Artificial intelligence-based advertisement copy automatic generation method

Abstract

The application relates to the technical field of text processing, in particular to an advertisement copy automatic generation method based on artificial intelligence, which comprises the following steps: collecting advertisement copies in a traffic delivery platform, identifying high-quality copies in the platform, determining the copy feature structure of the high-quality copies, respectively analyzing user keywords of access users and transaction users, determining user demand labels in different platforms, extracting key information words from function information of a target product based on the key demand labels of the users, identifying corresponding matching information words in the key information words of the target product based on the text structure of the high-quality copies, combining the identified matching information words according to the text structure of the high-quality copies, and automatically generating advertisement copies. The application automatically collects and analyzes the feature structure of high-quality copies in the platform through artificial intelligence, improves the production efficiency of the copies, and further improves the adaptability to the reading habits of the users of the platform based on the copies generated according to the feature structure of the platform.

Description

Technical Field

[0001] This invention relates to the field of text processing technology, and in particular to a method for automatically generating advertising copy based on artificial intelligence. Background Technology

[0002] With the rapid development of the internet advertising industry, the efficiency and accuracy of advertising copy production have become one of the core factors affecting marketing effectiveness.

[0003] However, existing methods rely heavily on human experience to select advertising platforms, without combining product lifecycle with user matching and traffic conversion efficiency of different platforms for quantitative analysis. This results in scattered traffic distribution, wasted resources, and an inability to achieve precise matching. At the same time, when automatically generating advertising copy, each step is independent and there is no AI-driven end-to-end collaborative mechanism. This makes it impossible to dynamically adjust the copy based on real-time data and adapt to rapidly changing market and user preferences. Summary of the Invention

[0004] The purpose of this invention is to solve the problems in the background art by proposing an automatic advertising copy generation method based on artificial intelligence.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: An AI-based method for automatically generating advertising copy, which includes the following steps: Step 1: Collect product information of the target product, obtain historical advertising information of the target product based on the product lifecycle, analyze the platform traffic coefficient of each platform in the advertising information, and determine the traffic placement platform based on the platform traffic coefficient. Step 2: Collect advertising copy from traffic delivery platforms, identify high-quality copy on the platform, and determine the structural characteristics of high-quality copy. Step 3: Obtain the access users and paying users of the traffic delivery platform, and analyze the user keywords of the access users and paying users respectively to determine the key demand coefficient PTm, and then determine the user demand tags in different platforms based on the key demand coefficient. Step 4: Based on the user's key needs tags, extract key information words from the functional information of the target product; Step 5: Based on the text structure of high-quality copy, identify corresponding matching words in the key information words of the target product, combine the identified matching words according to the text structure of high-quality copy, and then automatically generate advertising copy for the target platform.

[0006] As a further aspect of the present invention, the method for determining the traffic delivery platform includes: S1: Based on the target product's lifecycle, collect historical advertising information for the target product and mark it as sample information. The lifecycle refers to the entire market life process of a product from entering the market to gradually exiting the market. Advertising information includes advertising channels. S2: Extract the distribution channels from the sample information and collect the user access volume for each distribution channel. The user access volume refers to the number of times the target product is displayed on each distribution channel. Then, collect the access volume of the target product on each platform. The access volume includes the recommended access volume and the search access volume. Then, identify the corresponding transaction volume of the target product based on different access volumes. The transaction volume includes the recommended transaction volume and the search transaction volume. Choose any platform, divide the recommended transaction volume of this platform by the recommended visit volume to get the recommendation conversion rate, divide the search transaction volume by the search visit volume to get the search conversion rate, and then add the recommendation conversion rate to the search conversion rate. Mark the obtained value as the platform conversion rate Zi, where i represents different platforms, and i∈[1,I] indicates that there are I platforms in total. S3: Obtain the target product's page views on different platforms, get the platform page views Fi, and then use the formula... Get the platform access rate FLi; Obtain the platform conversion rate Zi and platform access rate FLi for different platforms, and then use the formula to obtain... Get the platform traffic coefficient Hi. and These are the proportionality coefficients; The platform traffic coefficient Hi is compared with the threshold coefficient Hy. If Hi < Hy, the platform is marked as an accidental platform for the target product. Otherwise, if Hi ≥ Hy, the platform is marked as the traffic delivery platform for the target product.

[0007] As a further aspect of the present invention, when the life cycle of the target product is in the introductory phase, there is no corresponding historical advertising information for the target product or the existing advertising information is less than the minimum sample threshold. At this time, based on the product type and product positioning of the target product, similar products are searched in the big data network, and advertising information of similar products is collected. At the same time, the advertising information of similar products is marked as sample information of the target product.

[0008] As a further aspect of the present invention, recommended visits refer to the access behavior generated after passively reaching the target product through the platform's recommendation page, while search visits refer to the access behavior of users actively entering keywords related to the target product in the platform's search box and entering the target product page through the search results page. Keywords include product names and category terms.

[0009] As a further aspect of the present invention, the method for determining the number of search visits is as follows: Obtain user behavior patterns and platform data logs. If, in the platform data logs, there is a search box input operation within t1 time before the user accesses the target product, if so, mark this access behavior as a search access and summarize it as search access volume. If not, mark this access behavior as a recommended access and summarize it as recommended access volume.

[0010] As a further aspect of the present invention, the method for identifying high-quality copywriting includes: The traffic delivery platforms are marked as target platforms in sequence. All advertising copy in the target platforms is collected, judgment indicators are set, and the advertising copy is sorted in descending order according to the judgment indicators. The top n advertising copy in each judgment indicator is selected and marked as key copy. The key copy in each judgment indicator is integrated into a copy set. Then identify whether there are overlapping key texts in the text sets with different judgment indicators. If they exist, mark the key text as high-quality text. If they do not exist, do not process the key text. High-quality copywriting is used as input data, and natural processing algorithms are used to perform semantic processing on the high-quality copywriting to determine the text structure. The text structure includes the copywriting's emotional tone, opening structure, and core content.

[0011] As a further aspect of the present invention, the judgment indicators are set as click-through rate, forwarding rate, and ad dwell time. When the number of high-quality copy is less than the sample copy threshold Y1, key copy is selected from the copy set and used as input data. Then, natural processing methods are used for semantic processing to determine the text structure.

[0012] As a further aspect of the present invention, the method for determining user demand tags includes: Based on the traffic of the target platform, users who visit the target product are obtained and marked as target users. Behavioral data of target users on the target platform is collected. The behavioral data includes the search keywords of target users and other products visited within time t2. Then, AI semantic parsing is used to extract the features of the visited products and mark the product features as user keywords. Product features include functional features, physical attributes, application scenarios and value positioning. Obtain the search keywords and user keywords of the target users, and then calculate the frequency of each keyword. Then compare the frequency with the frequency threshold Y2. If the frequency of a keyword is greater than or equal to the frequency threshold Y2, then mark the keyword as an access feature tag. Otherwise, if the frequency of a keyword is less than the frequency threshold Y2, then the corresponding keyword will not be processed. Acquire transacting users from the target platform and collect their behavioral data. Analyze the behavioral data of transacting users using the methods described above to determine the transacting user's transaction characteristic tags. Obtain transaction feature tags and access feature tags, and set tag weights for transaction feature tags and access feature tags respectively. and ,in, Weights for transaction feature labels To access feature label weights, ,and > ; Obtain the occurrence frequencies p1 and p2 of all feature tags, where p1 represents the occurrence frequency of the transaction feature tag and p2 represents the occurrence frequency of the visit feature tag. Then use the calculation formula... The key demand coefficient PTm is obtained, where m represents different feature labels; Obtain the key demand coefficient PTm for all feature labels, and compare the key demand coefficient PTm with the coefficient threshold Y3. If PTm ≥ Y3, then mark this feature label as a key demand label; otherwise, if PTm < Y3, then mark the corresponding feature label as a common label.

[0013] As a further aspect of the present invention, when a feature tag is both a transaction feature tag and an access feature tag, the occurrence frequencies p1 and p2 each have corresponding frequency values. If a feature tag is only a transaction feature tag, the occurrence frequency p1 of this transaction feature tag has a corresponding value, and the occurrence frequency p2 value is 0. Conversely, if a feature tag is only an access feature tag, the occurrence frequency p1 of this access feature tag has a value of 0, and the occurrence frequency p2 has a corresponding value.

[0014] As a further aspect of this invention, the key information words are extracted using the TF-IDF+keyword extraction model. This involves calculating the importance of words in the product text using TF-IDF, and then using keyword extraction tools to select high-frequency and core functional words as key information words. The keyword extraction tools include jieba and TextRank.

[0015] Compared with existing technologies, the advantages of this invention are: This invention combines the target product lifecycle with platform traffic coefficients to quantitatively screen traffic placement platforms, avoiding blind placement. It then calculates key demand coefficients by analyzing user visits and transactions, accurately extracting product keywords corresponding to users' high-priority needs. Based on the dual constraints of the platform's high-quality copywriting feature structure and user demand tags, the generated copy not only matches the platform scenario but also highlights the product's differentiated selling points, further reducing copywriting homogenization and ensuring brand consistency. Finally, artificial intelligence automatically collects and analyzes the feature structure of high-quality platform copywriting, replacing manual experience-based writing and further improving copywriting efficiency. Simultaneously, the copy generated based on the platform's feature structure has improved adaptability to platform users' reading habits, aligning with rapidly changing market and user preferences. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the method flow structure of the present invention. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Reference Figure 1 An AI-based method for automatically generating advertising copy, which includes the following steps: Step 1: Identify the target product and collect its product information. The target product refers to the product for which advertising copy needs to be generated. Product information includes product type, lifecycle stage, and brand positioning. Based on the target product's lifecycle, obtain its historical advertising data and analyze it to determine the traffic delivery platform. Specific methods for determining the traffic delivery platform include: S1: Based on the target product's lifecycle, collect historical advertising information for the target product and mark it as sample information. The lifecycle refers to the entire market life process of a product from entering the market to gradually exiting the market, including the introduction, growth, maturity, and decline stages. The advertising information includes advertising channels and advertising text. Furthermore, advertising channels include advertising locations and advertising platforms. It should be further explained that when the life cycle of the target product is in the introductory stage, there is no corresponding historical advertising information for the target product or the existing advertising information is less than the minimum sample threshold. At this time, based on the product type and product positioning of the target product, similar products are searched in the big data network and advertising information of similar products is collected. At the same time, the advertising information of similar products is marked as the sample information of the target product. Furthermore, the specific value of the minimum sample threshold is obtained by those skilled in the art after big data calculation. S2: Extract the advertising channels from the sample information and collect the user visits for each advertising channel. The user visits refer to the number of times the target product is placed in each advertising channel. For example, when the placement location is in a community elevator, the user visits of the target advertisement refer to the operating volume of the community elevator. Then, collect the access volume of the target product on each platform. The access volume includes the recommended access volume and the search access volume. Then, identify the corresponding transaction volume of the target product based on different access volumes. The transaction volume includes the recommended transaction volume and the search transaction volume. Choose any platform, divide the recommended transaction volume of this platform by the recommended visit volume to get the recommendation conversion rate, divide the search transaction volume by the search visit volume to get the search conversion rate, and then add the recommendation conversion rate to the search conversion rate. Mark the obtained value as the platform conversion rate Zi, where i represents different platforms, and i∈[1,I] indicates that there are I platforms in total. Among them, referral visits refer to the access behavior generated after passively reaching the target product through the platform's recommendation pages. Recommendation pages include the homepage information flow, the "You May Like" section, and the related product recommendation area. Search visits refer to the access behavior of users actively entering keywords related to the target product in the platform's search box and entering the target product page through the search results page. Keywords include product names and category terms. Furthermore, the method for judging search visits is as follows: obtain the user's behavior trajectory and platform data logs. If there is a search box input operation in the platform data logs within t1 time before the user accesses the target product, if so, this access behavior is marked as a search visit and summarized as search visits; if not, this access behavior is marked as a referral visit and summarized as referral visits. S3: Obtain the target product's page views on different platforms, get the platform page views Fi, and then use the formula... Get the platform access rate FLi; Obtain the platform conversion rate Zi and platform access rate FLi for different platforms, and then use the formula to obtain... Get the platform traffic coefficient Hi. and These are the proportionality coefficients; The platform traffic coefficient Hi is compared with the threshold coefficient Hy. If Hi < Hy, the platform is marked as an accidental platform for the target product. Otherwise, if Hi ≥ Hy, the platform is marked as the traffic delivery platform for the target product. Furthermore, the threshold coefficient Hy and the proportional coefficient and The specific values ​​were obtained by those skilled in the art based on big data calculations. Step Two: Collect advertising copy from traffic delivery platforms, identify high-quality copy within the platform, and determine the structural characteristics of high-quality copy. Specific methods for determining these structural characteristics include: Arbitrarily select a traffic delivery platform and mark it as the target platform. Collect all the advertising copy on the target platform, set judgment indicators, and sort the advertising copy in descending order according to the judgment indicators. Select the advertising copy in the first n positions of each judgment indicator and mark the selected advertising copy as the key copy. Integrate the key copy in each judgment indicator into a copy set. Then identify whether there are overlapping key texts in the text sets with different judgment indicators. If they exist, mark the key text as high-quality text. If they do not exist, do not process the key text. In this embodiment, the judgment indicators are set as click-through rate, forwarding rate, and ad dwell time. Furthermore, the judgment indicators are set by those skilled in the art based on big data experience, n is the threshold, and the specific values ​​are set by those skilled in the art based on big data experience. High-quality copywriting is used as input data, and natural processing algorithms are used to perform semantic processing on the high-quality copywriting to determine the text structure. The text structure includes the copywriting's emotional tone, opening structure, and core content. Furthermore, when the number of high-quality copy is less than the sample copy threshold Y1, key copy is selected from the copy collection and used as input data. Then, natural processing methods are used for semantic processing to determine the text structure. The specific value of Y1 is set by those skilled in the art based on big data experience. Step 3: Obtain the traffic and conversion rates of the traffic delivery platforms, analyze the user needs of these platforms, and determine the user need tags for different platforms. Specific methods for determining user need tags include: Based on the traffic volume of the target platform, acquire users who access the target product and mark them as target users. Collect behavioral data of target users on the target platform and determine the access feature tags of target users. The methods for determining access feature labels include: The behavioral data of the target users is obtained. The behavioral data includes the search keywords and other products visited by the target users within time t2. The characteristics of the visited products are then extracted using AI semantic parsing and marked as user keywords. Furthermore, t2 is a time threshold. The product characteristics include functional features, physical attributes, application scenarios, and value positioning. The search keywords and user keywords of the target users are obtained, and the occurrence frequency of each keyword is calculated. Then, the occurrence frequency is compared with the frequency threshold Y2. If the occurrence frequency of a keyword is greater than or equal to the frequency threshold Y2, the keyword is marked as an access feature tag. Otherwise, if the occurrence frequency of a keyword is less than the frequency threshold Y2, the corresponding keyword is not processed. The specific value of the frequency threshold Y2 is obtained by those skilled in the art based on big data calculations. Acquire transacting users from the target platform and collect their behavioral data. Analyze the behavioral data of transacting users using the methods described above to determine the transacting user's transaction characteristic tags. By combining visit feature tags with transaction feature tags for analysis, user demand tags can be determined. Further methods for determining user demand tags include: Obtain transaction feature tags and access feature tags, and set tag weights for transaction feature tags and access feature tags respectively. and ,in, Weights for transaction feature labels To access feature label weights, ,and > Furthermore, and The specific values ​​were obtained by those skilled in the art based on big data calculations. Obtain the occurrence frequencies p1 and p2 of all feature tags, where p1 represents the occurrence frequency of the transaction feature tag and p2 represents the occurrence frequency of the visit feature tag. Then use the calculation formula... The key demand coefficient PTm is obtained, where m represents different feature labels; It should be further explained that when a feature tag is both a transaction feature tag and an access feature tag, the occurrence frequencies p1 and p2 have corresponding frequency values. If a feature tag is only a transaction feature tag, the occurrence frequency p1 of this transaction feature tag has a corresponding value, and the occurrence frequency p2 value is 0. Conversely, if a feature tag is only an access feature tag, the occurrence frequency p1 of this access feature tag has a corresponding value, and the occurrence frequency p2 has a corresponding value. Next, the key demand coefficient PTm of all feature labels is obtained, and the key demand coefficient PTm is compared with the coefficient threshold Y3. If PTm≥Y3, the feature label is marked as a key demand label; otherwise, if PTm<Y3, the corresponding feature label is marked as a common label. The specific value of the coefficient threshold Y3 is obtained by those skilled in the art through big data calculation. Step 4: Obtain the functional information of the target product, and then extract key information words from the functional information of the target product based on the user's key needs tags. The functional information of the target product includes product manuals, details pages, parameter tables, and user reviews. The key information word extraction method adopts the TF-IDF + keyword extraction model, that is, the importance of words in the product text is calculated by TF-IDF, and combined with keyword extraction tools, high-frequency and core functional words are selected as key information words. Keyword extraction tools include jieba, TextRank, etc. Step 5: Obtain the text structure of high-quality copywriting on the target platform and the key information words of the target product. Based on the text structure of the high-quality copywriting, identify the corresponding matching information words in the key information words of the target product. Combine the identified matching information words according to the text structure of the high-quality copywriting to automatically generate advertising copy for the target platform.

[0019] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An AI-based method for automatically generating advertising copy, characterized in that: The method specifically includes the following steps: Step 1: Collect product information of the target product, obtain historical advertising information of the target product based on the product lifecycle, analyze the platform traffic coefficient of each platform in the advertising information, and determine the traffic placement platform based on the platform traffic coefficient. Step 2: Collect advertising copy from traffic delivery platforms, identify high-quality copy on the platform, and determine the structural characteristics of high-quality copy. Step 3: Obtain the access users and paying users of the traffic delivery platform, and analyze the user keywords of the access users and paying users respectively to determine the key demand coefficient, and then determine the user demand tags in different platforms based on the key demand coefficient. Step 4: Based on the user's key needs tags, extract key information words from the functional information of the target product; Step 5: Based on the text structure of high-quality copy, identify corresponding matching words in the key information words of the target product, combine the identified matching words according to the text structure of high-quality copy, and then automatically generate advertising copy for the target platform.

2. The method for automatically generating advertising copy based on artificial intelligence according to claim 1, characterized in that, Methods for determining traffic delivery platforms include: S1: Based on the target product's lifecycle, collect historical advertising information for the target product and mark it as sample information. The lifecycle refers to the entire market life process of a product from entering the market to gradually exiting the market. Advertising information includes advertising channels. S2: Extract the distribution channels from the sample information and collect the user access volume for each distribution channel. The user access volume refers to the number of times the target product is displayed on each distribution channel. Then, collect the access volume of the target product on each platform. The access volume includes the recommended access volume and the search access volume. Then, identify the corresponding transaction volume of the target product based on different access volumes. The transaction volume includes the recommended transaction volume and the search transaction volume. Choose any platform, divide the recommended transaction volume of this platform by the recommended visit volume to get the recommendation conversion rate, divide the search transaction volume by the search visit volume to get the search conversion rate, and then add the recommendation conversion rate to the search conversion rate. Mark the obtained value as the platform conversion rate Zi, where i represents different platforms, and i∈[1,I] indicates that there are I platforms in total. S3: Obtain the target product's page views on different platforms, get the platform page views Fi, and then use the formula... Get the platform access rate FLi; Obtain the platform conversion rate Zi and platform access rate FLi for different platforms, and then use the formula to obtain... Get the platform traffic coefficient Hi. and These are the proportionality coefficients; The platform traffic coefficient Hi is compared with the threshold coefficient Hy. If Hi < Hy, the platform is marked as an accidental platform for the target product. Otherwise, if Hi ≥ Hy, the platform is marked as the traffic delivery platform for the target product.

3. The method for automatically generating advertising copy based on artificial intelligence according to claim 2, characterized in that, When the target product is in the introductory phase of its lifecycle, there is no corresponding historical advertising information for the target product or the existing advertising information is less than the minimum sample threshold. At this time, based on the product type and product positioning of the target product, similar products are searched in the big data network, and advertising information of similar products is collected. At the same time, the advertising information of similar products is marked as sample information of the target product.

4. The method for automatically generating advertising copy based on artificial intelligence according to claim 2, characterized in that, Referral traffic refers to the access behavior generated after passively reaching the target product through the platform's recommendation page. Search traffic refers to the access behavior of users actively entering keywords related to the target product in the platform's search box and entering the target product page through the search results page. Keywords include product name and category terms.

5. The method for automatically generating advertising copy based on artificial intelligence according to claim 4, characterized in that, The method for determining search traffic is as follows: Obtain user behavior patterns and platform data logs. If, in the platform data logs, there is a search box input operation within t1 time before the user accesses the target product, if so, mark this access behavior as a search access and summarize it as search access volume. If not, mark this access behavior as a recommended access and summarize it as recommended access volume.

6. The method for automatically generating advertising copy based on artificial intelligence according to claim 1, characterized in that, Methods for identifying high-quality copywriting include: The traffic delivery platforms are marked as target platforms in sequence. All advertising copy in the target platforms is collected, judgment indicators are set, and the advertising copy is sorted in descending order according to the judgment indicators. The top n advertising copy in each judgment indicator is selected and marked as key copy. The key copy in each judgment indicator is integrated into a copy set. Then identify whether there are overlapping key texts in the text sets with different judgment indicators. If they exist, mark the key text as high-quality text. If they do not exist, do not process the key text. High-quality copywriting is used as input data, and natural processing algorithms are used to perform semantic processing on the high-quality copywriting to determine the text structure. The text structure includes the copywriting's emotional tone, opening structure, and core content.

7. The method for automatically generating advertising copy based on artificial intelligence according to claim 6, characterized in that, The evaluation metrics are set as click-through rate, forwarding rate, and ad dwell time. When the number of high-quality copy is less than the sample copy threshold Y1, key copy is selected from the copy collection and used as input data. Then, natural processing methods are used for semantic processing to determine the text structure.

8. The method for automatically generating advertising copy based on artificial intelligence according to claim 1, characterized in that, Methods for determining user demand tags include: Based on the traffic of the target platform, users who visit the target product are obtained and marked as target users. Behavioral data of target users on the target platform is collected. The behavioral data includes the search keywords of target users and other products visited within time t2. Then, AI semantic parsing is used to extract the features of the visited products and mark the product features as user keywords. Product features include functional features, physical attributes, application scenarios and value positioning. Obtain the search keywords and user keywords of the target users, and then calculate the frequency of each keyword. Then compare the frequency with the frequency threshold Y2. If the frequency of a keyword is greater than or equal to the frequency threshold Y2, then mark the keyword as an access feature tag. Otherwise, if the frequency of a keyword is less than the frequency threshold Y2, then the corresponding keyword will not be processed. Acquire transacting users from the target platform and collect their behavioral data. Analyze the behavioral data of transacting users using the methods described above to determine the transacting user's transaction characteristic tags. Obtain transaction feature tags and access feature tags, and set tag weights for transaction feature tags and access feature tags respectively. and ,in, Weights for transaction feature labels To access feature label weights, ,and > ; Obtain the occurrence frequencies p1 and p2 of all feature tags, where p1 represents the occurrence frequency of the transaction feature tag and p2 represents the occurrence frequency of the visit feature tag. Then use the calculation formula... The key demand coefficient PTm is obtained, where m represents different feature labels; Obtain the key demand coefficient PTm for all feature labels, and compare the key demand coefficient PTm with the coefficient threshold Y3. If PTm ≥ Y3, then mark this feature label as a key demand label; otherwise, if PTm < Y3, then mark the corresponding feature label as a common label.

9. The method for automatically generating advertising copy based on artificial intelligence according to claim 1, characterized in that, When a feature label is both a transaction feature label and an access feature label, the occurrence frequencies p1 and p2 each have corresponding frequency values. If a feature label is only a transaction feature label, the occurrence frequency p1 of this transaction feature label has a corresponding value, and the occurrence frequency p2 value is 0. Conversely, if a feature label is only an access feature label, the occurrence frequency p1 of this access feature label has a corresponding value, and the occurrence frequency p2 value has a corresponding value.

10. The method for automatically generating advertising copy based on artificial intelligence according to claim 1, characterized in that, The key information words are extracted using the TF-IDF + keyword extraction model. This model calculates the importance of words in the product text using TF-IDF and combines it with keyword extraction tools to select high-frequency and core functional words as key information words. The keyword extraction tools include jieba and TextRank.

Images