Interaction data processing method and system based on AIGC interaction analysis

Abstract

The invention relates to the field of digital data processing, in particular to an interactive data processing method and system based on AIGC interactive analysis, and the method comprises the steps that keywords are extracted from historical interactive data, a global association word bank is constructed, the global association word bank is an undirected complete graph, each keyword is a node, and the simultaneous occurrence frequency of the two keywords is an edge weight; obtaining current round interaction data, determining a local keyword and a local association word library, and taking the local keyword as an association word corresponding to the current round interaction data; local keywords are recommended to the user, user feedback of interaction data of the current round is obtained, and interaction data processing of AIGC interaction analysis is completed. The method has the effect of improving the data processing efficiency.

Description

Technical Field

[0001] The present application relates to the field of digital data processing, and in particular to an interactive data processing method and system based on AIGC interactive analysis. Background Art

[0002] AIGC (Artificial Intelligence Generated Content) refers to the use of artificial intelligence to produce content, where AI is the abbreviation of artificial intelligence and GC refers to creative content. AIGC can include various forms of content, such as articles, news, music, paintings, videos, etc. Currently, AIGC is mainly used in text, images, videos, audio, games, and virtual humans.

[0003] During the interaction process, generating associative words is a common requirement, which can help users find the required information faster, improve input speed, increase user stickiness, assist language learning, etc. For example, in search engines, the associative word API can provide search suggestions related to user input. In intelligent input methods, it can predict the next word that users want to enter. In social media platforms, the associative word API can help users quickly complete the input of tags, topics or comments. In addition, the associative word API can also be applied to content recommendation systems to recommend relevant content based on user interests.

[0004] The prior art is to determine associative words by establishing an associative word library based on semantic similarity. However, if an associative word library is established for each user's AIGC interaction data each time, it will lead to greater redundancy in the AIGC platform and lower data processing efficiency. Summary of the invention

[0005] In order to solve the above-mentioned problems of high redundancy and low efficiency of the AIGC platform caused by too many associative thesaurus, the present application provides an interactive data processing method and system based on AIGC interactive analysis.

[0006] In the first aspect, the present application provides an interactive data processing method based on AIGC interactive analysis, which adopts the following technical solutions:

[0007] The interactive data processing method based on AIGC interactive analysis includes the following steps: extracting keywords from historical interactive data, constructing a global associative vocabulary, wherein the global associative vocabulary is an undirected complete graph, wherein each keyword is a node, and the frequency of two keywords appearing at the same time is an edge weight; obtaining the interactive data of the current round, determining local keywords and the local associative vocabulary, wherein the local keywords are used as associative words corresponding to the interactive data of the current round; recommending local keywords to users and obtaining user feedback on the interactive data of the current round, thereby completing the interactive data processing of AIGC interactive analysis; the method for obtaining local keywords is as follows: extracting keywords from the interactive data of the current round, and constructing a keyword sequence and a local associative vocabulary of the interactive data of the current round; extracting an extraction sequence containing all keywords in the local associative vocabulary from the global associative vocabulary; using a pre-trained word vector model to identify keywords in the local associative vocabulary; and The keywords are vectorized to obtain a first vectorized vocabulary, and the keywords in the extraction sequence are vectorized to obtain a second vectorized vocabulary; the keyword with the highest frequency in the second vectorized vocabulary is used as the anchor keyword, and a first ratio sequence is constructed according to the ratio of the frequency of the keyword in the first vector vocabulary to the frequency of the anchor keyword; a second ratio sequence is constructed according to the ratio of the frequency of the keyword in the second vector vocabulary to the frequency of the anchor keyword; a keyword matrix is ​​constructed, and the rows of the keyword matrix represent the number of keywords, and the columns represent the preset number of related words related to each keyword; the association coefficient of the keywords in the keyword matrix is ​​calculated, and the association coefficient is inversely proportional to the difference between the corresponding ratio of the keywords in the first ratio sequence and the corresponding ratio of the keywords in the second ratio sequence; by sorting the association coefficients of each non-repeated keyword in the keyword matrix from high to low, a preset number of keywords are selected as local keywords.

[0008] The beneficial effect is that the global associative word library is a large associative word database. If the database is directly used to generate associative words for the current keyword sequence, the amount of calculation is too large, so the application constructs a local associative word library and obtains a part of the data in the global associative word library to generate associative words. After obtaining the local associative word library, the local associative word library is found from the global associative word library as an approximate local keyword, and the local keyword is used as the associative word corresponding to the current round of interactive data.

[0009] By building a local associative word library, the large amount of calculations when directly using the global associative word library to generate associative words is avoided, thereby improving the efficiency of data processing. In addition, the extraction and recommendation of local keywords can more accurately reflect the characteristics of the current round of interaction data, allowing the AIGC tool to provide more personalized and context-related answers, enhancing the coherence and accuracy of human-computer interaction. Furthermore, through the construction of a local associative word library and the recommendation of local keywords, computing resources can be used more effectively, focusing resources on the data most relevant to the current interaction, thereby improving the response speed and processing power of the overall system.

[0010] Optionally, the calculation formula of the association coefficient is: Among them, F i,j,t represents the association coefficient of the tth keyword in the jth round of interaction of user i, exp represents the exponential function with e as the base, E i,j,t Indicates the difference corresponding to the tth keyword in the first ratio sequence, E' i,j,t represents the difference value corresponding to the tth keyword in the second ratio sequence; H t It represents the frequency of the t-th keyword in the keyword matrix, G represents the number of keywords in the keyword matrix, and P is a hyperparameter.

[0011] The beneficial effect is that the calculation of the association coefficient takes into account the relative frequency of the keyword in the current interaction round and the relative frequency in the global interaction. This dual consideration helps to optimize the keyword recommendation mechanism to make it more in line with the user's current needs and context. i,j,t -E' i,j,t The larger the value, the more likely the tth keyword is in E i,j The higher the frequency of occurrence in the conversation, the higher the frequency of occurrence of related keywords in other conversations. The larger the value, the higher the frequency of related keywords in other conversations. i,j,t -E' i,j,t ) is larger, The larger the value, the more likely the t-th keyword should be associated (recommended), and the higher the association coefficient of the t-th keyword. The higher the frequency of the keyword in the keyword matrix, the stronger the relevance. However, if E' i,j If the frequency of the keywords in the search results is low, even if the frequency of related keywords is higher, the recommendation degree is still not high.

[0012] Optionally, the calculation formula of the association coefficient is: F i,j,t =exp(E i,j,t -E' i,j,t ), where F i,j,t represents the association coefficient of the tth keyword in the jth round of interaction of user i, exp represents the exponential function with e as the base, E i,j,t Indicates the difference corresponding to the tth keyword in the first ratio sequence, E' i,j,t Represents the difference corresponding to the tth keyword in the second ratio sequence.

[0013] The beneficial effect is: since some calculation steps are omitted, such as the ratio calculation of keyword frequency and the adjustment of hyperparameters, this method can significantly reduce the amount of calculation, thereby speeding up the processing speed. The association coefficient calculated by this method has a lower accuracy, but it is suitable for application scenarios with lower requirements for recommendation accuracy and smaller requirements for calculation.

[0014] Optionally, local keywords are recommended to users and user feedback on the current round of interaction data is obtained to complete the interaction data processing of AIGC interaction analysis, including the steps of: obtaining user feedback, the user feedback is a determination instruction for an associative word, and the determined associative word is used as an updated associative word; an updated associative word sequence is constructed according to the updated associative words, and the keyword matrix is ​​updated according to the updated associative words to obtain an associative word matrix for the next associative word recommendation.

[0015] The beneficial effect is that by updating the associative words and keyword matrix, the system can maintain the continuity of the conversation, making multiple rounds of conversation more natural and smooth.

[0016] Optionally, an associative word matrix is ​​updated according to the updated associative words to obtain a keyword matrix for the next associative word recommendation, including the steps of: calculating the associative coefficient of each keyword in the associative word matrix and the associative coefficient of each keyword in the keyword matrix; calculating the cosine distance of the keywords between the associative word matrix and the keyword matrix; reconstructing the graph structure data of the updated associative words and the keywords in the keyword matrix to obtain an updated associative word library, wherein the nodes in the updated associative word library are keywords, and the edge weights are the cosine distances between the two keywords; performing data classification on the keywords in the updated associative word library using a clustering algorithm to obtain multiple cluster clusters, and determining the optimal keyword for each cluster cluster; recommending the optimal keyword to the user and obtaining user feedback on the corresponding round of interaction data.

[0017] The beneficial effect is that new keywords may be introduced during the keyword matrix update process. These new words may not be strongly related to the old keywords. Clustering can help identify these new words and reduce their interference with the existing recommendation system. In the recommendation system, only those keywords belonging to larger clusters need to be paid attention to, rather than the entire keyword matrix, which further improves the efficiency of data processing.

[0018] Optionally, the optimal keyword is determined by taking the keyword corresponding to the maximum association coefficient within the cluster as the optimal keyword of the cluster.

[0019] The beneficial effect is that each cluster contains semantically similar keywords, and the association coefficient is an indicator that measures the relevance of keywords to user interests. In each cluster, find the keyword with the largest association coefficient. This keyword is considered to be the most representative of the cluster, and these optimal keywords are recommended to users to obtain user feedback. User feedback will be used to further optimize the recommendation system, thereby improving the relevance and accuracy of recommendations.

[0020] Optionally, the clustering algorithm is a K-means clustering algorithm.

[0021] Optionally, the number of recommended keywords each time is the same or different.

[0022] In the second aspect, the present application provides an interactive data processing system based on AIGC interactive analysis, which adopts the following technical solutions:

[0023] The interactive data processing system based on AIGC interactive analysis includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the interactive data processing method based on AIGC interactive analysis is implemented.

[0024] The beneficial effect is: the above-mentioned interactive data processing method based on AIGC interactive analysis is generated into a computer program and stored in a memory so as to be loaded and executed by a processor, thereby making a system based on the memory and the processor for easy use. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present application will become easily understood. In the accompanying drawings, several embodiments of the present application are shown in an exemplary and non-restrictive manner, and the same or corresponding numbers represent the same or corresponding parts.

[0026] Figure 1 It is a method flow chart of the interactive data processing method based on AIGC interactive analysis in an embodiment of the present application.

[0027] Figure 2 It is a method flow chart of step S2 in the interactive data processing method based on AIGC interactive analysis in an embodiment of the present application.

[0028] Figure 3 It is a flow chart of a method for determining updated associative words in an interactive data processing method based on AIGC interactive analysis in an embodiment of the present application. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0030] It should be understood that when the terms "first", "second", etc. are used in the claims, specification and drawings of the present application, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the specification and claims of the present application indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their collections.

[0031] The present application embodiment discloses an interactive data processing method based on AIGC interactive analysis, referring to Figure 1 , including steps S1 to S3, which are specifically as follows:

[0032] S1: Extract keywords from historical interaction data and build a global associative vocabulary. The global associative vocabulary is an undirected complete graph. Each keyword is a node, and the association coefficient between two keywords is the edge weight.

[0033] The AIGC interactive platform collects historical interactive data to obtain interactive data of all users. The semantic model is used to extract keywords in each round of dialogue. The associative vocabulary can be constructed based on the frequency of occurrence of keywords. The associative vocabulary is a graph data, and the associative vocabulary is constructed in the form of a graph. The specific process is as follows:

[0034] The historical interaction data A generated by all interactions of all users is obtained through the AIGC interaction platform. The AIGC interaction data of the jth interaction of the i-th user is recorded as A i,j , the interaction data A generated by each interaction of 10,000 users 10000 Keywords are extracted using a pre-trained BERT model. 10,000 is a hyperparameter, which is the number of users used to build an associative vocabulary library. It represents the data used to build an associative vocabulary library. The pre-trained BERT model is used to extract keywords, which is a prior art and will not be described in detail in this application.

[0035] After extracting the keywords, the keywords in each interaction data are combined into a keyword sequence B. The keyword sequence corresponding to the AIGC interaction data of the i-th user's j-th interaction is B i,j .

[0036] The keyword sequences of each interaction data of all users can be combined into a total keyword sequence. The number of times any two keywords appear in the interaction data corresponding to each user is recorded as the association coefficient of the two keywords.

[0037] The graph structure data of the total keyword sequence is constructed, and the graph structure data corresponding to the total keyword sequence is denoted as C, where C is the constructed global associative vocabulary.

[0038] The global associative vocabulary is an undirected complete graph, in which each keyword is a node, and the edge weight between any two nodes is the association coefficient (cumulative number of occurrences) between the corresponding two keywords.

[0039] S2: Obtain the current round of interaction data and determine local keywords and local associative vocabulary.

[0040] After obtaining the global associative word library, if the current user is the i-th user and the latest conversation round is the j-th round, then for the keyword sequence B i,j Construct graph structure data and obtain local associative word library C i,j , the local associative vocabulary is an undirected complete graph.

[0041] The global associative word library is a large associative word database. If the database is directly used to generate associative words for the current keyword sequence, the amount of calculation is too large, so the application constructs a local associative word library and obtains a part of the data in the global associative word library to generate associative words. After obtaining the local associative word library, it is necessary to find the local associative word library as an approximate local keyword from the global associative word library, and the local keyword is used as the associative word corresponding to the current round of interaction data.

[0042] Reference Figure 2 The process of obtaining local keywords includes steps S20 to S22, which are as follows:

[0043] S20: extract keywords from the current round of interaction data, and construct a keyword sequence and a local associative vocabulary of the current round of interaction data; extract an extraction sequence containing all keywords in the local associative vocabulary from the global associative vocabulary.

[0044] Since the global associative vocabulary C contains many keywords, choose to obtain the local associative vocabulary C i,j The keyword with the highest frequency (most occurrences) is used as the anchor, the global associative vocabulary C and the local associative vocabulary C i,j They are all undirected complete graphs in the graph structure. The global associative vocabulary C queries the local associative vocabulary C i,j After finding the keyword with the highest frequency (most occurrences), the position of the keyword in the global associative vocabulary C can be quickly obtained. From the global associative vocabulary C, a local associative vocabulary C can be extracted. i,j Extract sequence C' of all keywords in i,j .

[0045] S21: Using the pre-trained word vector model, the keywords in the local associative vocabulary are vectorized to obtain a first vectorized vocabulary, and the keywords in the extraction sequence are vectorized to obtain a second vectorized vocabulary.

[0046] In one embodiment, the word vector model uses Word2Vec, which learns the vector representation of words by predicting context. Word2Vec is an existing technology and will not be described in detail here.

[0047] S22: Take the keyword with the highest frequency in the second vectorized vocabulary as the anchor keyword, and construct a first ratio sequence according to the ratio of the frequency of the keyword in the first vectorized vocabulary to the frequency of the anchor keyword; construct a second ratio sequence according to the ratio of the frequency of the keyword in the second vectorized vocabulary to the frequency of the anchor keyword.

[0048] Get the first vectorized vocabulary XC i,j With the second vectorized vocabulary XC' i,j The frequencies D corresponding to the anchored keywords i,j and D' i,j , get D i,j With XC i,j The ratio of the corresponding frequencies of all keywords except the non-anchor keywords in the ratio sequence E is obtained. i,j (including the ratio of the anchored keyword, which is 1), and the corresponding frequencies of all other keywords are the numerator, XC i,j The frequency of the anchor keyword in is the denominator, and D' is obtained in the same way. i,j With XC' i,j The ratio of the corresponding frequencies of all keywords except the non-anchor keywords in the middle is used to obtain the second ratio sequence E' i,j (The ratio corresponding to the anchored keyword is 1).

[0049] S23: construct a keyword matrix, wherein the rows of the keyword matrix represent the number of keywords, and the columns represent the number of preset related words related to each keyword.

[0050] The rows of the keyword matrix represent the number of keywords, and the columns represent the number of related words preset to each keyword. i,j Each keyword is selected from the top 10 most relevant keywords (but the keywords here can only be selected from the global associative vocabulary C). 10 is a hyperparameter that can be adjusted by the implementer according to the specific implementation scenario. The top 10 keywords are the top ten most relevant keywords (the preset number of related words related to each keyword is 10).

[0051] S24: Calculate the association coefficient of the keywords in the keyword matrix, where the association coefficient is inversely proportional to the difference between the corresponding ratio of the keywords in the first ratio sequence and the corresponding ratio of the keywords in the second ratio sequence; sort the association coefficients of each non-repeating keyword in the keyword matrix from high to low, and select a preset number of keywords as local keywords.

[0052] In one embodiment, the calculation formula of the association coefficient is:

[0053] Among them, F i,j,t represents the association coefficient of the tth keyword in the jth round of interaction of user i, E i,j,t Indicates the difference corresponding to the tth keyword in the first ratio sequence, E' i,j,t represents the difference value corresponding to the tth keyword in the second ratio sequence; H t It represents the frequency of the t-th keyword in the keyword matrix, and G represents the number of keywords in the keyword matrix.

[0054] exp represents an exponential function with e as the base. exp is used to prevent the occurrence of zero values. P is a hyperparameter, which can be 5 and can be adjusted by the implementer according to the specific implementation scenario.

[0055] E i,j,t -E' i,j,t The larger the value, the more likely the tth keyword is in E i,j The higher the frequency of occurrence in the conversation, the higher the frequency of occurrence of related keywords in other conversations. The larger the value, the higher the frequency of related keywords in other conversations. i,j,t -E' i,j,t ) is larger, The larger the value is, the more likely the t-th keyword should be associated (recommended), and the higher the association coefficient of the t-th keyword is. Conversely, the lower the association coefficient of the t-th keyword is. The higher the frequency of the keyword in the keyword matrix, the stronger the relevance is. However, if E' i,j If the frequency of the keywords in the search results is low, even if the frequency of related keywords is higher, the recommendation degree is still not high.

[0056] After anchoring, due to the inconsistency of the frequencies corresponding to different keywords, the recommendation is based on XC' i,j The frequency information is adjusted. When constructing the global associative vocabulary C, it relies on a large amount of data, while a single round relies on less data. The frequency needs to be normalized to obtain the recommendation degree (association coefficient) of the keyword.

[0057] By sorting the association coefficients of each non-repeated keyword in the keyword matrix from high to low, a preset number of keywords are selected as local keywords as the association words corresponding to the current round of interaction data. The preset number can be 5, which can be adjusted by the implementer according to the specific implementation scenario.

[0058] In other embodiments, only exp(E i,j,t -E 'i,j,t) as the association coefficient, but the association coefficient calculated by this method has low accuracy and is suitable for application scenarios with low requirements for recommendation accuracy and small requirements for calculation amount.

[0059] S3: Recommend local keywords to users and obtain user feedback on the current round of interaction data to complete the interaction data processing of AIGC interaction analysis.

[0060] Obtain user feedback, which is an instruction for determining the associative words, and use the determined associative words as updated associative words; construct an updated associative word sequence based on the updated associative words, and update the keyword matrix based on the updated associative words to obtain an associative word matrix for the next associative word recommendation.

[0061] After obtaining the associated words, the associated words are pushed to the user as recommended words, and the user determines the associated words to obtain a determined sequence of associated words. After obtaining the determined sequence of associated words, it means that the user has a clearer understanding of his / her interaction intention, and the associated word matrix can be updated according to the determined sequence of associated words. The associated word sequence needs to be used as the prompt word of the AGIC interaction model under AIGC interaction, that is, the determined keywords are added to the prompt of the AIGC interaction model.

[0062] After a certain interaction using the determined associative words, the keyword matrix is ​​updated to recommend the next associative words (ie, the present application preferably updates once after multiple conversations, but updating once per conversation is also acceptable).

[0063] In one embodiment, referring to Figure 3 , the process of determining the updated associative words is as follows:

[0064] S30: Calculate the association coefficient of each keyword in the association word matrix and the association coefficient of each keyword in the keyword matrix.

[0065] For the convenience of description, the associative word matrix is ​​recorded as W B , the keyword matrix is ​​recorded as W Z , update the keyword matrix to get the associative word matrix, so the keyword matrix is ​​the new round of matrix, the associative word matrix is ​​the previous matrix, calculate W B With W Z The keyword cosine distance and the corresponding recommendation degree between them are used to determine the new round of keywords. The keywords in the newly added keyword matrix are not strongly related to the keywords in the previous keyword matrix. Normally, there is not much interference, but there may be a situation where the user asks the question again, so the old keyword matrix is ​​retained and analyzed.

[0066] S31: Calculate the cosine distance of keywords between the association word matrix and the keyword matrix.

[0067] By calculating the cosine distance, we can quantify the similarity between two keywords. The cosine distance is determined by measuring the cosine value of the angle between two vectors. The closer the value is to 0, the more similar the two vectors (keywords) are.

[0068] S32: Reconstruct graph structure data of the updated associative words and the keywords in the keyword matrix to obtain an updated associative vocabulary, where the nodes in the updated associative vocabulary are keywords and the edge weights are the cosine distances between the two keywords.

[0069] S33: Using a clustering algorithm to classify the keywords in the updated associative vocabulary to obtain multiple clusters, and determining the optimal keyword for each cluster.

[0070] From W B and W Z Select keywords with similar meanings and high corresponding association coefficients as the new recommendation of a single keyword, but it may be due to W B The new keywords section and W Z There may be strong semantic similarity between the keywords in W. B New keywords and W Z The new graph structure data is constructed based on the keywords in the text to obtain the graph structure data, that is, the associative vocabulary is updated, the nodes in the updated associative vocabulary are keywords, the edge weights are the vectorized cosine distance values ​​between two keywords, the associative vocabulary is updated to be an undirected complete graph, and the updated associative vocabulary is classified by spectral clustering, and the semantically similar ones are divided into one category. During spectral clustering, the number of clusters is set to 10, and the K-means clustering algorithm is selected as the clustering method. The number of clusters can be adjusted by the implementer according to the specific implementation scenario.

[0071] The optimal keyword is determined in the following way: the keyword corresponding to the maximum association coefficient within the cluster is used as the optimal keyword of the cluster.

[0072] Each cluster contains semantically similar keywords. The association coefficient is a measure of the relevance of keywords to user interests. In each cluster, find the keyword with the largest association coefficient. This keyword is considered to be the most representative of the cluster. These optimal keywords are recommended to users to obtain user feedback. User feedback will be used to further optimize the recommendation system, thereby improving the relevance and accuracy of recommendations.

[0073] S34: Recommend optimal keywords to users and obtain user feedback on corresponding rounds of interaction data.

[0074] Select the first 10 keywords as associated words for recommendation, determine the next keyword, and complete the AIGC interactive data processing. 10 is a hyperparameter, and it can also be 5. The number of recommended keywords each time can be inconsistent.

[0075] The embodiment of the present application also discloses an interactive data processing system based on AIGC interactive analysis, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an interactive data processing method based on AIGC interactive analysis according to the present application is implemented.

[0076] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, and their configuration and functions are known in the art, so they will not be described in detail here.

[0077] In the present application, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high bandwidth memory HBM (High Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module or both. Any such computer storage medium may be part of a device or accessible or connectable to a device.

[0078] Although this specification has shown and described a plurality of embodiments of the present application, it is obvious to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will think of many changes, modifications and alternatives without departing from the thought and spirit of the present application. It should be understood that in the process of practicing the present application, various alternatives to the embodiments of the present application described herein may be adopted.

[0079] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. The interactive data processing method based on AIGC interactive analysis is characterized by: Includes steps: Extract keywords from historical interaction data and build a global associative vocabulary. The global associative vocabulary is an undirected complete graph, where each keyword is a node and the frequency of two keywords appearing at the same time is the edge weight. Obtain the current round of interaction data, determine the local keywords and the local associative word library, and use the local keywords as the associative words corresponding to the current round of interaction data; Recommend local keywords to users and obtain user feedback on the current round of interaction data to complete the interaction data processing of AIGC interaction analysis; The method for obtaining local keywords is: Extract keywords from the current round of interaction data, and construct a keyword sequence and a local associative vocabulary for the current round of interaction data; extract an extraction sequence containing all keywords in the local associative vocabulary from the global associative vocabulary; Using the pre-trained word vector model, the keywords in the local associative word library are vectorized to obtain a first vectorized word library, and the keywords in the extraction sequence are vectorized to obtain a second vectorized word library; The keyword with the highest frequency in the second vectorized vocabulary is used as the anchor keyword, and a first ratio sequence is constructed according to the frequency ratio of the keyword in the first vectorized vocabulary to the anchor keyword; a second ratio sequence is constructed according to the frequency ratio of the keyword in the second vectorized vocabulary to the anchor keyword; Construct a keyword matrix, wherein the rows of the keyword matrix represent the number of keywords, and the columns represent the number of preset related words related to each keyword; Calculate the association coefficient of the keywords in the keyword matrix. The association coefficient is inversely proportional to the difference between the corresponding ratio of the keyword in the first ratio sequence and the corresponding ratio of the keyword in the second ratio sequence. Sort the association coefficients of each non-repeated keyword in the keyword matrix from high to low, and select a preset number of keywords as local keywords.

2. The interactive data processing method based on AIGC interactive analysis according to claim 1 is characterized in that: The calculation formula of the association coefficient is: Among them, F i,j,t represents the association coefficient of the tth keyword in the jth round of interaction of user i, exp represents the exponential function with e as the base, E i,j,t Indicates the difference corresponding to the tth keyword in the first ratio sequence, E' i,j,t represents the difference value corresponding to the tth keyword in the second ratio sequence; H t It represents the frequency of the t-th keyword in the keyword matrix, G represents the number of keywords in the keyword matrix, and P is a hyperparameter.

3. The interactive data processing method based on AIGC interactive analysis according to claim 1 is characterized in that: The calculation formula of the association coefficient is: F i,j,t =exp(E i,j,t -E' i,j,t ), where F i,j,t represents the association coefficient of the tth keyword in the jth round of interaction of user i, exp represents the exponential function with e as the base, E i,j,t Indicates the difference corresponding to the tth keyword in the first ratio sequence, E' i,j,t Represents the difference corresponding to the tth keyword in the second ratio sequence.

4. The interactive data processing method based on AIGC interactive analysis according to any one of claims 1 to 3, characterized in that: Recommend local keywords to users and obtain user feedback on the current round of interaction data to complete the interaction data processing of AIGC interaction analysis, including the following steps: Obtain user feedback, the user feedback is an instruction to determine the associated words, and use the determined associated words as updated associated words; An updated associative word sequence is constructed according to the updated associative words, and a keyword matrix is ​​updated according to the updated associative words to obtain an associative word matrix for the next associative word recommendation.

5. The interactive data processing method based on AIGC interactive analysis according to claim 4 is characterized in that: The keyword matrix is ​​updated according to the updated associative words to obtain an associative word matrix for the next associative word recommendation, including the following steps: Calculate the association coefficient of each keyword in the association word matrix and the association coefficient of each keyword in the keyword matrix; Calculate the cosine distance of keywords between the association word matrix and the keyword matrix; Reconstruct the graph structure data of the updated associative words and the keywords in the keyword matrix to obtain an updated associative vocabulary, where the nodes in the updated associative vocabulary are keywords and the edge weight is the cosine distance between the two keywords; Using a clustering algorithm to classify the keywords in the updated associative word library to obtain multiple clusters, and determining the optimal keywords for each cluster; Recommend the best keywords to users and obtain user feedback on the corresponding round of interaction data.

6. The interactive data processing method based on AIGC interactive analysis according to claim 5 is characterized in that: The optimal keyword is determined in the following way: the keyword corresponding to the maximum association coefficient within the cluster is used as the optimal keyword of the cluster.

7. The interactive data processing method based on AIGC interactive analysis according to claim 6 is characterized in that: The clustering algorithm is the K-means clustering algorithm.

8. The interactive data processing system based on AIGC interactive analysis is characterized by: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the interactive data processing method based on AIGC interactive analysis according to any one of claims 1 to 7 is implemented.

Images