enhanced user interaction information semantic coding vectors as the set of medium-granularity semantically enhanced user interaction information semantic coding features.

5. The anchor management method based on data analysis according to claim 4 is characterized in that: Determining the granular semantic association window in the user interaction information of each user interaction information semantic encoding vector in the set of user interaction information semantic encoding vectors based on the set distribution of the user interaction information semantic grammar depth implicit representation value includes: Extracting a user interaction information semantic grammar depth implicit representation value of a first user interaction information semantic encoding vector from the set of user interaction information semantic encoding vectors as a starting position of a granular semantic association window in the user interaction information; Along the set distribution of the semantic grammatical depth implicit representation values ​​of the user interaction information, searching for the semantic grammatical depth implicit representation value of the user interaction information that is closest to the semantic grammatical depth implicit representation value of the first user interaction information semantic encoding vector as the termination position of the granular semantic association window in the user interaction information; Based on the starting position of the granular semantic association window in the user interaction information and the ending position of the granular semantic association window in the user interaction information, the granular semantic association window in the user interaction information of the first user interaction information semantic encoding vector is determined from the set of user interaction information semantic encoding vectors.

6. The anchor management method based on data analysis according to claim 5 is characterized in that: Calculating the graph walk topology features of the set of semantic encoding features of the medium-granularity semantically enhanced user interaction information includes: Performing hyperbolic space mapping on each of the medium-granularity semantically enhanced user interaction information semantic coding vectors in the set of medium-granularity semantically enhanced user interaction information semantic coding vectors to obtain a set of medium-granularity semantically enhanced user interaction information semantic coding vectors after hyperbolic space mapping; Calculating the Poincare distance between any two semantic encoding vectors of medium-granularity semantically enhanced user interaction information after hyperbolic space mapping in the set of semantic encoding vectors of medium-granularity semantically enhanced user interaction information after hyperbolic space mapping to obtain a walk topology matrix of the medium-granularity semantically enhanced user interaction information graph; The medium-granularity semantically enhanced user interaction information graph walk topology matrix is ​​subjected to dilated convolutional coding to obtain a medium-granularity semantically enhanced user interaction information graph walk topology feature matrix as the graph walk topology feature.

7. The anchor management method based on data analysis according to claim 6 is characterized in that: Based on the graph walk topology feature, the set of medium-granularity semantically enhanced user interaction information semantic encoding features is autocorrelatedly significantly aggregated to obtain the user interaction information semantic aggregation feature, including: Performing graph convolution global context walk on the medium-granularity semantically enhanced user interaction information graph walk topological feature matrix and the set of medium-granularity semantically enhanced user interaction information semantic encoding vectors after hyperbolic space mapping to obtain a set of medium-granularity contextual semantically enhanced user interaction information feature vectors; Performing autocorrelation gating processing on each corresponding set of medium-granularity contextual semantics-enhanced user interaction information feature vectors and medium-granularity semantics-enhanced user interaction information semantic encoding vectors in the set of medium-granularity contextual semantics-enhanced user interaction information feature vectors and the set of medium-granularity semantics-enhanced user interaction information semantic encoding vectors to obtain a set of user interaction information autocorrelation gating significant confidence factors; Normalizing the set of user interaction information autocorrelation gated significant confidence factors based on a Softmax function to obtain a set of user interaction information autocorrelation gated significant confidence weight factors; Based on the set of user interaction information autocorrelation gated significant confidence weight factors, the position-weighted sum of the set of medium-granularity semantically enhanced user interaction information semantic encoding vectors is calculated to obtain a user interaction information semantic aggregation feature vector as the user interaction information semantic aggregation feature.

8. The anchor management method based on data analysis according to claim 7 is characterized in that: Based on the semantic aggregation features of the user interaction information, anchor content optimization suggestions are obtained, including: after adding a prompt word at the end of the semantic aggregation feature vector of the user interaction information, it is input into an anchor content optimization suggestion generation module based on a large model to obtain the anchor content optimization suggestion.

9. An anchor management system based on data analysis, characterized in that: include: A user interaction information extraction module, used to extract user interaction information from the live broadcast platform to obtain a collection of user interaction information; A user interaction information semantic coding module, used for semantically coding each user interaction information in the set of user interaction information to obtain a set of user interaction information semantic coding features; A user interaction information enhancement module, used for performing text set local semantic enhancement on the set of user interaction information semantic coding features to obtain a set of medium-granularity semantically enhanced user interaction information semantic coding features, wherein the user interaction information enhancement module comprises: a deep implicit representation value calculation unit, used for calculating the grammatical depth implicit representation value of each user interaction information semantic coding feature in the set of user interaction information semantic coding features to obtain a set distribution of user interaction information semantic grammatical depth implicit representation values; a medium-granularity semantic enhancement unit, used for performing medium-granularity contextual semantic enhancement on each user interaction information semantic coding feature in the set of user interaction information semantic coding features based on the set distribution of the user interaction information semantic grammatical depth implicit representation values ​​to obtain the set of medium-granularity semantically enhanced user interaction information semantic coding features; A user interaction information semantic aggregation module, used for performing graph walk adaptive semantic feature aggregation on the set of medium-granularity semantically enhanced user interaction information semantic coding features to obtain user interaction information semantic aggregation features, wherein the user interaction information semantic aggregation module includes: a graph walk topology calculation unit, used for calculating the graph walk topology features of the set of medium-granularity semantically enhanced user interaction information semantic coding features; an autocorrelation aggregation unit, used for performing autocorrelation significant aggregation on the set of medium-granularity semantically enhanced user interaction information semantic coding features based on the graph walk topology features to obtain the user interaction information semantic aggregation features; The anchor content optimization suggestion generation module is used to obtain anchor content optimization suggestions based on the semantic aggregation features of the user interaction information.

Images