Text processing method and device based on AI, computer equipment and storage medium

Abstract

The invention provides an AI-based text processing method and device, computer equipment and a storage medium. The method comprises the steps of obtaining a comment data set of a target object in a first time period, then determining a target comment data set from the comment data set according to a target comment label, then generating prompt information according to a comment text of each piece of comment data in the target comment data set, and inputting the prompt information into a preset AIGC engine. The prompt information comprises a prompt statement for indicating the AIGC engine to generate the reference comment text and a to-be-processed text, the text is composed of keywords and keyword phrases and is directly associated to a core concern of a user, and the intelligent prompt information generation mode has the advantages that the user experience is improved, and the user experience is improved. And the AIGC engine is helped to more accurately understand the intention of the user and generate a high-quality reference comment text.

Description

Technical Field

[0001] This application relates to data processing technologies, and in particular, to an AI-based text processing method, apparatus, computer device, and storage medium. Background Art

[0002] With the rapid development of Internet technologies and the wide popularization of social media, user-generated content has become an indispensable part of the cyberspace. Among them, comment data, as an important form of user feedback and opinion expression, is of great significance for understanding user needs, evaluating product or service quality, monitoring brand reputation, etc.

[0003] However, in the face of a vast amount of comment data, how to efficiently and accurately extract valuable information from it has become an urgent problem to be solved. Traditional text processing methods often require manual intervention for cumbersome data screening, classification, and analysis work, which is not only time-consuming and laborious but also difficult to ensure the objectivity and accuracy of the processing results. Summary of the Invention

[0004] This application provides an AI-based text processing method, apparatus, computer device, and storage medium, which are used to generate prompt information that can be input into an AIGC engine based on a vast amount of comment data, so as to generate high-quality reference comment texts, thereby improving the processing efficiency of the vast amount of comment data.

[0005] In a first aspect, this application provides an AI-based text processing method, which is characterized by including: Obtain a comment data set of a target object within a first time period, where the comment data set includes comment data made by different accounts on the target object within the first time period, and the comment data includes comment tags and comment texts; Determine a target comment data set from the comment data set according to a target comment tag, where the comment tags of the comment data in the target comment data set are the target comment tags; Generate prompt information according to the comment texts of each comment data in the target comment data set, where the prompt information includes a prompt statement and a text to be processed; Input the prompt information into a preset AIGC engine to output a reference comment text corresponding to the target comment tag.

[0006] In the above solution, by obtaining the comment dataset of the target object in the first time period, the comment data of different accounts in the same time period is effectively integrated, improving the data processing efficiency, enabling the rapid collection of a large number of relevant comments, and providing a data basis for subsequent analysis. At the same time, by performing tagging processing on the comment data, the data classification is further refined, facilitating subsequent targeted processing. Screening the target comment dataset from the comment dataset according to the target comment tag realizes the accurate positioning of the comment data. The target comment tag can be associated with multiple indicators such as the target satisfaction level, the target user attribute, and the single-customer comment count, thereby ensuring that the selected comment data highly meets specific analysis requirements. Then, by using the preset keyword database to extract keywords from the comment texts in the target comment dataset and generate prompt information, where the prompt information not only includes the prompt statement indicating the AIGC engine to generate reference comment texts, but also includes the text to be processed, which consists of keywords and keyword phrases and is directly related to the user's core concerns. This intelligent prompt information generation method helps the AIGC engine more accurately understand the user's intention and generate high-quality reference comment texts. Inputting the prompt information into the preset AIGC engine to output the reference comment texts corresponding to the target comment tag, thereby enabling the generation of high-quality, coherent, and logical text content. These reference comment texts not only reflect the user's true feedback, but also have high readability and reference value, which is of great significance for enhancing the market competitiveness of the target object.

[0007] Optionally, the target object is an e-commerce product; The target comment tag is used to associate one or more of the target satisfaction level, the target user attribute, and the single-customer comment count, where the single-customer comment count is the number of comments output by the same account before the end node of the first time period.

[0008] In the above solution, by setting the target object as an e-commerce product, it is possible to accurately collect and process the review data related to the product. These review data directly reflect the true feedback of consumers on the product, including multiple aspects such as satisfaction, usage experience, and function evaluation. By associating with the target review tags, such as the target satisfaction level, it is possible to quickly locate the overall evaluation tendency of consumers on the product, providing intuitive market feedback for merchants. Moreover, the target review tags are also associated with the target user attributes and the number of reviews per single customer, which can be used to analyze the characteristics of the user group and their review behaviors. Through user attributes (such as age, gender, region, etc.), it is possible to understand the preferences and acceptance degrees of different user groups for the product; while the number of reviews per single customer reflects the continuous attention and participation of users in the product. The comprehensive analysis of this information helps merchants to more comprehensively understand consumer needs. In addition, through intelligent text processing and analysis means, the efficiency of processing a large amount of e-commerce product review data has been significantly improved. Compared with the traditional manual analysis method, AIGC technology can generate high-quality reference review texts more quickly, providing timely and accurate market feedback for merchants.

[0009] Optionally, generate prompt information according to the review texts of each review data in the target review dataset. The prompt information includes a prompt statement and a text to be processed, including: Utilize a preset keyword database and perform keyword extraction on each review text in the target review dataset to generate a target review keyword vector corresponding to each target review data , where is the th target review data in the target review dataset , is the target review keyword vector corresponding to ; Generate a target review keyword matrix according to the target review keyword vectors corresponding to each target review data ; ; Generate the prompt information according to the target review keyword matrix , where the text to be processed includes the keywords in the target review keyword matrix , and the prompt statement is used to instruct the preset AIGC engine to generate the reference review text according to the text to be processed.

[0010] In the above solution, by using a preset keyword database to extract keywords from each review text in the target review dataset, the core information and key viewpoints in the reviews can be accurately identified, which not only improves the efficiency of text processing but also ensures the accuracy and pertinence of subsequent analysis. By generating the target review keyword vectors corresponding to each target review data, the complex review texts are transformed into a structured data form, facilitating subsequent further processing and analysis. After extracting the keywords, a target review keyword matrix is further generated according to the target review keyword vectors corresponding to each target review data, realizing the integration and correlation of keyword data, enabling the keywords originally scattered in multiple review texts to be centrally displayed, and facilitating subsequent analysis and mining. In addition, the construction of the keyword matrix provides an intuitive tool for understanding the overall distribution and correlation relationship of review data. Based on the target review keyword matrix, prompt information can be intelligently generated, including prompt statements and text to be processed. The prompt statements are used to instruct a preset AIGC engine to generate reference review texts according to the text to be processed, and the text to be processed contains the keywords in the keyword matrix. This intelligent way of generating prompt information ensures that the AIGC engine can receive accurate, relevant, and efficient input information, thereby generating high-quality reference review texts. By optimizing the input of the AIGC engine, the efficiency and accuracy of text generation are further improved. Finally, through the prompt information generated by using the above method, the AIGC engine can output reference review texts corresponding to the target review tags.

[0011] Optionally, the generating the target review keyword matrix according to the target review keyword vectors corresponding to each target review data includes: Determining the number of rows of the target review keyword matrix according to the keyword vector with the highest dimension in the target review keyword vectors corresponding to each target review data to generate the target review keyword matrix :

[0012] Wherein, is the j-th keyword in the i-th target review keyword vector in the target review keyword matrix ; if the number of keywords in the target review keyword vector is less than m, then the elements after are configured as preset values; Correspondingly, according to the target comment keyword matrix to generate the prompt information, including: traverse each element in the target comment keyword matrix and add the target comment keywords whose occurrence times are greater than the preset times to the text to be processed, so as to generate the prompt information, where the prompt statement is used to instruct the preset AIGC engine to generate the reference comment text according to all the target comment keywords in the text to be processed.

[0013] In the above solution, by determining the number of rows of the keyword matrix according to the keyword vector with the highest dimension in the target comment keyword vectors corresponding to each target comment data, a structured keyword matrix is generated, which not only realizes the effective integration of keyword data, but also optimizes the data storage and retrieval efficiency. The construction of the keyword matrix makes the association relationship between keywords clearer, facilitating subsequent analysis and mining work. At the same time, by filling the vectors with insufficient keyword quantities with preset values, the integrity and consistency of the matrix are ensured, providing convenience for subsequent processing. After generating the keyword matrix, by traversing each element in the matrix, the target comment keywords whose occurrence times are greater than the preset times are added to the text to be processed to construct the prompt information, realizing the accurate screening of high-frequency keywords and ensuring that the prompt information can accurately reflect the core concerns of consumers. High-frequency keywords usually represent the mainstream views or common problems in the comments. Incorporating them into the text to be processed helps the AIGC engine generate more targeted and reference-worthy reference comment texts. By using the prompt information constructed by the above method, the AIGC engine can receive more accurate and relevant input data, thereby generating reference comment texts that are more matched with the target comment tags. This improvement in relevance and accuracy not only enhances the readability and reference value of the reference comment texts, but also improves the overall efficiency of text generation. For e-commerce products, this means that merchants can obtain the real feedback of consumers more quickly, and then make more accurate marketing strategy adjustments and product improvement decisions. In addition, the entire process of generating the keyword matrix and constructing the prompt information fully demonstrates the intelligent level of AIGC technology in the field of data processing and analysis. Through the automated keyword extraction, matrix construction and screening process, the workload of manual processing is greatly reduced, and the efficiency of data processing and analysis is improved.

[0014] Optionally, according to the target comment keyword matrix to generate the prompt information, further includes: traverse the target comment keyword matrix each element in, and add the associated target comment keyword phrase whose keyword association feature value is greater than the preset association feature threshold to the text to be processed, where the associated target comment keyword phrase includes a first target comment keyword and a second target comment keyword.

[0015] In the above solution, by calculating the association feature values between keywords and screening out keyword phrases with high association degrees, the semantic relationships in the comment data can be understood more deeply. Associated keyword phrases often can reveal the comprehensive evaluations of reviewers on various aspects such as product features and usage experiences. Incorporating them into the text to be processed helps the AIGC engine generate richer, more accurate, and insightful reference review texts.

[0016] Optionally, traversing each element in the target comment keyword matrix and adding the associated target comment keyword phrase whose keyword association feature value is greater than the preset association feature threshold to the text to be processed further includes: Updating the keyword association feature value according to the preset weight factor of each keyword in the preset keyword database.

[0017] In the above solution, by introducing the weight factors in the preset keyword database, the association strength between keywords can be evaluated more accurately. The weight factors are usually determined based on the importance of keywords in a specific field or context. Therefore, using these factors to update the keyword association feature value can make the association evaluation more in line with the actual context and domain knowledge. This improvement in accuracy helps generate prompt information closer to the real comment intention, thereby improving the quality of the reference review texts generated by the AIGC engine.

[0018] Optionally, the AI-based text processing method further includes: Obtaining a monitoring instruction for the target object, where the monitoring instruction is used to indicate obtaining another reference review text corresponding to another target object within the first time period, where the another target object is a benchmark object of the target object; Generating and outputting a monitoring report at preset time intervals, where the monitoring report includes the reference review text and the another reference review text.

[0019] In the above solution, by obtaining the monitoring instructions of the target object and its benchmark object, it is possible to collect the reference review texts of both in a specific period in real time or regularly. These review texts not only reflect the direct feedback of consumers on the target object and its benchmark product, but also imply key information such as market trends and consumer preferences. Generating and outputting a monitoring report at a preset time interval enables the enterprise to quickly obtain the latest market dynamics and consumer feedback. This real-time monitoring and reporting mechanism helps the enterprise quickly respond to market changes, adjust product strategies, marketing strategies or customer service strategies to adapt to the changing market demands.

[0020] In a second aspect, the present application provides an AI-based text processing device, including: An acquisition module, configured to acquire a comment data set of a target object within a first period, where the comment data set includes comment data made by different accounts on the target object within the first period, and the comment data includes a comment tag and a review text; A processing module, configured to determine a target comment data set from the comment data set according to a target comment tag, where the comment tag of the comment data in the target comment data set is the target comment tag; A generation module, configured to generate a prompt message according to the review texts of the respective comment data in the target comment data set, where the prompt message includes a prompt statement and a text to be processed; Input the prompt message into a preset AIGC engine to output a reference review text corresponding to the target comment tag.

[0021] Optionally, the target object is an e-commerce product; The target comment tag is used to associate one or more of a target satisfaction level, target user attributes, and the number of single-customer comments, where the number of single-customer comments is the number of comments output by the same account before the end node of the first period.

[0022] Optionally, the generation module is specifically configured to: Utilize a preset keyword database and perform keyword extraction on the review texts in the target comment data set to generate a target comment keyword vector corresponding to each target comment data , where is the th target comment data in the target comment data set , and is the target comment keyword vector corresponding to ; Generate a target comment keyword matrix according to the target comment keyword vectors corresponding to the respective target comment data ​ ; Generate the prompt information according to the target comment keyword matrix where the text to be processed includes the keywords in the target comment keyword matrix and the prompt statement is used to instruct the preset AIGC engine to generate the reference comment text according to the text to be processed

[0023] Optionally, the generating module is specifically configured to: Determine the number of rows of the target comment keyword matrix according to the keyword vector with the highest dimension in each target comment keyword vector of the target comment keyword matrix to generate the target comment keyword matrix :

[0024] where is the th target comment keyword vector in the target comment keyword matrix and the th keyword in it. If the number of keywords in the target comment keyword vector is less than , then configure the elements after as preset values; Traverse each element in the target comment keyword matrix and add the target comment keywords with the number of occurrences greater than the preset number of times to the text to be processed to generate the prompt information, where the prompt statement is used to instruct the preset AIGC engine to generate the reference comment text according to all the target comment keywords in the text to be processed

[0025] Optionally, the generating module is further specifically configured to: Traverse each element in the target comment keyword matrix and add the associated target comment keyword phrase with the keyword association eigenvalue greater than the preset association feature threshold to the text to be processed, where the associated target comment keyword phrase includes the first target comment keyword and the second target comment keyword

[0026] Optionally, the generating module is further specifically configured to: Update the keyword association eigenvalue according to the preset weight factor of each keyword in the preset keyword database

[0027] ​Optionally, the obtaining module is further configured to obtain a monitoring instruction for the target object, where the monitoring instruction is used to indicate obtaining another reference review text corresponding to another target object within the first time period, and the another target object is a benchmark object of the target object; The output module 340 is further configured to generate and output a monitoring report at a preset time interval, where the monitoring report includes the reference review text and the another reference review text.

[0028] In a third aspect, the present application provides an electronic device, including: A processor; and, A memory for storing executable instructions of the processor; Wherein, the processor is configured to execute any possible method described in the first aspect by executing the executable instructions.

[0029] In a fourth aspect, the present application provides a computer-readable storage medium, where computer-executable instructions are stored in the computer-readable storage medium, and when the computer-executable instructions are executed by a processor, they are used to implement any possible method described in the first aspect.

[0030] The AI-based text processing method, device, computer device, and storage medium provided by the present application obtain a comment dataset of a target object within a first time period, then determine a target comment dataset from the comment dataset according to a target comment label, and then generate a prompt message based on the review texts of the comment data in the target comment dataset, and input the prompt message into a preset AIGC engine to output a reference review text corresponding to the target comment label. The prompt message not only includes a prompt statement indicating the AIGC engine to generate a reference review text, but also includes text to be processed, which consists of keywords and keyword phrases and is directly related to the core concerns of the user. This intelligent way of generating prompt messages helps the AIGC engine better understand the user's intention and generate high-quality reference review texts. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings here are incorporated into the specification and constitute a part of this specification, showing embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application.

[0032] Figure 1 is a schematic flowchart of an AI-based text processing method shown according to an example embodiment of the present application; Figure 2 is a schematic flowchart of an AI-based text processing method shown according to another example embodiment of the present application; Figure 3It is a schematic structural diagram of an AI-based text processing device shown according to an exemplary embodiment of the present application; Figure 4 It is a schematic structural diagram of an electronic device shown according to an exemplary embodiment of the present application.

[0033] Through the above-mentioned drawings, specific embodiments of the present application have been shown, and there will be a more detailed description hereinafter. These drawings and textual descriptions are not intended to limit the scope of the concept of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. Detailed Embodiments

[0034] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0035] To solve the above problems, in the embodiments provided by the present application, the system receives a monitoring instruction input by the user, and clarifies the target object, the benchmark object, and the monitoring time period (the first period) to be monitored. According to the monitoring instruction, the system captures the review texts of the target object and the benchmark object in real time or regularly within the specified time period. This process involves technical means such as web crawlers and API calls to ensure the comprehensiveness and accuracy of the review texts. The system sorts out and analyzes the collected review texts to generate a detailed monitoring report. The report content includes the reference review texts of the target object and the benchmark object, sentiment tendency analysis, keyword frequency statistics, etc., as well as the comparative analysis between the two. Finally, the system outputs the generated monitoring report in the manner specified by the user, such as email, the internal message center of the system, file download, etc., to ensure that the user can conveniently obtain the report content.

[0036] It is worth noting that the embodiments provided by the present application make full use of the advantages of AIGC technology to intelligently analyze and process the review texts through algorithm models, improving the efficiency and accuracy of text processing. The system supports real-time monitoring of changes in review texts and generates regular reports at preset time intervals to ensure that users can timely understand public opinion trends. It can also reveal the differences, similarities, or potential trends between the two by comparing and analyzing the review texts of the target object and the benchmark object, providing valuable decision-making support for users. In addition, through the automated monitoring and report generation process, the manual intervention cost is significantly reduced, and the monitoring efficiency is improved. Moreover, by providing comprehensive review text analysis and comparative analysis functions, it can help users more accurately grasp market dynamics and public opinion trends, providing strong support for decision-making.

[0037] Figure 1 This is a schematic flowchart of an AI-based text processing method shown by this application according to an exemplary embodiment. As Figure 1 shown, the method provided in this embodiment includes: S101. Obtain a comment data set of a target object within a first time period.

[0038] In this step, obtain a comment data set of a target object within a first time period. The comment data set includes comment data made by different accounts on the target object within the first time period, and the comment data includes a comment tag and a comment text.

[0039] Specifically, to obtain a comment data set of a target object within a first time period, this obtaining behavior is based on the legal authorization of the data publisher, and the publisher can be the merchant itself, other merchants, or users, etc. The comment data set contains all comment data made by different accounts on the target object within the first time period. These comment data include two parts: a comment tag and a comment text. The comment tag is usually used to identify the category or theme of the comment, while the comment text is the specific evaluation content of the user. For example, a target object can be set, such as a certain e-commerce product. Determine the first time period, such as the past week, month, etc. Through web crawler technology or API interfaces, etc., obtain all comment data of the target object within the first time period from relevant platforms (such as e-commerce platforms, social media, etc.). Perform preprocessing on the obtained comment data, including steps such as deduplication and cleaning, to ensure the accuracy and effectiveness of the data.

[0040] S102. Determine a target comment data set from the comment data set according to a target comment tag.

[0041] In this step, determine a target comment data set from the comment data set according to a target comment tag, where the comment tag of the comment data in the target comment data set is the target comment tag.

[0042] Specifically, a target comment data set can be screened out from the comment data set according to a target comment tag. Among them, the comment data in the target comment data set has the same comment tag, that is, they all belong to the same category or theme. Among them, for a preset target comment tag, it can be customized according to the needs of the enterprise, such as satisfaction level (very satisfied, satisfied, average, dissatisfied, very dissatisfied), user attributes (age, gender, region, etc.), and the number of single-customer comments, etc. Traverse the comment data set, match the comment tag of the comment data with the target comment tag, and screen out the qualified comment data to form a target comment data set.

[0043] S103. Generate a prompt message according to the comment text of each comment data in the target comment data set.

[0044] In this step, prompt information is generated based on the review texts of each review data in the target review dataset. The prompt information includes a prompt statement and the text to be processed.

[0045] Specifically, prompt information can be generated according to the review texts of each review data in the target review dataset. Among them, the prompt information includes two parts: a prompt statement and the text to be processed. The prompt statement is used to indicate to the AIGC engine how to generate a reference review text, while the text to be processed contains the key information that needs to be processed by the AIGC engine.

[0046] In a possible implementation, it can be to extract keywords from the review texts in the target review dataset using a preset keyword database. The keyword database contains a series of keywords related to the target object and their weight factors. Generate a target review keyword vector based on the extracted keywords, and each target review data corresponds to a keyword vector. Generate a target review keyword matrix based on the target review keyword vectors corresponding to each target review data. The number of rows of the matrix can be determined according to the dimension of the keyword vector. If the number of keywords in a certain keyword vector is insufficient, use a preset value for filling. Traverse the target review keyword matrix, and add the target review keywords whose occurrence times are greater than the preset number of times to the text to be processed. These keywords represent the mainstream views or common problems in the review data. Further traverse the target review keyword matrix, calculate the correlation eigenvalue between keywords, and filter out the associated target review keyword phrases whose keyword correlation eigenvalues are greater than the preset correlation threshold. The associated keyword phrases can reveal the comprehensive evaluation of the reviewers on various aspects such as product features and usage experience. Use the weight factors in the preset keyword database to update the keyword correlation eigenvalues to more accurately evaluate the association strength between keywords. Add the filtered associated target review keyword phrases to the text to be processed as well. Generate a prompt statement according to the text to be processed, indicating to the AIGC engine to generate a reference review text based on the keywords and keyword phrases in the text to be processed.

[0047] S104. Input the prompt information into a preset AIGC engine to output the reference review text corresponding to the target review label.

[0048] Finally, the generated prompt information can be input into a preset AIGC engine to output the reference review text corresponding to the target review label. The AIGC engine generates high-quality reference review text based on the keywords, keyword phrases, and prompt statements in the prompt information. Appropriate AIGC engines can be selected, such as advanced natural language processing models like GPT and BERT. The generated prompt information is input into the AIGC engine, and the AIGC engine performs text generation based on the keywords, keyword phrases, and prompt statements in the prompt information to output the reference review text. Further, post-processing can be performed on the generated reference review text, such as grammar checking and spelling proofreading, to ensure the accuracy and readability of the text.

[0049] In this embodiment, by obtaining the review dataset of the target object in the first time period, then determining the target review dataset from the review dataset according to the target review label, generating prompt information based on the review texts of each review data in the target review dataset, and inputting the prompt information into a preset AIGC engine to output the reference review text corresponding to the target review label. Among them, the prompt information not only includes the prompt statement indicating the AIGC engine to generate the reference review text, but also includes the text to be processed, which consists of keywords and keyword phrases and is directly related to the user's core concerns. This intelligent way of generating prompt information helps the AIGC engine better understand the user's intention and generate high-quality reference review text.

[0050] Figure 2 is a schematic flowchart of an AI-based text processing method shown according to another exemplary embodiment of the present application. As Figure 2 shown, the method provided in this embodiment includes: S201. Obtain the review dataset of the target object in the first time period.

[0051] In this step, obtain the review dataset of the target object in the first time period. The review dataset includes the review data made by different accounts on the target object in the first time period, and the review data includes review labels and review texts.

[0052] In a possible application scenario, the target review label is used to associate one or more of a target satisfaction level, target user attributes, and the number of single-customer reviews, where the number of single-customer reviews is the number of reviews output by the same account before the end node of the first time period.

[0053] Specifically, obtain the comment dataset of the target object within the first time period. This acquisition behavior is based on the legal authorization of the data publisher, who can be the merchant itself, other merchants, or users, etc. The comment dataset contains all the comment data made by different accounts on the target object within the first time period. These comment data include two parts: comment tags and comment texts. Comment tags are usually used to identify the category or theme of the comment, while the comment text is the specific evaluation content of the user. For example, a target object can be set, such as a certain e-commerce product. Determine the first time period, such as the past week, month, etc. Through web crawler technology or API interfaces, etc., obtain all the comment data of the target object within the first time period from relevant platforms (such as e-commerce platforms, social media, etc.). Preprocess the obtained comment data, including steps such as deduplication and cleaning, to ensure the accuracy and effectiveness of the data.

[0054] S202. Determine the target comment dataset from the comment dataset according to the target comment tag.

[0055] In this step, determine the target comment dataset from the comment dataset according to the target comment tag, where the comment tags of the comment data in the target comment dataset are the target comment tags.

[0056] Specifically, the target comment dataset can be filtered out from the comment dataset according to the target comment tag. Among them, the comment data in the target comment dataset have the same comment tag, that is, they all belong to the same category or theme. Among them, for the preset target comment tags, they can be customized according to the needs of the enterprise, such as satisfaction levels (very satisfied, satisfied, average, dissatisfied, very dissatisfied), user attributes (age, gender, region, etc.), and the number of single-customer comments, etc. Traverse the comment dataset, match the comment tags of the comment data with the target comment tags, and filter out the qualified comment data to form the target comment dataset.

[0057] S203. Utilize the preset keyword database and process the target comment dataset.

[0058] In this step, the preset keyword database can be utilized and the target comment dataset Extract keywords from each comment text in it to generate the target comment keyword vector corresponding to each target comment data , where is the th target comment data in the target comment dataset , is the target comment keyword vector corresponding to

[0059] Specifically, a keyword database related to the field needs to be established. This database should contain common and representative vocabulary in the field, as well as their synonyms, near-synonyms, etc., to ensure the accuracy and comprehensiveness of keyword extraction. Then, advanced natural language processing techniques, such as TF-IDF (Term Frequency-Inverse Document Frequency), TextRank, or existing keyword extraction models based on deep learning, are used to extract keywords from each review text in the target review dataset. For each review text, the extracted keywords will be organized into a vector, where each element of the vector represents a keyword, and the value of the element can be the weight of the keyword (such as the TF-IDF value) or a simple existence identifier (0 or 1).

[0060] S204. Generate a target review keyword matrix based on the target review keyword vectors corresponding to each target review data.

[0061] In this step, based on the target review keyword vectors corresponding to each target review data generate a target review keyword matrix , for the target review keyword matrix is the number of target review keyword vectors in it.

[0062] Specifically, based on the target review keyword vectors generated in the previous step, these vectors can be arranged in order to form a matrix, where each column of the matrix represents the keyword vector of a review text. Optionally, before constructing the matrix, further processing of the keywords may be required, such as removing duplicates, merging synonyms, filtering out irrelevant vocabulary, etc., to ensure the accuracy and effectiveness of the matrix.

[0063] In a possible implementation, based on the target review keyword vectors corresponding to each target review data determine the number of rows of the target review keyword matrix based on the keyword vector with the highest dimension in to generate a target review keyword matrix :

[0064] where is the target review keyword matrix the th target review keyword vector in the th keyword, if the number of keywords in the target review keyword vector is less than less than , then configure the elements after to the preset values.

[0065] S205. Generate prompt information according to the target comment keyword matrix.

[0066] In this step, according to the target comment keyword matrix generate prompt information, where the text to be processed includes the keywords in the target comment keyword matrix and the prompt statement is used to instruct the preset AIGC engine to generate a reference comment text according to the text to be processed.

[0067] Furthermore, traverse each element in the target comment keyword matrix and add the target comment keywords whose occurrence times are greater than the preset times to the text to be processed to generate prompt information, where the prompt statement is used to instruct the preset AIGC engine to generate a reference comment text according to all the target comment keywords in the text to be processed.

[0068] Specifically, select representative keywords from the target comment keyword matrix to form the text to be processed. These keywords can be high-frequency words, high-weight words, or words closely related to a specific theme. Design one or more prompt statements according to the content and purpose of the text to be processed. These statements should be able to instruct the AIGC engine on how to generate a reference comment text based on the text to be processed. For example: "Please generate a comment on the product experience based on the following keywords: {keyword list}." Combine the text to be processed and the prompt statements into a complete prompt information. This information will be input into the AIGC engine as the basis for it to generate a reference comment text.

[0069] Suppose we have a comment dataset about a certain mobile phone. Through the above steps, we can generate the following prompt information: Text to be processed: {"The screen is clear", "The battery life is long-lasting", "The photo-taking effect is good", "The cost performance is high"} Prompt statement: "Please generate a positive comment on a certain mobile phone based on the following keywords." Complete prompt information: "Please generate a positive comment on a certain mobile phone based on the following keywords: The screen is clear, The battery life is long-lasting, The photo-taking effect is good, The cost performance is high." This prompt information will be input into the AIGC engine, and the engine will generate a reference comment text that meets the requirements based on these keywords and instructions.

[0070] Furthermore, it can also be to traverse the target comment keyword matrix each element in, and add the associated target comment keyword phrases with the keyword association feature value greater than the preset association feature threshold to the text to be processed. The associated target comment keyword phrases include the first target comment keyword and the second target comment keyword. The first target comment keyword and the second target comment keyword of the keyword association feature value is determined according to formula (1), and formula (1) is: Formula (1)

[0071] wherein, is the first target comment keyword and the second target comment keyword co-occur in the same target comment keyword vector in the target comment keyword matrix times, is the number of times the first target comment keyword appears in the target comment keyword matrix times, is the number of times the second target comment keyword appears in the target comment keyword matrix times.

[0072] In the above solution, by calculating the association feature values between keywords and screening out the keyword phrases with high association degrees, the semantic relationships in the comment data can be understood more deeply. Associated keyword phrases often can reveal the comprehensive evaluations of reviewers on product features, usage experiences, etc. Incorporating them into the text to be processed helps the AIGC engine generate richer, more accurate, and more insightful reference review texts. This ability to deepen the semantic understanding of the text is crucial for improving the quality of content generation and can provide more comprehensive and objective market feedback for e-commerce products. Among them, the screening of associated keyword phrases not only considers the occurrence frequency of individual keywords but also takes into account the co-occurrence relationship between keywords. This comprehensive consideration makes the generated prompt information more relevant and coherent. When the AIGC engine generates reference review texts based on the prompt information, these associated keyword phrases can be more naturally integrated, making the text content smoother and more organized. This improvement in relevance and coherence helps to enhance the readability and pertinence of the reference review texts. Moreover, by introducing the calculation of keyword association feature values, the refined processing and analysis of comment data are realized. Compared with the screening method that only relies on the occurrence frequency of keywords, this method can more accurately capture the subtle differences and potential associations in the comment data, making the generated prompt information more accurately reflect the true intentions and concerns of reviewers and providing more valuable input data for the AIGC engine.

[0073] Further, it can also be to use formula (2) and update the keyword association eigenvalue according to the preset weight factors of each keyword in the preset keyword database. The update is performed as follows, and formula (2) is: Formula (2)

[0074] Where is the keyword weight value of the first target comment keyword and the second target comment keyword in the th target comment keyword vector If the first target comment keyword and the second target comment keyword appear simultaneously in , then ; if the first target comment keyword and the second target comment keyword do not appear simultaneously in , then . is the preset weight factor of the first target comment keyword in the preset keyword database. is the preset weight factor of the second target comment keyword in the preset keyword database. is the preset weight factor of the th keyword in the th target comment keyword vector in the preset keyword database. is the number of columns of the target comment keyword matrix . is the number of rows of the target comment keyword matrix .

[0075] In the above solution, by introducing the weight factors in the preset keyword database, the association strength between keywords can be evaluated more accurately. The weight factors are usually determined based on the importance of keywords in a specific field or context. Therefore, updating the keyword association eigenvalue using these factors can make the association evaluation more in line with the actual context and domain knowledge. This improvement in accuracy helps to generate prompt information that is closer to the true comment intention, thereby improving the quality of the reference comment text generated by the AIGC engine. The keyword association eigenvalue adjusted by the weight factors can better reflect the importance of keywords in the comment data and their relationships with each other. When generating prompt information, incorporating these keyword phrases with high association and importance into the text to be processed helps the AIGC engine generate more targeted and in-depth reference comment text.

[0076] Among them, formula (2) finely adjusts the keyword association eigenvalue by introducing the weight factor in the preset keyword database. This adjustment process not only considers the co-occurrence times of keywords in the comments, but also combines the importance of keywords in a specific field or context, that is, the weight factor. This comprehensive consideration method makes the calculation of the keyword association eigenvalue more accurate and comprehensive, and can better reflect the actual association strength between keywords. The keyword association eigenvalue adjusted by formula (2) can more accurately guide the AIGC engine to select keyword phrases with higher relevance and importance when generating text. These keyword phrases can often more accurately reflect the true intentions and concerns of the reviewers, thereby improving the pertinence and relevance of the generated text. For e-commerce products, this means that the generated reference review text can more accurately reflect users' evaluations of product features, usage experiences, etc., and provide more valuable market feedback for merchants.

[0077] S206. Input the prompt information into the preset AIGC engine to output the reference review text corresponding to the target review label.

[0078] Finally, the generated prompt information can be input into the preset AIGC engine to output the reference review text corresponding to the target review label. The AIGC engine generates high-quality reference review text based on the keywords, keyword phrases, and prompt statements in the prompt information. Appropriate AIGC engines can be selected, such as advanced natural language processing models like GPT and BERT. Input the generated prompt information into the AIGC engine, and the AIGC engine generates text according to the keywords, keyword phrases, and prompt statements in the prompt information to output the reference review text. Further, post-processing can be performed on the generated reference review text, such as grammar checking, spelling correction, etc., to ensure the accuracy and readability of the text.

[0079] Based on the above embodiments, a monitoring instruction of the target object can also be obtained. The monitoring instruction is used to indicate obtaining another reference review text corresponding to another target object within the first time period, where the other target object is the benchmark object of the target object. Then, at preset time intervals, a monitoring report is generated and output. The monitoring report includes the reference review text and the other reference review text.

[0080] Specifically, the system needs to receive monitoring instructions input by the user or administrator. This instruction clarifies the monitoring target, scope, and time frame, especially specifying another target object to be monitored (i.e., the benchmark object) and the first time period to be monitored. The acquisition of monitoring instructions can be achieved through methods such as a graphical user interface (GUI), command-line interface (CLI), or application programming interface (API), ensuring flexibility and ease of use. For example, the user can select the "Add Benchmark Object Monitoring" function through the system's monitoring settings interface, and then input the identification information of the benchmark object (such as name, ID, etc.) and the time period (the first time period) they hope to monitor. The system then records and parses these instructions in preparation for subsequent operations.

[0081] After receiving the monitoring instructions, the system starts the monitoring program to capture relevant review texts in real-time or at regular intervals for the target object and its benchmark object within the first time period. This process may involve various technical means such as web crawler technology, API calls, or database queries to ensure that all relevant review data can be comprehensively and accurately collected.

[0082] For the acquisition of review texts for the target object and the benchmark object, the system needs to design an intelligent recognition mechanism to distinguish and store the review data of both separately. This may include methods such as keyword filtering, user identity recognition, or content context analysis to ensure the accuracy and integrity of the data.

[0083] The system automatically sorts and analyzes the collected review data at preset time intervals (such as daily, weekly, or monthly). The analysis content may include, but is not limited to, the number of reviews, sentiment tendency, keyword frequency, hot topics, etc., aiming to provide users with a comprehensive overview of the public opinion environment of the target object and its benchmark object during the specified time period.

[0084] Based on the above analysis, the system generates a monitoring report. The monitoring report adopts an intuitive and easy-to-read format (such as charts, lists, summaries, etc.) to display the reference review texts of the target object and the benchmark object, as well as the analysis results based on these texts. The report should also include a comparative analysis section to highlight the differences, similarities, or potential trends between the two, providing strong support for user decision-making.

[0085] Finally, the system outputs the generated monitoring report in the manner specified by the user. This can be sent via email, displayed in the system's internal message center, directly downloaded as a PDF or Excel file, or pushed to a third-party application through an API, etc. Ensure that users can conveniently obtain and view the report content and make timely responses or adjust strategies.

[0086] In the above solution, by obtaining the monitoring instructions of the target object and its benchmark object, it is possible to collect the reference review texts of both in a specific period in real time or regularly. These review texts not only reflect the direct feedback of consumers on the target object and its benchmark product, but also imply key information such as market trends and consumer preferences. Generating and outputting a monitoring report at a preset time interval enables the enterprise to quickly obtain the latest market dynamics and consumer feedback. This real-time monitoring and reporting mechanism helps the enterprise quickly respond to market changes, adjust product strategies, marketing strategies or customer service strategies to adapt to the changing market demands.

[0087] Figure 3 It is a schematic structural diagram of an AI-based text processing device shown according to an exemplary embodiment of the present application. As Figure 3 shown, the AI-based text processing device 300 provided in this embodiment includes: An acquisition module 310, configured to acquire a comment data set of a target object within a first period, where the comment data set includes comment data made by different accounts on the target object within the first period, and the comment data includes a comment tag and a review text; A processing module 320, configured to determine a target comment data set from the comment data set according to a target comment tag, where the comment tag of the comment data in the target comment data set is the target comment tag; A generation module 330, configured to generate a prompt message according to the review texts of each comment data in the target comment data set, where the prompt message includes a prompt statement and a text to be processed; An output module 340, configured to input the prompt message into a preset AIGC engine to output a reference review text corresponding to the target comment tag.

[0088] Optionally, the target object is an e-commerce product; The target comment tag is used to associate one or more of a target satisfaction level, a target user attribute, and the number of single-customer comments, where the number of single-customer comments is the number of comments output by the same account before the end node of the first period.

[0089] Optionally, the generation module 330 is specifically configured to: Utilize a preset keyword database and perform keyword extraction on each review text in the target comment data set to generate a target comment keyword vector corresponding to each target comment data , where is the th target comment data in the target comment data set , is The corresponding target comment keyword vector; Generate a target comment keyword matrix based on the target comment keyword vectors corresponding to each piece of target comment data ; ; Based on the target comment keyword matrix Generate the prompt information, where the text to be processed includes the keywords in the target comment keyword matrix The prompt statement is used to instruct the preset AIGC engine to generate the reference comment text according to the text to be processed.

[0090] Optionally, the generating module 330 is specifically configured to: Generate the target comment keyword matrix based on the keyword vector with the highest dimension in the target comment keyword vectors corresponding to each piece of target comment data Determine the number of rows of the target comment keyword matrix ; to generate the target comment keyword matrix :

[0091] where is the th target comment keyword vector in the target comment keyword matrix th keyword, if the number of keywords in the target comment keyword vector is less than , then configure the elements after to a preset value; Traverse each element in the target comment keyword matrix and add the target comment keywords with the number of occurrences greater than the preset number of occurrences to the text to be processed to generate the prompt information, where the prompt statement is used to instruct the preset AIGC engine to generate the reference comment text according to all the target comment keywords in the text to be processed.

[0092] Optionally, the generating module 330 is further specifically configured to: Traverse each element in the target comment keyword matrix and add the associated target comment keyword phrases with the keyword association eigenvalue greater than the preset association feature threshold to the text to be processed, where the associated target comment keyword phrases include the first target comment keyword and the second target comment keyword.

[0093] Optionally, the generating module 330 is further specifically configured to: Update the keyword correlation feature values according to the preset weight factors of each keyword in the preset keyword database.

[0094] Optionally, the obtaining module 310 is further configured to obtain a monitoring instruction of the target object, where the monitoring instruction is used to indicate obtaining another reference review text corresponding to another target object within the first time period, and the another target object is a benchmark object of the target object; The output module is further configured to generate and output a monitoring report at a preset time interval, where the monitoring report includes the reference review text and the another reference review text.

[0095] Figure 4 It is a schematic structural diagram of an electronic device shown by this application according to an exemplary embodiment. As Figure 4 shown, an electronic device 400 provided in this embodiment includes: a processor 401 and a memory 402; where: The memory 402 is used to store a computer program, and the memory may also be a flash (flash memory).

[0096] The processor 401 is configured to execute the execution instruction stored in the memory to implement each step in the above method. Specifically, reference may be made to the relevant descriptions in the foregoing method embodiments.

[0097] Optionally, the memory 402 may be either independent or integrated with the processor 401.

[0098] When the memory 402 is a device independent of the processor 401, the electronic device 400 may further include: A bus 403 for connecting the memory 402 and the processor 401.

[0099] This embodiment further provides a readable storage medium, where a computer program is stored in the readable storage medium. When at least one processor of the electronic device executes the computer program, the electronic device executes the methods provided by the above various embodiments.

[0100] This embodiment further provides a program product, which includes a computer program, and the computer program is stored in a readable storage medium. At least one processor of the electronic device can read the computer program from the readable storage medium, and at least one processor executes the computer program so that the electronic device implements the methods provided by the above various embodiments.

[0101] Those skilled in the art will readily conceive of other embodiments of the present application after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application, which follow the general principles of the present application and include known common general knowledge or conventional technical means in the technical field not disclosed in the present application. The specification and examples are only regarded as exemplary, and the true scope and spirit of the present application are pointed out by the claims.

[0102] It should be understood that the present application is not limited to the exact structures already described and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present application is only limited by the appended claims.

Claims

1. A text processing method based on AI, characterized in that: include: Acquire a comment data set of a target object in a first period of time, wherein the comment data set includes comment data made by different accounts on the target object in the first period of time, and the comment data includes comment tags and comment texts; Determining a target comment data set from the comment data set according to a target comment tag, wherein the comment tag of the comment data in the target comment data set is the target comment tag; Generate prompt information according to the comment text of each comment data in the target comment data set, wherein the prompt information includes a prompt sentence and a text to be processed; The prompt information is input into a preset AIGC engine to output a reference comment text corresponding to the target comment tag.

2. The AI-based text processing method according to claim 1, characterized in that: The target object is an e-commerce product; The target comment tag is used to associate one or more of the target satisfaction level, the target user attribute, and the number of single-customer comments, wherein the number of single-customer comments is the number of comments output by the same account before the end node of the first time period.

3. The AI-based text processing method according to claim 2, characterized in that: Prompt information is generated according to the comment text of each comment data in the target comment data set, including: Using a preset keyword database and reviewing the target dataset Extract keywords from each comment text in the target comment data to generate the target comment keyword vector corresponding to each target comment data. ,in, Review dataset for the target The Target review data, for The corresponding target review keyword vector; According to the target comment keyword vector corresponding to each target comment data Generate target review keyword matrix ; Review keyword matrix according to the stated goal Generate the prompt information, wherein the text to be processed includes the target review keyword matrix The prompt statement is used to instruct the preset AIGC engine to generate the reference comment text according to the text to be processed.

4. The AI-based text processing method according to claim 3, characterized in that: The target comment keyword vector corresponding to each target comment data Generate target review keyword matrix ,include: According to the target comment keyword vector corresponding to each target comment data The keyword vector with the highest dimension in the target review keyword matrix is ​​determined number of rows , to generate the target review keyword matrix : ； in, Review keyword matrix for the target Middle Target review keyword vector The keywords, if the target review keyword vector The number of keywords in Less than , then The following elements are configured as preset values; Correspondingly, the target review keyword matrix Generating the prompt information includes: Traverse the target review keyword matrix and adds the target comment keywords whose occurrence times are greater than the preset times to the text to be processed to generate the prompt information, wherein the prompt statement is used to instruct the preset AIGC engine to generate the reference comment text according to all the target comment keywords in the text to be processed.

5. The AI-based text processing method according to claim 4, characterized in that: The target review keyword matrix Generating the prompt information further includes: Traverse the target review keyword matrix , and add the associated target comment keyword phrase whose keyword associated feature value is greater than the preset associated feature threshold to the text to be processed, wherein the associated target comment keyword phrase includes the first target comment keyword and the second target comment keyword.

6. The AI-based text processing method according to claim 5, characterized in that: The traversal of the target review keyword matrix Each element in the text, and adding the associated target comment keyword phrase whose keyword associated feature value is greater than the preset associated feature threshold to the text to be processed, further comprising: The keyword-related feature value is updated according to the preset weight factor of each keyword in the preset keyword database.

7. The AI-based text processing method according to any one of claims 1 to 6, characterized in that: Also includes: Acquire a monitoring instruction of the target object, wherein the monitoring instruction is used to instruct acquisition of another reference comment text corresponding to another target object within the first time period, wherein the other target object is a benchmark object of the target object; Generate and output a monitoring report at a preset time interval, wherein the monitoring report includes the reference comment text and the other reference comment text.

8. An AI-based text processing device, characterized in that: include: An acquisition module, configured to acquire a comment data set of a target object within a first period of time, wherein the comment data set includes comment data made by different accounts on the target object within the first period of time, and the comment data includes comment tags and comment texts; A processing module, configured to determine a target comment data set from the comment data set according to a target comment tag, wherein the comment tag of the comment data in the target comment data set is the target comment tag; A generating module, configured to generate prompt information according to the comment text of each comment data in the target comment data set, wherein the prompt information includes a prompt statement and a text to be processed; The output module is used to input the prompt information into a preset AIGC engine to output a reference comment text corresponding to the target comment tag.

9. An electronic device, characterized in that: include: processor; as well as, A memory, configured to store executable instructions of the processor; The processor is configured to perform the method of any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 7 when executed by a processor.

Images