Semantic recognition method and system for automatically auditing activity copywriting based on artificial intelligence technology

Abstract

The invention provides a semantic recognition method and system for automatically auditing an activity copywriting based on an artificial intelligence technology. The semantic recognition method comprises the following steps: receiving an uploaded file which needs to be subjected to content inspection; extracting first document data of the to-be-checked file; carrying out word splitting processing on the first-time copywriting data based on a word segmentation algorithm to obtain second-time copywriting data; inputting the second copywriting data into a preset model for semantic understanding to obtain a copywriting analysis result; and performing content compliance auditing on the copywriting analysis result according to a pre-formulated auditing rule to generate an auditing report. According to the invention, the checking efficiency of the compliance of the to-be-published activity copywriting content can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a semantic recognition method and system for automatically reviewing activity documents based on artificial intelligence technology. Background Art

[0002] In order to protect the legitimate rights and interests of citizens, the state has promulgated a series of laws and regulations, requiring application providers to fulfill their main responsibilities for content management, strengthen the review of rules, advertisements, consultations and other copywriting content in promotional activities, and maintain a good network ecology. Pre-review of the copywriting content to be released can prevent illegal content such as infringement of user rights, destruction of the network ecology, suspected improper speech, etc., and maintain the ecological order of the Internet and user rights. At present, the difficulties in handling copywriting review are mainly concentrated in the following aspects:

[0003] (1) The existing audit method is generally conducted manually, which relies on the experience and professional quality of the auditors and is prone to subjective misjudgment.

[0004] (2) Promotional activities take various forms and involve a variety of types of copywriting, so traditional manual review is inefficient.

[0005] (3) The state has promulgated more and more regulations, and the corresponding audit rules are constantly increasing. The existing audit methods lack timely coverage of new regulations.

[0006] The solution of the prior art is: matching the text to be filtered with sensitive words according to the sensitive word library, and determining the sensitive word sentences of the text to be filtered according to the sensitive word matching results; generating sensitive word text according to sensitive words and sensitive word variants, inputting the sensitive word text into the large model for detection, and obtaining the sensitive text semantics; filtering the text to be checked with sensitive words according to the sensitive text semantics.

[0007] The existing technology for copywriting review is mainly based on sensitive word libraries. The detection accuracy depends on the richness of the word library and is not flexible enough. Reviewing through sensitive word libraries requires regular updates of the sensitive word library, and it is impossible to independently learn and update the judgment rules based on the input laws and regulations.

[0008] In summary, it is urgent to develop a semantic recognition method and system based on artificial intelligence technology to automatically identify and review activity copywriting. Summary of the invention

[0009] The present invention aims to provide a semantic recognition method and system for automatically reviewing activity documents based on artificial intelligence technology, so as to improve the review efficiency and accuracy of the document content.

[0010] In the first aspect, a semantic recognition method for automatically reviewing activity copywriting based on artificial intelligence technology includes the following process:

[0011] S100: receiving a to-be-published activity copywriting file uploaded by a user; extracting the first copywriting data in the file;

[0012] S200: performing word segmentation processing on the first copywriting data based on a word segmentation algorithm to obtain second copywriting data;

[0013] S300: Input the second copywriting data into a preset model for semantic understanding to obtain a copywriting analysis result;

[0014] S400: Conduct content compliance review on the text analysis results according to pre-established review rules and generate a review report. The review rules are formulated based on different compliance scenarios.

[0015] Preferably, in S100, extracting the first copy data in the file to be published specifically includes the following process: judging the file format according to the file extension of the content to be published, when the format is word format, calling the first API interface to extract data, and when the format is PDF, calling the second API interface to extract data to obtain the first copy data.

[0016] Preferably, the splitting process of the first copy data in S200 specifically includes the following process: filtering out redundant characters and erroneous characters in the first copy data based on a preset string processing method, and standardizing the text format to obtain the second copy data.

[0017] Preferably, S300 includes the following specific process: inputting the second copywriting data into a preset model, converting the context information of each word into a high-dimensional vector, inputting the result into the next model, and performing semantic understanding, entity recognition, and sentiment analysis accordingly to obtain the copywriting analysis result.

[0018] Preferably, S400 includes the following specific processes: performing content compliance rule detection and sensitive word detection on each word in the text analysis result one by one; when there are abnormal results, marking the abnormal description content and generating an audit report; wherein the abnormal results include violations of relevant laws, infringements on user rights, and the presence of sensitive words.

[0019] Second, a semantic recognition system for automatically reviewing activity copywriting based on artificial intelligence technology, including:

[0020] The copywriting data extraction module is used to receive the activity copywriting files to be published uploaded by users and extract the first copywriting data in the files;

[0021] A word splitting processing module for performing word splitting processing on the first copywriting data to obtain the second copywriting data;

[0022] A semantic understanding module for inputting the second copywriting data into a preset model for semantic understanding to obtain a copywriting analysis result;

[0023] A compliance audit module for performing content compliance audit on the copywriting analysis result according to the pre-established audit rules and generating an audit report.

[0024] The advantages of the present invention over the prior art include:

[0025] (1) Using artificial intelligence can quickly process a large amount of content, save manpower, reduce the waiting time of users, and can handle the audit of a large number of copywriting contents in Dalian, improving work efficiency.

[0026] (2) Through semantic understanding and sentiment analysis of the copywriting, analyzing the content with multiple meanings and making judgments based on the context is more time-saving and labor-saving than simply relying on manual work.

[0027] (3) Analyzing based on preset rules and models can ensure that all copywriting audits are analyzed using a unified standard process, and specific audit rules can be formulated according to different activity types, which can adapt to different activity copywriting and increase flexibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a schematic flow diagram of a semantic recognition method for automatically auditing activity copywriting based on artificial intelligence technology of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0029] A semantic recognition method for automatically auditing activity copywriting based on artificial intelligence technology includes the following processes:

[0030] S100: Receive the activity copywriting file to be published uploaded by the user and extract the first copywriting data from the file; specifically including: judging the file format according to the file extension of the content to be published, the file format including one or more of word, txt, pdf, calling the corresponding API interface to extract the content to be published according to the file format, when the format is word format, calling the first API interface to extract data, when the format is PDF, calling the second API interface to extract data, to obtain the first copywriting data; among them, word and PDF formats need to be parsed using tools or libraries of development languages, such as python-docx and PyPDF2.

[0031] S200: Perform word splitting on the first copywriting data based on a word segmentation algorithm to obtain the second copywriting data. Specifically, it includes: removing redundant characters, incorrect characters, and special characters from the first copywriting based on a preset string processing method, and performing spelling correction on the result to obtain the second copywriting data. Regular expressions and character set filtering methods can be used for the removal operation.

[0032] S300: Input the second copywriting data into a preset model for semantic understanding to obtain a copywriting analysis result. Specifically, it includes: inputting the second copywriting data into the preset model, converting the context information of each word into a high-dimensional vector, and inputting the result into the next model, and performing semantic understanding, entity recognition, and sentiment analysis in turn to obtain a copywriting analysis result. Among them, the semantic understanding model in the preset model generally uses BERT and GPT-4; BERT is suitable for understanding the context relationship of the text. In this example, BERT is selected as the semantic understanding model, and the training process steps include: data preparation, model construction, model training, and evaluation; BERT encodes the context information of words into high-dimensional vectors, enabling the model to understand the semantics of words in a specific context.

[0033] S400: Perform content compliance review on the copywriting analysis result according to the pre-established review rules to generate a review report. Specifically, it includes: detecting content compliance rules, sensitive word detection, and sentiment analysis for each word in the copywriting analysis result one by one; when there is an abnormal result, mark the abnormal description content and generate a review report; among them, the abnormal results include violations of relevant laws, infringement of user rights and interests, existence of sensitive vocabulary, and existence of extreme sentiment; the preset review rules can be compliant or non-compliant copywriting cases, including activity rules, product design specifications, requirement specifications, and user post content; when there is an abnormal result, mark the abnormal content, including non-compliant keywords, the laws and regulations relied on, and the non-compliant category.

Claims

1. A semantic recognition method for automatically reviewing activity copywriting based on artificial intelligence technology, characterized in that: The process includes the following: S100: receiving a to-be-published activity copywriting file uploaded by a user; extracting the first copywriting data in the file; S200: performing word segmentation processing on the first copywriting data based on a word segmentation algorithm to obtain second copywriting data; S300: Input the second copywriting data into a preset model for semantic understanding to obtain a copywriting analysis result; S400: Conduct content compliance review on the text analysis results according to pre-established review rules and generate a review report. The review rules are formulated based on different compliance scenarios.

2. According to claim 1, a semantic recognition method for automatically reviewing activity copywriting based on artificial intelligence technology is characterized in that: In the S100, extracting the first copy data in the file to be published specifically includes the following process: judging the file format according to the file extension of the content to be published, when the format is Word format, calling the first API interface to extract data, and when the format is PDF, calling the second API interface to extract data to obtain the first copy data.

3. According to claim 1, a semantic recognition method for automatically reviewing activity copywriting based on artificial intelligence technology is characterized in that: The splitting process of the first copy data in S200 specifically includes the following process: filtering out redundant characters and erroneous characters in the first copy data based on a preset character string processing method, and standardizing the text format to obtain the second copy data.

4. According to claim 1, a semantic recognition method for automatically reviewing activity copywriting based on artificial intelligence technology is characterized in that: The S300 includes the following specific processes: inputting the second copywriting data into a preset model, converting the context information of each word into a high-dimensional vector, inputting the result into the next model, and performing semantic understanding, entity recognition, and sentiment analysis accordingly to obtain the copywriting analysis result.

5. According to claim 1, a semantic recognition method for automatically reviewing activity copywriting based on artificial intelligence technology is characterized in that: The S400 includes the following specific processes: performing content compliance rule detection and sensitive word detection on each word in the text analysis result one by one; when there are abnormal results, marking the abnormal description content and generating an audit report; Abnormal results include violations of relevant laws, infringement of user rights, and the presence of sensitive words.

6. A semantic recognition system for automatically reviewing activity copywriting based on artificial intelligence technology, characterized in that: include: The copywriting data extraction module is used to receive the activity copywriting files to be published uploaded by users and extract the first copywriting data in the files; A word splitting processing module is used to perform word splitting processing on the first copywriting data to obtain the second copywriting data; The semantic understanding module is used to input the second copywriting data into the preset model for semantic understanding and obtain the copywriting analysis results; The compliance review module conducts content compliance review on the text analysis results according to pre-established review rules and generates an review report.

Images