Article matching network red method and device based on large model, computer equipment and readable storage medium

Abstract

The invention discloses an article matching online red method and device based on a large model, computer equipment and a readable storage medium. The method comprises the following steps: firstly, acquiring multi-modal description information of a target article, an online red marketing knowledge base and a marketing tool library containing article understanding, searching and analyzing tools; and calling an article understanding tool by using a preset Netred big model, and obtaining comprehensive information of the article according to the multi-modal information. Thirdly, calling a network red search tool by the large model, and determining a plurality of to-be-determined network red accounts from the knowledge base according to the comprehensive information of the article; and finally, calling a network red data analysis tool by the large model, determining a target network red account from the undetermined accounts, and realizing accurate article matching network red.

Description

Technical Field

[0001] The present invention relates to the field of large model application technology, and in particular to a method, device, computer equipment and readable storage medium for matching items with internet celebrities based on large models. Background Art

[0002] In today's commercial marketing environment, influencer marketing has become an important means of promotion. With the wide variety of items on the market and the increasing number of influencers, how to accurately match items with suitable influencers has become a key issue. Traditional matching methods often lack systematicity and accuracy, and it is difficult to comprehensively consider the multi-faceted characteristics of items and the multi-faceted data of influencers. Summary of the invention

[0003] The object of the present invention is to provide a method, device, computer equipment and readable storage medium for matching items with influencers based on a large model.

[0004] In a first aspect, an embodiment of the present invention provides a method for matching items with influencers based on a large model, comprising:

[0005] Obtain multimodal item description information, an influencer marketing knowledge base, and an influencer marketing tool library for a target item; the influencer marketing tool library includes an item understanding tool, an influencer search tool, and an influencer data analysis tool;

[0006] The item understanding tool is called by using a preset internet celebrity model, and comprehensive item information of the target item is obtained according to the multimodal item description information;

[0007] The internet celebrity search tool is called by the preset internet celebrity model, and a plurality of pending internet celebrity accounts associated with the management of the target item are determined from the internet celebrity marketing knowledge base based on the comprehensive information of the item;

[0008] The internet celebrity data analysis tool is called through the preset internet celebrity model to determine the target internet celebrity account from the multiple pending internet celebrity accounts.

[0009] In a possible implementation, the object understanding tool includes an OCR image recognition tool, a web crawling tool, and a search engine tool;

[0010] The method of calling the item understanding tool by using the preset internet celebrity model and obtaining the comprehensive item information of the target item according to the multimodal item description information includes:

[0011] In the case where the multimodal item description information is an item picture, the OCR image recognition tool is called by using a preset internet celebrity model to obtain comprehensive item information of the target item;

[0012] In the case where the multimodal item description information is the URL of an item details webpage, the webpage crawling tool is called through a preset internet celebrity model to obtain comprehensive item information of the target item;

[0013] In the case where the multimodal item description information is a text description of the item, the search engine tool is called through a preset internet celebrity model to obtain comprehensive item information of the target item.

[0014] In a possible implementation, the comprehensive item information includes basic item information and a profile of a user of the item, and calling the internet celebrity search tool through the preset internet celebrity model to determine a plurality of pending internet celebrity accounts associated with the target item management from the internet celebrity marketing knowledge base based on the comprehensive item information includes:

[0015] Generate search keywords based on the basic information of the item and the profile of the user of the item;

[0016] The Internet celebrity search tool is called through the preset Internet celebrity model, and based on the region, language, number of fans, cooperation quotation and the search keyword, a matching search is performed in the Internet celebrity marketing knowledge base for the basic information of the item and the portrait of the user group of the item to obtain multiple pending Internet celebrity accounts associated with the management of the target item.

[0017] In a possible implementation, calling the internet celebrity data analysis tool through the preset internet celebrity model to determine a target internet celebrity account from the plurality of pending internet celebrity accounts includes:

[0018] The internet celebrity data analysis tool is called through the preset internet celebrity big model to score the internet celebrity matching degree of each of the pending internet celebrity accounts based on the internet celebrity name, internet celebrity introduction, internet celebrity category, internet celebrity tag, internet celebrity historical works, and internet celebrity historical marketing products;

[0019] The internet celebrity data analysis tool is called by the preset internet celebrity model to score the internet celebrity quality of each of the pending internet celebrity accounts based on the internet celebrity quality, internet celebrity cooperation quotation and zombie fan ratio;

[0020] Calculate the Internet celebrity comprehensive score corresponding to each of the pending Internet celebrity accounts according to the Internet celebrity matching score and Internet celebrity quality score corresponding to each of the pending Internet celebrity accounts;

[0021] The pending internet celebrity account whose internet celebrity comprehensive score exceeds a preset internet celebrity comprehensive score threshold is determined as the target internet celebrity account.

[0022] In a possible implementation, the method further includes:

[0023] Determine the popular area and hot marketing time period of the target item based on the comprehensive information of the item through the preset Internet celebrity model;

[0024] The calling of the internet celebrity search tool through the preset internet celebrity model and determining a plurality of pending internet celebrity accounts associated with the target item management from the internet celebrity marketing knowledge base based on the comprehensive item information includes:

[0025] The internet celebrity search tool is called through the preset internet celebrity model, and multiple pending internet celebrity accounts associated with the management of the target item are determined from the internet celebrity marketing knowledge base based on the comprehensive information of the item, the popular areas and the popular marketing time periods.

[0026] In a possible implementation, the method further includes:

[0027] Calling the internet celebrity data analysis tool through the preset internet celebrity model to obtain the fan regional distribution information of the internet celebrities corresponding to the multiple internet celebrity accounts to be determined;

[0028] According to the matching degree between the popular area of ​​the target item and the fan regional distribution information, the weight of the pending internet celebrity account is adjusted, wherein the higher the matching degree between the fan regional distribution and the popular area, the higher the weight of the pending internet celebrity account;

[0029] The calling the internet celebrity data analysis tool through the preset internet celebrity model to determine a target internet celebrity account from the plurality of pending internet celebrity accounts includes:

[0030] Based on the adjusted weights and the influencer matching score and influencer quality score, the adjusted influencer comprehensive score corresponding to each of the to-be-determined influencer accounts is calculated;

[0031] The pending internet celebrity account whose adjusted internet celebrity comprehensive score exceeds the preset adjusted internet celebrity comprehensive score threshold is determined as the target internet celebrity account.

[0032] In a possible implementation, the method further includes:

[0033] Classifying the fan interests of the influencer corresponding to the target influencer account;

[0034] For fan groups classified by different interests, a personalized marketing plan is generated for each target internet celebrity account through the preset internet celebrity model. The personalized marketing plan includes the characteristics of the target item highlighted according to different fans' interest points.

[0035] In a second aspect, an embodiment of the present invention provides an object matching internet celebrity device based on a large model, comprising:

[0036] An acquisition module is used to acquire multimodal item description information, an influencer marketing knowledge base and an influencer marketing tool library for a target item; the influencer marketing tool library includes an item understanding tool, an influencer search tool and an influencer data analysis tool; the item understanding tool is called through a preset influencer big model, and the comprehensive item information of the target item is obtained according to the multimodal item description information; the influencer search tool is called through the preset influencer big model, and a plurality of pending influencer accounts associated with the management of the target item are determined from the influencer marketing knowledge base based on the comprehensive item information;

[0037] The matching module is used to call the internet celebrity data analysis tool through the preset internet celebrity model to determine the target internet celebrity account from the multiple pending internet celebrity accounts.

[0038] In a third aspect, an embodiment of the present invention provides a computer device, comprising a processor and a non-volatile memory storing computer instructions, wherein when the computer instructions are executed by the processor, the computer device executes the method described in the first aspect.

[0039] In a fourth aspect, an embodiment of the present invention provides a readable storage medium, wherein the readable storage medium includes a computer program, and when the computer program is executed, the computer device where the readable storage medium is located is controlled to execute the method described in the first aspect.

[0040] Compared with the prior art, the beneficial effects provided by the present invention include: adopting a method, device, computer equipment and readable storage medium for matching items with influencers based on a large model disclosed in the present invention, by obtaining multimodal description information of the target item, an influencer marketing knowledge base and a marketing tool library containing tools for understanding, searching and analyzing items. The item understanding tool is called using the preset influencer large model, and the comprehensive information of the item is obtained from the multimodal information. Next, the large model calls the influencer search tool, and determines multiple pending influencer accounts from the knowledge base based on the comprehensive information of the item. Finally, the large model calls the influencer data analysis tool, determines the target influencer account from the pending accounts, and realizes accurate item matching with influencers. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can also be obtained based on these drawings without creative work.

[0042] Figure 1 A schematic diagram of the steps of the method for matching items with influencers based on a large model provided in an embodiment of the present invention;

[0043] Figure 2 A schematic block diagram of the structure of a device for matching internet celebrities with items based on a large model provided in an embodiment of the present invention;

[0044] Figure 3 A schematic block diagram of the structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0046] The specific implementation modes of the present invention are described in detail below in conjunction with the accompanying drawings.

[0047] In order to solve the technical problems in the aforementioned background technology, Figure 1 The flowchart of the method for matching items with influencers based on a large model provided in an embodiment of the present disclosure is provided below. The method for matching items with influencers based on a large model is introduced in detail.

[0048] Step S201, obtaining multimodal item description information, an influencer marketing knowledge base, and an influencer marketing tool library for a target item; the influencer marketing tool library includes an item understanding tool, an influencer search tool, and an influencer data analysis tool;

[0049] Step S202, calling the item understanding tool through a preset internet celebrity model, and obtaining comprehensive item information of the target item according to the multimodal item description information;

[0050] Step S203, calling the internet celebrity search tool through the preset internet celebrity model, and determining a plurality of pending internet celebrity accounts associated with the target item management from the internet celebrity marketing knowledge base based on the comprehensive information of the item;

[0051] Step S204: calling the internet celebrity data analysis tool through the preset internet celebrity model to determine a target internet celebrity account from the multiple pending internet celebrity accounts.

[0052] In an embodiment of the present invention, for example, it is assumed that the server is matching a new sports shoe of a fashion brand with an online celebrity. The multimodal item description information of this sports shoe may be in various forms. For example, the brand provides a high-definition sports shoe product picture, which shows the appearance of the shoe, including the unique shoe shape, color matching, the location of the brand logo, etc. At the same time, a webpage URL containing a detailed introduction to the sports shoe is also provided. On this webpage, there are descriptions of the material of the shoe (such as the use of high-tech breathable fabrics, special sole materials to provide better shock absorption, etc.), design inspiration (derived from a certain sports culture or fashion trend), the range of adaptation of different sizes, and price information. In addition, the brand also provides a text description, emphasizing the sports scenes suitable for this sports shoe (such as jogging, fitness training, etc.), the target consumer group (sports enthusiasts with an age range of about 20-40 years old) and the competitive advantages compared with other similar products (such as lighter, more fashionable, etc.).

[0053] The influencer marketing knowledge base in the server is a huge data set. It contains detailed information about many influencers, such as the influencer's name, region (e.g. some influencers are mainly active in Europe and the United States, while others are concentrated in Asia), the language used (English, Chinese, French, etc.), the number of fans (ranging from thousands to millions), the cooperation quotation (which varies greatly depending on the influencer's popularity and influence), the influencer's introduction (including the influencer's personal style, such as humor, professionalism, etc.), the influencer category (e.g. fashion, food, technology, etc.), the influencer label (e.g. "sports expert", "fashion pioneer", "beauty blogger", etc.), the influencer's historical works (including the videos or articles they have previously published, such as various fashion items that fashion influencers have shown, food tutorials that food influencers have made, etc.), and the influencer's historical marketing products (e.g. a clothing line of a certain brand that a fashion influencer has successfully promoted).

[0054] The item understanding tools, influencer search tools and influencer data analysis tools in the influencer marketing tool library are important means for the server to match items with influencers.

[0055] Object understanding tools include OCR image recognition tools, web crawling tools, and search engine tools. OCR image recognition tools can recognize text information in images, such as identifying the model logo or some special text descriptions that may exist on the shoes from the product image of sports shoes; web crawling tools can extract all relevant text information from the provided sports shoe details web page URL, including metadata in HTML, product descriptions, user reviews, etc.; search engine tools can search for relevant information in a wider range of Internet resources based on keywords in the text description (such as "sports shoes", "suitable for jogging", etc.) to further enrich the understanding of the target object.

[0056] The influencer search tool can search the influencer marketing knowledge base based on various conditions. For example, you can find influencers who may be related to the target product based on region, language, number of fans, cooperation quotes, and specific search keywords.

[0057] The influencer data analysis tool can analyze various data of influencers, such as analyzing the matching degree between influencers and target items based on the influencer name, influencer introduction, influencer category, influencer label, influencer historical works, and influencer historical marketing products. It can also evaluate the quality of influencers based on the quality of influencers, influencer cooperation quotations, and the proportion of zombie fans.

[0058] When the multimodal item description information received by the server is a product image of a sports shoe, the preset Internet celebrity model will call the OCR image recognition tool. For example, there is a string of small text on the side of the shoe in the picture, which may be the model of the product or a special technical mark. The OCR image recognition tool will recognize the text information in the picture and then match it with the existing database of sports shoe brands and models in the server. At the same time, the OCR tool can also identify the color distribution of the shoes in the picture, and determine the proportion of the main color and the auxiliary color through the color recognition algorithm. This recognized information, combined with the general knowledge about sports shoes pre-stored by the server (such as the popularity of certain color combinations among specific sports people), can obtain partial comprehensive information about this sports shoe, such as it may belong to a certain series (based on model identification), and the color style may be suitable for young and fashionable sports enthusiasts (based on color analysis and general knowledge).

[0059] If the server receives the URL of the sports shoe details page, the preset Internet celebrity model will call the web crawler. The web crawler will first visit the URL and then start crawling information according to the HTML structure of the web page. It will extract key information from the web page title. For example, the title may emphasize "new high-tech sports shoes", which directly indicates the new features of the product. Next, it will crawl the product description part to obtain detailed information such as the specific material, functional characteristics, etc. of the shoes. For example, the material part mentions that the upper uses a special weaving technology with good air permeability and elasticity, and the sole uses a new type of rubber material to provide better grip. The web crawler will also obtain information from the user evaluation part. If users generally evaluate that this shoe is comfortable to wear and stylish in appearance, then this information will also be included in the comprehensive information of the item. By crawling and sorting the entire web page information, the server obtains more comprehensive information about this sports shoe, including the technical characteristics of the product, user reputation, etc.

[0060] When the server receives a text description about sports shoes, the preset Internet celebrity model will call the search engine tool. Suppose the text description mentions "this sports shoe is suitable for jogging and fitness training, suitable for people with wider feet, and very light". The search engine tool will search for relevant information in a wider range of Internet resources based on these keywords ("jogging", "fitness training", "wide feet", "lightweight"). It may find some professional review articles about sports shoes suitable for jogging and fitness. These articles may mention some comparisons of similar products, as well as design points for people with wider feet. By integrating this searched information, the server can further improve the comprehensive information about this sports shoe, such as determining the unique selling points of this shoe among similar products suitable for jogging and fitness, and the design advantages for people with wider feet.

[0061] Based on the comprehensive information of sports shoes obtained above, the basic information of the item includes product name (sports shoes), brand, material (breathable fabric, special rubber sole), suitable sports scene (jogging, fitness training), etc. The user profile of the item is sports enthusiasts aged 20-40, with wide feet, etc. The server will generate search keywords based on this information. For example, the generated search keywords may include "sports shoes suitable for jogging", "sports equipment suitable for wide feet", "sports shoes recommended for 20-40 year old sports enthusiasts", "[brand name] sports shoes", etc.

[0062] The preset influencer model calls the influencer search tool to conduct a matching search in the influencer marketing knowledge base based on the basic information of the item and the portrait of the group using the item, according to the region, language, number of fans, cooperation quotation and these search keywords.

[0063] Assuming that the server first screens by region, if this pair of sneakers is mainly for the domestic market, then Internet celebrities active in the domestic region will be prioritized. Then based on language, if the target market is mainly Chinese speakers, Internet celebrities who use Chinese as the main language of communication will be screened out. For the number of fans, Internet celebrities with a fan base within a certain range (such as 100,000-500,000) may be screened out based on the brand's budget and expected promotion scope. These Internet celebrities have a certain audience base and are relatively in line with the budget. For cooperation quotations, Internet celebrities whose quotations are within the acceptable range of the brand will also be screened out.

[0064] Next, a more detailed match is performed based on the search keywords. For example, those influencers who are tagged as "sports experts" or have promoted sports equipment are screened. If the influencer's historical works contain content about jogging or fitness training, or if they have recommended products suitable for people with wider feet, then these influencers will be identified as pending influencer accounts associated with target item management. Suppose that through such a search and match, the server has identified 10 pending influencer accounts. These influencers are from different regions (but mainly domestic regions), use Chinese, have a fan base between 100,000 and 500,000, and the cooperation quotation is within the brand's budget. There is a certain correlation with the product characteristics of sports shoes and the target consumer group.

[0065] The preset influencer model calls the influencer data analysis tool to score the influencer matching degree of each pending influencer account based on the influencer name, influencer introduction, influencer category, influencer tag, influencer historical works, and influencer historical marketing products.

[0066] For example, there is an influencer named "Sports Little Expert". In his influencer profile, he mentioned that he is a blogger who focuses on sports equipment reviews. His influencer category is sports, and his tags include "jogging enthusiast" and "fitness equipment recommendation". His historical works include many evaluation videos of different brands of sports shoes, and he has also promoted similar sports shoes suitable for jogging and fitness in his historical marketing products. Then this influencer will get a higher score in terms of influencer matching. Another influencer may be named "Fashion Lifestyle". Although he occasionally involves sports-related content, he mainly focuses on fashion and lifestyle content. His influencer category and tags are less relevant to sports shoes, and there is less content about sports shoes in his historical works, so his score in terms of influencer matching will be relatively low.

[0067] The server calls the internet celebrity data analysis tool through the preset internet celebrity model, and assigns an internet celebrity quality score to each pending internet celebrity account based on the internet celebrity quality, internet celebrity cooperation quotation, and the proportion of zombie fans.

[0068] Regarding the quality of influencers, if an influencer's videos or articles are well-made, have high views and interaction rates (such as likes, comments, shares, etc.), then his quality will be considered high. For example, if an influencer's videos have an average of more than 100,000 views, an average of more than 5,000 likes, and are relatively active in commenting and sharing, then he will get a higher score in terms of influencer quality. Regarding the quote for cooperation, if an influencer's quote is relatively high, but his number of fans and influence are not significantly higher than other influencers, then his score in this regard will be lower. Regarding the proportion of zombie fans, if data analysis finds that a certain influencer's fans have a high proportion (such as more than 30%) of zombie fans (i.e. fans who rarely interact), then his score in terms of influencer quality will also be lower.

[0069] The comprehensive score of each pending internet celebrity account is calculated based on the internet celebrity matching score and internet celebrity quality score corresponding to each pending internet celebrity account. Assuming that the full score of the internet celebrity matching score is 100 points and the full score of the internet celebrity quality score is 100 points, the calculation is performed according to certain weights (such as the internet celebrity matching score accounts for 60% and the internet celebrity quality score accounts for 40%). For example, the internet celebrity matching score of "Sports Little Expert" is 80 points and the internet celebrity quality score is 70 points, then his comprehensive score = 80 × 0.6 + 70 × 0.4 = 76 points.

[0070] The pending influencer accounts whose comprehensive scores exceed the preset threshold (assuming 70 points) are identified as target influencer accounts. According to the previous calculation, if the comprehensive score of "Sports Little Master" is 76 points, which exceeds 70 points, then he will be identified as the target influencer account. After calculating and screening all 10 pending influencer accounts, 5 target influencer accounts may be finally determined, and these influencers will be recommended to the brand for the promotion of sports shoes.

[0071] In an embodiment of the present invention, the object understanding tool includes an OCR image recognition tool, a web crawling tool and a search engine tool; the object understanding tool is called through a preset Internet celebrity model, and the comprehensive object information of the target object is obtained according to the multimodal object description information, which can be implemented through the following examples.

[0072] In the case where the multimodal item description information is an item picture, the OCR image recognition tool is called by using a preset internet celebrity model to obtain comprehensive item information of the target item;

[0073] In the case where the multimodal item description information is the URL of an item details webpage, the webpage crawling tool is called through a preset internet celebrity model to obtain comprehensive item information of the target item;

[0074] In the case where the multimodal item description information is a text description of the item, the search engine tool is called through a preset internet celebrity model to obtain comprehensive item information of the target item.

[0075] In the embodiment of the present invention, for example, when the server processes a handbag promotion task of a fashion brand, the received multimodal item description information is a beautiful picture of a handbag. The handbag has a unique design, with complex textures and decorative patterns on the surface, and the brand logo is prominently printed on the bag surface.

[0076] The preset Internet celebrity model in the server calls the OCR image recognition tool to start processing this image. The OCR image recognition tool first pre-processes the image, including adjusting the contrast and brightness of the image to ensure that the text in the image can be clearly recognized. For this handbag image, since the brand logo and some decorative text on the bag may use special fonts, the OCR tool will use its built-in multiple font recognition algorithms to recognize it.

[0077] During the recognition process, the OCR tool successfully identified the brand logo and determined the brand name of the handbag by comparing it with the fashion brand database stored in the server. At the same time, there may be some small labels describing the material of the handbag or the name of the series on the picture, and the OCR tool also accurately recognized these text information. For example, it recognized words such as "genuine leather material" and "limited edition series".

[0078] In addition, the zipper part of the handbag may have a brand-customized zipper head with some small words engraved on it. The OCR tool can also recognize this text information, which may imply some special design or high-end positioning of the handbag.

[0079] Based on the identified information, the server begins to build comprehensive information about the handbag. First, the brand information is determined, which is crucial for subsequent search for matching influencers, as different fashion brands may have different audiences. The identification result of the leather material indicates that this is a relatively high-end handbag, and when looking for influencers, you can prefer those who focus on high-end fashion products. The limited edition series information can be used as a unique selling point, which can be emphasized when communicating marketing plans with influencers.

[0080] Assume that the server needs to match influencers for a newly launched smartphone, and the multimodal item description information is a promotional picture of the smartphone. The front of the smartphone shows its large screen and narrow bezel design, and the back has a unique camera arrangement and the smartphone brand and model logo.

[0081] The preset Internet celebrity model calls the OCR image recognition tool to process the mobile phone image. For the screen part, although there is no direct text information, the OCR tool can obtain relevant information by identifying some icons or system interface elements displayed on the screen. For example, identifying that the mobile phone is using a specific operating system is helpful for determining the function and user group of the mobile phone.

[0082] When identifying the brand and model logo on the back of the phone, the OCR tool accurately obtains the specific model of the phone. At the same time, there may be some small words about the camera parameters around the camera, such as "48 million pixel main camera" and "ultra-wide-angle lens". The OCR tool recognizes these parameter information and passes it to the server.

[0083] The server builds comprehensive information about smartphones based on the identified information. The brand and model of the phone are basic information, and different models of phones have different market positioning and target users. Camera parameters are important product features, which help to screen out those who often share photography or are interested in the camera function of mobile phones when looking for influencers. For example, some influencers who focus on travel photography may be interested in the camera function of this phone because its powerful camera configuration can attract their fans.

[0084] For the promotion of a smart lamp, the server receives its product picture as the multimodal item description information. The smart lamp has a unique appearance, with some touch control buttons on the lamp body and some logo text about the light mode and function next to it.

[0085] After the preset internet celebrity model calls the OCR image recognition tool, the text near the control button on the lamp body is first recognized. Functional labels such as "cold and warm light switching", "brightness adjustment", and "intelligent voice control" are recognized. This information directly reflects the main functional characteristics of the smart lamp.

[0086] At the same time, the base or side of the lamp may have the brand name and model of the product, and the OCR tool can accurately identify this information. Through the identified brand name, the server can determine the brand series to which the smart lamp belongs. Different home furnishing brands have different positioning and audiences in the home furnishing market. For example, some brands are known for high-end, intelligent home furnishing products, while others focus on cost-effectiveness.

[0087] Based on these recognition results, the server constructs comprehensive information about the smart lamp. The functional characteristics of the light will be an important basis for finding influencers, such as those who focus on smart home life or home decoration, because the function and appearance of smart lamps have a great impact on the home environment. Brand and model information will help further refine the scope of target influencers to ensure that the most suitable influencer for promoting this smart lamp is found.

[0088] When the server wants to promote a cosmetic, the multimodal item description information received is the URL of the cosmetic's detailed webpage. The preset Internet celebrity model calls the web crawler to start accessing this URL.

[0089] The web crawler first parses the HTML structure of the web page and finds the title of the web page. The title may contain the brand name of the cosmetics, the product name, and some key features, such as "[Brand Name] Whitening Essence - Highly Effective for Brightening Skin Tone". This provides important information for the server to initially understand the product.

[0090] Next, the web crawler began to crawl the product description section of the web page. In the product description, the ingredients of this whitening essence are introduced in detail, such as "contains niacinamide, arbutin and other effective whitening ingredients". At the same time, it also mentioned the skin type that the product is suitable for, "suitable for all skin types, especially dull skin", and the method of use, "after cleansing every morning and evening, take an appropriate amount of essence and apply it evenly on the face".

[0091] Web crawling tools also obtain the content of user reviews. There may be some positive feedback in user reviews, such as "After using it for a week, the skin tone has been significantly brightened" and "The texture is very light and easy to absorb", and there may also be some negative feedback, such as "It is a bit irritating to sensitive skin." These user review information is also an important part of building comprehensive information about the item.

[0092] The server builds comprehensive information about cosmetic items based on the information crawled from the web pages. The ingredient information of a product can help determine its positioning in the cosmetics market. For example, an essence containing popular whitening ingredients such as niacinamide may be more popular with consumers who are concerned about whitening effects. Information such as applicable skin types and usage methods can help screen out those influencers whose fan base matches the target users when looking for influencers. User reviews can give the server a more comprehensive understanding of the advantages and disadvantages of a product, so that when communicating marketing plans with influencers, the server can emphasize the advantages and solve possible problems in a targeted manner.

[0093] Suppose the server is processing a task of matching influencers with fitness equipment, and the multimodal item description information is the URL of the detailed webpage of the fitness equipment. After the web crawler accesses the URL, it first obtains the key information in the title of the webpage, such as "[Brand Name] Multifunctional Fitness Equipment - Essential for Family Fitness".

[0094] In the product description section, the crawler tool obtains the functional features of fitness equipment, such as "can perform a variety of strength training, including bench press, squats, etc.", "made of high-strength steel, sturdy and durable", "adjustable weight and angle, suitable for different training needs", etc. At the same time, it also obtains product size information, "small footprint, suitable for home use."

[0095] The user reviews section is also a key part of web crawling. Users may mention information such as "It is very convenient to use and you can do full body exercise at home" and "The installation process is a bit complicated and requires some professional guidance".

[0096] The server builds comprehensive information about fitness equipment based on the crawled information. The functional characteristics of fitness equipment determine the user groups it is suitable for, such as fitness enthusiasts who want to do strength training at home. Size information is also important for determining the target market. For example, the characteristics suitable for home use can attract people who have limited living space but want to exercise. Installation issues in user reviews can remind influencers to emphasize to their fans that they can provide installation guidance and other services when communicating marketing plans with influencers.

[0097] For the promotion of a new book, the server receives the URL of the book's detailed webpage as the multimodal item description information. After the web crawler accesses the URL, it parses the title of the webpage, such as "[Book Title] - A literary masterpiece exploring human nature".

[0098] In the product description section, the crawler tool obtains the book's content introduction, "This book explores the complexity and beauty of human nature by telling the protagonist's growth experience", "It combines a variety of literary techniques, and the plot is ups and downs", etc. It also obtains the author information of the book, "The author is a well-known writer who has won many literary awards."

[0099] The user review section may include readers' comments on the book, such as "the writing is beautiful and fascinating", "the story is profound and thought-provoking", and other positive comments. There may also be some negative comments, such as "the beginning is a bit dragged out".

[0100] The server builds comprehensive information about the book based on this information. The book's content and author information are important basic information that can help determine the book's audience, such as literature lovers who like to explore the theme of human nature. User reviews can be used as a reference when formulating marketing plans. Positive reviews can be highlighted in publicity, and negative reviews can be considered for compensation in marketing, such as emphasizing the exciting content in the second half of the book.

[0101] When the server is matching a travel product with an influencer, the multimodal item description information it receives is a text description of the travel product. For example, the text description is "Our travel package includes a seven-day trip to Europe, covering famous cities such as Paris, France, Rome and Venice, Italy. Tourists will stay in local four-star hotels, taste local specialties, and visit world-renowned art museums and historical sites. The tour group is equipped with professional tour guides who provide full explanation services."

[0102] The preset Internet celebrity model calls the search engine tool, which searches based on the keywords in the text. First, the keywords "European trip", "Paris", "Rome", "Venice", "four-star hotel", "special food", "art museum", "historical monuments", "professional tour guide" and so on are extracted.

[0103] The search engine tool will search for information related to these keywords. For example, search for popular tourist attractions in Paris, Rome and Venice, detailed introductions to local food culture, and local reviews of four-star hotels. Through the search, you may find that the Eiffel Tower in Paris, the Colosseum in Rome, and the canals in Venice are must-see attractions for tourists, and that local French cuisine, Italian pizza and pasta are very popular foods.

[0104] The server builds comprehensive information about tourism products based on the information searched. Tourist destinations and attractions are important factors in attracting tourists. When looking for influencers, you can look for those who often share their travel experiences or are interested in European culture. Information such as the hotel's star rating and cuisine can also be used as selling points for tourism products, and can be emphasized when communicating marketing plans with influencers.

[0105] Suppose the server wants to promote a gourmet product, and the text description is "Our cake is made of fresh strawberries, high-quality flour, pure cream and eggs. The cake has a beautiful appearance, rich strawberry aroma, and a soft and dense texture. It is suitable as an afternoon tea snack or birthday cake, and is suitable for people of all ages."

[0106] After the preset internet celebrity model calls the search engine tool, it extracts keywords such as “strawberry cake”, “fresh strawberries”, “high-quality flour”, “pure cream”, “eggs”, “afternoon tea snacks”, “birthday cake”, “all age groups”, etc.

[0107] Search engine tools search for information related to these keywords. For example, search for the nutritional value of fresh strawberries, the quality standards of high-quality flour and pure cream, and the popularity of strawberry cakes in different occasions (afternoon tea, birthdays). Through the search, it is found that strawberries are rich in vitamin C. In the afternoon tea scene, strawberry cake with a cup of tea is the choice of many people, while in the birthday scene, the beautiful appearance of strawberry cake is suitable as a dessert for birthday celebrations.

[0108] The server constructs comprehensive information about the food product based on the searched information. The raw material information of the cake can reflect its quality. When looking for influencers, you can look for those who pay attention to the quality of food ingredients. Suitable eating scenes and audience age range are also important information. For example, you can look for influencers whose fan base has a wide age span and who often share food to promote this cake.

[0109] The promotion of an online education course is described as follows: "Our online education course is about the introduction to programming. The course is taught by experienced programming experts and uses an interactive teaching method. Through a large number of examples and projects, students can quickly master the basics of programming. The course is suitable for beginners with zero basic knowledge, both students and adults who want to switch to the field of programming. The course content includes the basic syntax of programming languages, data structures, algorithms and other knowledge."

[0110] The preset internet celebrity model calls the search engine tool to extract keywords such as “introductory programming course”, “experienced experts”, “interactive teaching”, “beginners with zero foundation”, “students”, “adults changing careers”, “basics of programming languages”, “data structure”, and “algorithm”.

[0111] Search engine tools search for information related to these keywords. For example, searching for the market demand for introductory programming courses shows that more and more people want to learn programming, especially students with no basic knowledge and adults who want to change careers. They also search for information about the importance of teaching data structures and algorithms in programming courses.

[0112] The server builds comprehensive information about educational courses based on the information searched. The target audience of the course (students and adults who want to change careers) and the teaching method (interactive teaching) are important information. When looking for influencers, you can look for those who focus on the field of education, especially those who are interested in programming education and have many students or people who want to change careers in their fan base. The key knowledge points in the course content (data structure, algorithm, etc.) can also be used as selling points for the course, and they should be emphasized when communicating marketing plans with influencers.

[0113] In an embodiment of the present invention, the comprehensive information of the item includes basic information of the item and a portrait of the group of people who use the item. The Internet celebrity search tool is called through the preset Internet celebrity model, and multiple pending Internet celebrity accounts associated with the management of the target item are determined from the Internet celebrity marketing knowledge base based on the comprehensive information of the item. The implementation can be performed through the following example.

[0114] Generate search keywords based on the basic information of the item and the profile of the user of the item;

[0115] The Internet celebrity search tool is called through the preset Internet celebrity model, and based on the region, language, number of fans, cooperation quotation and the search keyword, a matching search is performed in the Internet celebrity marketing knowledge base for the basic information of the item and the portrait of the user group of the item to obtain multiple pending Internet celebrity accounts associated with the management of the target item.

[0116] In the embodiment of the present invention, for example, for a high-end watch of a fashion brand, the basic information of the item includes the brand name (such as "Rolex"), the watch type (such as a mechanical watch), the watch body material (such as 18K gold, stainless steel), functional characteristics (such as waterproof depth, timing accuracy), etc. The profile of the user of the item is 30-60 years old, mostly male, high-income group, and people who have a high pursuit of fashion and quality.

[0117] Based on this information, the search keywords generated by the server may include "Rolex mechanical watch", "18K gold watch recommendation", "watches suitable for high-income men", "high-end fashion watches", "waterproof mechanical watches", etc. These keywords cover the basic attributes of watches and the characteristics of the target population, and can accurately locate relevant Internet celebrities in subsequent searches.

[0118] The basic information of infant milk powder includes brand (such as "Mead Johnson"), milk powder type (such as infant formula milk powder, organic milk powder), applicable age group (such as 0-12 months old infants), nutritional ingredients (such as containing DHA, ARA, etc.), etc. The user profile of the item is new mothers or expectant mothers, mainly aged 20-35 years old, who are concerned about the healthy growth of their babies.

[0119] The search keywords generated by the server would be "Mead Johnson baby milk powder", "milk powder suitable for babies aged 0-12 months", "baby milk powder containing DHA", "milk powder recommended by new mothers", "organic baby milk powder", etc. These keywords combine the characteristics of the milk powder itself and the characteristics of the target user group, which helps to find suitable influencers for promotion.

[0120] The basic information of tablet computers includes brand (such as "Apple iPad"), screen size (such as 10.2 inches), operating system (such as iOS), special functions (such as support for Apple Pencil, suitable for drawing and note-taking), etc. The user group of the items is students (aged 6-25), creative workers (such as designers, illustrators), etc.

[0121] The search keywords generated by the server include "Apple iPad 10.2 inches", "tablets suitable for students", "tablets that support Apple Pencil", "iOS system tablets", "tablets recommended by designers", etc. These keywords can accurately reflect the characteristics of tablets and the people they are suitable for, providing an effective basis for searching for Internet celebrities.

[0122] The preset influencer model calls the influencer search tool and starts searching in the influencer marketing knowledge base based on the search keywords generated previously.

[0123] First, consider the region. If the watch is mainly for the domestic market, then prioritize searching for influencers who are active in China. For language, if the target audience is mainly Chinese speakers, select influencers who use Chinese as their main language of communication.

[0124] In terms of the number of fans, since watches are high-end products, brands may want to find influencers with a large number of fans and a certain influence to match their brand image, so they will screen out influencers with more than 500,000 fans. For cooperation quotations, according to the brand's budget, select influencers with quotations within an acceptable range.

[0125] Then match according to the search keywords. For example, there are some fashion influencers whose tags are "high-end fashion recommendation" and "men's fashion". Their historical works often display and recommend high-end watches, such as Rolex and other brands. The age distribution of the fan groups of these influencers also overlaps with the target population (30-60 years old men), and they will be identified as pending influencer accounts associated with the management of this watch. Assume that through such search and screening, the server has identified 5 pending influencer accounts, which meet the requirements in terms of region, language, number of fans, cooperation quotation, and relevance to watches.

[0126] After calling the influencer search tool with the preset influencer model, filter by region. If milk powder is mainly sold in China, search for influencers in China. In terms of language, since the target audience is new mothers or expectant mothers, who mainly speak Chinese, choose Chinese influencers.

[0127] In terms of the number of fans, considering that the promotion of maternal and infant products requires a certain coverage, we screened out influencers with 200,000 to 500,000 fans. For cooperation quotations, we screened according to the brand's budget.

[0128] According to the search keyword matching, those mother and baby Internet celebrities with tags such as "parenting expert" and "mother and baby recommendation" have a lot of content about recommending infant formula in their historical works, and a large part of their fan base are women aged 20-35, which is consistent with the target user profile of infant formula. Through such search and screening, the server identified 8 pending Internet celebrity accounts, which have a high correlation with infant formula in various aspects.

[0129] The preset Internet celebrity model calls the Internet celebrity search tool. From the region, if the tablet is mainly aimed at the domestic and some Asian markets, search for Internet celebrities active in these regions. In terms of language, since most students and creative workers use Chinese or English, filter out Internet celebrities who use these two languages.

[0130] In terms of the number of fans, we screened out influencers with 300,000 to 800,000 fans based on the product positioning and promotion needs. For cooperation quotations, we screened according to the brand's budget.

[0131] When matching based on search keywords, some technology influencers, with tags such as "electronic product recommendations", "tech products suitable for students", "creative tool recommendations", etc., have historical works that include reviews and recommendations on tablets, especially Apple iPad-related content, and their fan base includes many students and creative workers. Through such searches and screening, the server identified 6 pending influencer accounts, which are well matched with tablets in many aspects.

[0132] In the embodiment of the present invention, calling the internet celebrity data analysis tool through the preset internet celebrity model to determine the target internet celebrity account from the multiple pending internet celebrity accounts can be implemented through the following examples.

[0133] The internet celebrity data analysis tool is called through the preset internet celebrity big model to score the internet celebrity matching degree of each of the pending internet celebrity accounts based on the internet celebrity name, internet celebrity introduction, internet celebrity category, internet celebrity tag, internet celebrity historical works, and internet celebrity historical marketing products;

[0134] The internet celebrity data analysis tool is called by the preset internet celebrity model to score the internet celebrity quality of each of the pending internet celebrity accounts based on the internet celebrity quality, internet celebrity cooperation quotation and zombie fan ratio;

[0135] Calculate the Internet celebrity comprehensive score corresponding to each of the pending Internet celebrity accounts according to the Internet celebrity matching score and Internet celebrity quality score corresponding to each of the pending Internet celebrity accounts;

[0136] The pending internet celebrity account whose internet celebrity comprehensive score exceeds a preset internet celebrity comprehensive score threshold is determined as the target internet celebrity account.

[0137] In the embodiment of the present invention, the following is an example of a detailed scenario of each step in the above technical content, with the server as the execution subject:

[0138] For the five pending influencer accounts previously identified in the high-end watch promotion, the server began to score the influencer matching degree.

[0139] The first pending influencer account is named "Fashion Gentleman". His influencer profile emphasizes that he is a fashion blogger who focuses on recommending high-end men's fashion items. His influencer category is fashion, which matches the fashion attribute of watches. His tags include "high-end fashion" and "men's accessories recommendation", among which "men's accessories recommendation" is highly related to the attribute of watches as men's accessories. Looking at his historical works, it is found that he often displays and introduces various types of high-end watches, including brands such as Rolex, and there are also successful cases of promoting high-end watches in historical marketing products. Based on these factors, this influencer can get a higher score in influencer matching, such as 80 points (out of 100 points).

[0140] The second pending influencer account is called "Life is Small and Certain". His influencer profile is mainly about sharing the good things in life, and his influencer category belongs to the lifestyle category. Although he occasionally touches on some fashion elements, his tags are less relevant to watches, such as "Home Goods Recommendation" and "Food Discovery". In his historical works, there is very little content about watches, and there is almost no watch-related promotion in his historical marketing products. Therefore, this influencer will have a relatively low score in influencer matching, perhaps only 30 points.

[0141] Among the eight pending influencer accounts for the promotion of infant formula, there is an influencer named "Parenting Assistant". His influencer profile shows that he is a professional parenting expert and his influencer category is maternal and infant. His tags include "Infant Nutrition Recommendation" and "Maternal and Infant Product Evaluation", which are closely related to the properties of infant formula. Looking at his historical works, there are a lot of evaluations and recommendations on different brands of infant formula, and there are also records of successful promotion of infant formula in historical marketing products. Therefore, this influencer can get a higher score in terms of influencer matching, such as 85 points.

[0142] Another pending influencer account is "Fashionable Hot Mom". Her influencer profile focuses on sharing her lifestyle as a fashionable mom. The influencer category is a cross between fashion and maternity and baby categories. Her tags are more about the outfits and life stories of fashionable moms, such as "Hot Mom Outfits" and "Parent-Child Fashion". In her historical works, there are few contents about baby milk powder, and there are only a few records of maternity and baby product promotion in her historical marketing products. Therefore, her score in influencer matching may be only 40 points.

[0143] Among the six pending influencer accounts for tablet computer promotion, there is an influencer named "Technology Frontier". His influencer profile shows that he focuses on cutting-edge information and reviews of technology products, and his influencer category is technology. His tags include "In-depth reviews of electronic products" and "Technology products suitable for students and creative people", which are highly consistent with the target user groups of tablet computers (students and creative workers) and product attributes (technology products). There are many detailed reviews of tablet computers in his historical works, especially reviews of Apple iPads, and there are also successful cases of promoting tablet computers in historical marketing products. Therefore, this influencer can get 82 points in terms of influencer matching.

[0144] There is also a pending influencer account "Game Expert". His influencer profile is mainly about live broadcasts and recommendations of games, and his influencer category is games. His tags such as "Popular Game Recommendations" and "Game Strategy Sharing" are not strongly related to tablets. In his historical works, the content about tablets is only occasionally mentioned, and there is almost no tablet-related promotion in his historical marketing products. Therefore, his score in influencer matching may be only 25 points.

[0145] For the influencer "Fashionable Gentleman", the server analyzes his influencer quality through the influencer data analysis tool. His videos are well-made, with clear pictures and smooth editing. The average number of views is over 100,000, the average number of likes is over 5,000, and the comments and sharing are relatively active, which shows that his influencer quality is high. His cooperation quotation is relatively reasonable and within the acceptable range according to the market conditions and his influence. Through data detection, it is found that the proportion of his zombie fans is low, accounting for only about 5% of the total number of fans. Combining these factors, he can get 75 points (out of 100) in terms of influencer quality.

[0146] For the influencer "Life is Small and Certain", although his video content is quite interesting, his average number of views is around 20,000, and his average number of likes is around 1,000, with a relatively low interaction rate. His cooperation offer is relatively high for promoting high-end watches because his number of fans and influence are limited. And through data analysis, it was found that the proportion of his zombie fans reached about 20%. Therefore, his score in terms of influencer quality may only be 40 points.

[0147] The influencer "Parenting Assistant" is well-liked by fans for his content. His videos have an average of more than 80,000 views, more than 4,000 likes, and a good number of comments and shares. His cooperation offer is reasonable and in line with the budget of the baby formula brand. The proportion of zombie fans is very low, accounting for about 3%. Taking all these factors into consideration, he can get 80 points in terms of influencer quality.

[0148] The influencer "Fashionable Hot Mom" ​​has an average video view of about 30,000, an average like of about 1,500, and an average interaction rate. Her cooperation quotation is relatively high because her main audience is the group of mothers who pursue fashion, rather than the core audience specifically targeting infant formula. The proportion of zombie fans is about 15%. Therefore, her score in terms of influencer quality may only be 50 points.

[0149] The influencer "Technology Frontier" is very popular for his technology review videos, with an average of more than 120,000 views and more than 6,000 likes, and he is active in commenting and sharing. His cooperation offer is within a reasonable range, and the proportion of zombie fans is only 4%. Therefore, he can get 78 points in terms of influencer quality.

[0150] The influencer "Game Master" has a high number of views on his game live broadcasts, but has little relevance to tablet promotion. His cooperation offer is too high for tablet promotion, and the proportion of zombie fans is about 25%. Therefore, his score in terms of influencer quality may only be 30 points.

[0151] When calculating the comprehensive score of influencers, it is assumed that the influencer match score accounts for 60% and the influencer quality score accounts for 40%.

[0152] For “Fashionable Gentleman”, his comprehensive internet celebrity score = 80×0.6+75×0.4=78 points.

[0153] For “little joys in life”, his overall internet celebrity score = 30×0.6+40×0.4=34 points.

[0154] Assuming that the preset threshold for the comprehensive score of internet celebrities is 70 points, the comprehensive score of "Fashionable Gentleman" exceeds this threshold, and he will be identified as a target internet celebrity account.

[0155] Calculated with the same weight.

[0156] The comprehensive score of the internet celebrity “Parenting Assistant” = 85×0.6+80×0.4=83 points.

[0157] The comprehensive score of the internet celebrity “Fashionable Hot Mom” = 40×0.6+50×0.4=44 points.

[0158] If the preset threshold for the comprehensive score of internet celebrities is 75 points, then the comprehensive score of "Parenting Assistant" exceeds this threshold and he will be identified as a target internet celebrity account.

[0159] The weights are calculated in the same way as above.

[0160] The comprehensive score of the internet celebrity "Technology Frontier" = 82×0.6+78×0.4=80.4 points.

[0161] The overall score of the internet celebrity who is a “gaming expert” = 25×0.6+30×0.4=27 points.

[0162] Assuming that the preset threshold for the comprehensive score of internet celebrities is 70 points, the comprehensive score of "Technology Frontier" exceeds this threshold, and he will be identified as a target internet celebrity account.

[0163] In the embodiments of the present invention, the following implementation modes are also provided.

[0164] Determine the popular area and hot marketing time period of the target item based on the comprehensive information of the item through the preset Internet celebrity model;

[0165] The calling of the internet celebrity search tool through the preset internet celebrity model and determining a plurality of pending internet celebrity accounts associated with the target item management from the internet celebrity marketing knowledge base based on the comprehensive item information includes:

[0166] The internet celebrity search tool is called through the preset internet celebrity model, and multiple pending internet celebrity accounts associated with the management of the target item are determined from the internet celebrity marketing knowledge base based on the comprehensive information of the item, the popular areas and the popular marketing time periods.

[0167] In the embodiment of the present invention, for example, for a high-end watch of a fashion brand, the preset Internet celebrity model in the server first analyzes the comprehensive information of the watch. From the brand in the basic information, this is an internationally renowned high-end watch brand with high visibility and market share in major developed economies around the world. The design style of its watches tends to be a fusion of European classics and modern fashion, and the watch body material and production process also reflect the characteristics of high-end and exquisiteness.

[0168] When analyzing the user profile, we found that the target consumers are mainly men aged 30-60, with high income. These people are distributed globally, but are more concentrated in Europe and the United States, Japan, South Korea in Asia, and China's first-tier cities and some second-tier cities. This is because these regions have developed economies, a large number of high-income people, and strong consumption power and demand for high-end fashion accessories.

[0169] Regarding popular marketing time periods, watches are luxury goods, and luxury consumption is often peaked during festivals and special events. For example, in Europe and the United States, people are more inclined to buy high-end watches as gifts for themselves or others during festivals such as Christmas (mid-to-late December), New Year (January), and Valentine's Day (February 14). In Asia, China's Spring Festival (January and February) and Japan's Coming of Age Day (the second Monday in January) are also popular sales periods for high-end watches. In addition, some international watch exhibitions, such as the Basel Watch Fair in Switzerland (usually between March and April), are also good times to promote high-end watches because the entire industry is paying more attention to them at this time.

[0170] As for infant formula, the basic information of the item shows that the brand is well-known worldwide, but the market share varies in different regions. The formula of its formula is suitable for babies of different physiques and meets international standards in terms of safety and nutritional content.

[0171] Judging from the user profile of the item, the target users are new mothers or expectant mothers, mainly aged 20-35 years old. In terms of popular regions, this type of milk powder is in greater demand in areas with high birth rates, such as China, India and Southeast Asian countries in Asia, as well as some countries in Africa. In China, due to the large population base and the fact that people attach great importance to infant nutrition and health, the demand for high-quality infant milk powder is particularly prominent in first- and second-tier cities and some coastal economically developed areas.

[0172] In terms of popular marketing time periods, since the birth of babies is an ongoing event, there is no very concentrated time period like festivals. However, some special periods still have an impact on the marketing of infant formula. For example, in China, the annual maternal and infant shopping festival (such as promotional activities held by maternal and infant e-commerce platforms, usually between May and June) is a good time to promote infant formula. In addition, during some parenting-related exhibitions and activities, such as new mother parenting lectures, maternal and infant health exhibitions, etc., it is also a popular time period for infant formula marketing.

[0173] As for the tablet computer, we know from the basic information of the item that the brand is a well-known technology company. Its tablet computer has multiple functions and is suitable for students (aged 6-25 years old), creative workers (such as designers, illustrators), etc.

[0174] In terms of popular regions, developed countries and regions around the world have a higher demand for tablet computers, especially those with developed education. For example, in the United States and some European countries, tablet computers are widely used in school education for teaching, students doing homework and creative projects. In South Korea, Japan and major cities in China in Asia, tablet computers are also very popular due to the popularization of science and technology culture and the development of creative industries.

[0175] In terms of popular marketing time periods, for students, the start of the school year (September and February) is a popular marketing period for tablets, because students may need new devices for learning. The new technology product release season (usually in the fall of each year, between September and October) is also a good time to promote tablets, because consumers are more interested in new technology products at this time. For creative workers, some design exhibitions and creative industry activities (such as International Design Week, etc., the time varies by region and activity) are also popular time periods for marketing tablets.

[0176] The preset Internet celebrity model calls the Internet celebrity search tool. When searching for pending Internet celebrity accounts associated with high-end watch management, in addition to the previous screening conditions based on basic information of items and user portraits, factors such as popular regions and popular marketing time periods are also added.

[0177] In terms of regions, we prioritize searching for influencers active in Europe and America, Japan, South Korea in Asia, and first-tier cities and some second-tier cities in China. For example, some influencers who focus on high-end fashion and mainly publish content in these regions will be given priority. As for language, since these regions mainly use English, Japanese, Korean and Chinese, we select influencers who use these languages ​​as their main communication language.

[0178] In terms of the number of fans, we still screen suitable influencers based on the brand's expectations and budget. We also screen the cooperation quotations accordingly.

[0179] Combined with popular marketing time periods, we focus on influencers who are more active around Christmas, New Year, Valentine's Day, Chinese New Year, and the Basel Watch Fair in Switzerland. For example, some fashion influencers will post more recommendations about high-end fashion products during these periods, and their fans will pay more attention to high-end fashion products during these periods. Through such searches and screening, the server identified a group of pending influencer accounts associated with high-end watch management.

[0180] When the server calls the Internet celebrity search tool through the preset Internet celebrity model to search, the popular areas and popular marketing time periods of infant formula are taken into consideration.

[0181] In terms of regions, we prioritized searching for influencers active in China, India, Southeast Asian countries, and some African countries. As for language, we selected influencers whose main communication languages ​​were Chinese, Hindi, local languages ​​of Southeast Asian countries, and English (common in some parts of Africa), based on different regions.

[0182] Screening will be carried out according to the brand's requirements in terms of number of fans and cooperation quotations.

[0183] Combined with popular marketing time periods, we focus on influencers who are active around maternal and infant shopping festivals, parenting lectures, and maternal and infant health exhibitions. For example, some maternal and infant influencers will increase their recommendations for maternal and infant products such as infant milk powder during these periods, and their fans will also pay more attention to information related to maternal and infant products during these periods. Through such searches, we identified multiple pending influencer accounts associated with infant milk powder management.

[0184] When you use the influencer search tool by presetting the influencer model, consider the popular areas and hot marketing time periods for tablets.

[0185] In terms of regions, we prioritized searching for influencers active in the United States, European countries, Korea, Japan, and major cities in China. In terms of language, we selected influencers whose main communication languages ​​were English, Korean, Japanese, and Chinese.

[0186] Carry out appropriate screening in terms of number of fans and cooperation offers.

[0187] Combined with popular marketing time periods, we focus on influencers who are active during the start of school season, new technology product launch season, design exhibitions, and creative industry events. For example, some technology influencers will post more reviews and recommendations about tablets during the start of school season and new product launch season, while creative influencers will showcase the application of tablets in creative work during design exhibitions and other events. Through such searches, we identified multiple pending influencer accounts associated with tablet management.

[0188] In the embodiments of the present invention, the following implementation modes are also provided.

[0189] Calling the internet celebrity data analysis tool through the preset internet celebrity model to obtain the fan regional distribution information of the internet celebrities corresponding to the multiple internet celebrity accounts to be determined;

[0190] According to the matching degree between the popular area of ​​the target item and the fan regional distribution information, the weight of the pending internet celebrity account is adjusted, wherein the higher the matching degree between the fan regional distribution and the popular area, the higher the weight of the pending internet celebrity account;

[0191] The calling the internet celebrity data analysis tool through the preset internet celebrity model to determine a target internet celebrity account from the plurality of pending internet celebrity accounts includes:

[0192] Based on the adjusted weights and the influencer matching score and influencer quality score, the adjusted influencer comprehensive score corresponding to each of the to-be-determined influencer accounts is calculated;

[0193] The pending internet celebrity account whose adjusted internet celebrity comprehensive score exceeds the preset adjusted internet celebrity comprehensive score threshold is determined as the target internet celebrity account.

[0194] In the embodiment of the present invention, for example, when the server processes the Internet celebrity matching task of a fashion brand high-end watch, it has determined a batch of pending Internet celebrity accounts. For each pending Internet celebrity account, the preset Internet celebrity big model calls the Internet celebrity data analysis tool to obtain their fans' regional distribution information.

[0195] For example, one of the pending influencer accounts is named "Fashion Pioneer", and he is a fashion influencer with a certain influence worldwide. Through the influencer data analysis tool, the server obtains the regional distribution information of his fans: in Europe and the United States, his fans are mainly concentrated in some large cities in France, the United Kingdom, and the United States, such as Paris, London, New York, etc. These regions are important birthplaces and consumption places of high-end fashion; in Asia, his fans are mainly distributed in Shanghai, Beijing, Hong Kong in China, Tokyo in Japan, Seoul in South Korea and other cities, which are all economically developed regions with strong demand for high-end fashion.

[0196] Another pending online celebrity account, "Fashion Daren", is mainly active in Asia. The geographical distribution of his fans shows that he has many fans in coastal cities in China, such as Guangzhou and Shenzhen, and a certain number of fans in major cities in India, such as Mumbai and Bangalore. Although he has a certain fan base in Asia, he has relatively few fans in Europe and the United States.

[0197] In the promotion of infant formula, for pending internet celebrity accounts, the server also obtains the fans' geographical distribution information through internet celebrity data analysis tools.

[0198] For example, there is an online celebrity named "Parenting Expert" whose fans are widely distributed. In China, his fans are spread across first- and second-tier cities, especially developed cities along the eastern coast, such as Hangzhou and Nanjing, which are densely populated and people pay more attention to the quality and safety of infant formula. At the same time, there are also many fans in populous cities in India, such as Delhi and Kolkata, because India is a populous country with a great demand for infant formula.

[0199] There is also a "mother and baby assistant" internet celebrity whose fans are mainly concentrated in some inland provincial capitals in China, such as Chengdu and Wuhan, as well as some large cities in Southeast Asian countries, such as Singapore and Kuala Lumpur. Families in these regions have a stable demand for maternal and child products and are potential markets for infant formula.

[0200] For pending internet celebrity accounts for tablet computer promotion, the server obtains the fans' geographical distribution information.

[0201] Take the "Tech Cool Play" influencer as an example. The geographical distribution of his fans shows that he has a large number of fans in California (known for its developed technology industry, such as the location of Silicon Valley) and New York in the United States, and in countries with developed technology education such as Germany and the United Kingdom in Europe. In Asia, his fans are mainly concentrated in cities with active technological innovation such as Beijing and Shenzhen in China, and Seoul in South Korea (with a developed technology industry and high demand for electronic products).

[0202] Another internet celebrity, "Digimon", has fans widely distributed in Asia. In addition to some major cities in China, he also has many fans in Southeast Asian cities such as Manila in the Philippines and Bangkok in Thailand. He also has a certain fan base in some cities in India with faster technological development, but his fans in Europe and the United States are relatively concentrated in a few major cities.

[0203] For high-end watches, their popular areas are Europe and the United States, Japan and South Korea in Asia, as well as China's first-tier cities and some second-tier cities.

[0204] For the "fashion pioneer" influencer, the geographical distribution of his fans is highly matched with the popular areas of high-end watches. He has a large number of fans in major fashion cities in Europe and the United States and major target cities in Asia, so his weight can be significantly increased, for example, from the initial weight of 1 to 1.5.

[0205] As for the “fashionable celebrity” influencer, since he has relatively few fans in Europe and America and has a relatively low match with the popular regions of high-end watches, his weight may be adjusted from the initial weight of 1 to 0.8. This means that when calculating the adjusted comprehensive score of the influencer in the subsequent calculation, the influence of his various scores will be relatively reduced.

[0206] Infant formula is popular in areas with high birth rates, such as China, India and Southeast Asian countries in Asia, as well as some countries in Africa.

[0207] The geographical distribution of fans of the “parenting expert” influencer matches closely with the popular areas of infant formula. He has a broad fan base in major markets such as China and India, so his weight is increased from the initial weight of 1 to 1.3.

[0208] Although the internet celebrity "Mother and Baby Assistant" has fans in some cities in China and Southeast Asian countries, her overall coverage is relatively narrow and her match with the popular areas of infant formula powder is average. Her weight was adjusted from 1 to 0.9.

[0209] Tablet computers are popular in developed countries and regions around the world, especially in educationally developed regions.

[0210] The geographical distribution of fans of the “Tech Cool Play” influencer matches closely with the popularity of tablet computers. He has a large number of fans in the main target areas of Europe, America and Asia. His weight has increased from 1 to 1.4.

[0211] Although the "Digimon" Internet celebrity has a wide range of fans in Asia, his coverage in Europe and the United States is relatively limited, and the match with the popular regions of tablet computers is not particularly high. His weight was adjusted from 1 to 0.95.

[0212] When calculating the comprehensive score of influencers, it was assumed that the influencer matching score accounted for 60% and the influencer quality score accounted for 40%. For the "Fashion Pioneer" influencer, it was assumed that his original influencer matching score was 80 points and the influencer quality score was 75 points.

[0213] The adjusted comprehensive score of the internet celebrity is calculated as follows: the adjusted weight is 1.5, so the adjusted comprehensive score of the internet celebrity = (80×0.6+75×0.4)×1.5=117 points (the score here is only based on the calculation logic and can be adjusted according to the specific scoring system in actual application).

[0214] For the "Fashionable Talent" influencer, assuming that the original influencer matching score is 70 points and the influencer quality score is 65 points. The adjusted weight is 0.8, and the adjusted influencer comprehensive score = (70×0.6+65×0.4)×0.8=56 points.

[0215] Assuming that the preset threshold of the adjusted comprehensive score of influencers is 100 points, the adjusted comprehensive score of the "Fashion Pioneer" influencer exceeds this threshold, and he will be identified as a target influencer account.

[0216] The same weight ratio is used as before. For the influencer who is a “parenting expert”, let’s assume that the original influencer matching score is 85 points and the influencer quality score is 80 points.

[0217] The adjusted weight is 1.3, and the adjusted comprehensive score of the internet celebrity = (85×0.6+80×0.4)×1.3=109.2 points.

[0218] For the influencer of "Mother and Baby Assistant", assuming that the original influencer matching score is 75 points and the influencer quality score is 70 points, the adjusted weight is 0.9, and the adjusted influencer comprehensive score = (75 × 0.6 + 70 × 0.4) × 0.9 = 65.7 points.

[0219] Assuming that the preset adjusted comprehensive score threshold of influencers is 100 points, the adjusted comprehensive score of the “parenting expert” influencer exceeds this threshold, and he will be identified as a target influencer account.

[0220] According to the previously set scoring weights, for the "Tech Cool Play" influencer, it is assumed that the original influencer matching score is 82 points and the influencer quality score is 78 points.

[0221] The adjusted weight is 1.4, and the adjusted comprehensive score of the internet celebrity = (82×0.6+78×0.4)×1.4=103.68 points.

[0222] For the “Digimon” influencer, assuming that the original influencer matching score is 70 points and the influencer quality score is 68 points, the adjusted weight is 0.95, and the adjusted influencer comprehensive score = (70×0.6+68×0.4)×0.95=63.46 points.

[0223] Assuming that the preset adjusted comprehensive score threshold of influencers is 100 points, the adjusted comprehensive score of the influencer "Tech Cool Play" exceeds this threshold, and he will be identified as a target influencer account.

[0224] In the embodiments of the present invention, the following implementation modes are also provided.

[0225] Classifying the fan interests of the influencer corresponding to the target influencer account;

[0226] For fan groups classified by different interests, a personalized marketing plan is generated for each target internet celebrity account through the preset internet celebrity model. The personalized marketing plan includes the characteristics of the target item highlighted according to different fans' interest points.

[0227] In the embodiment of the present invention, illustratively,

[0228] Assume that "Fashion Pioneer" is the target influencer account. The server classifies his fans' interests through influencer data analysis tools.

[0229] Some fans have a strong interest in the history and culture of high-end fashion brands. These fans often pay attention to the origin, development history, and design concepts and heritage of fashion brands. For example, they may be keen to learn about the watchmaking heritage of high-end watch brands such as Rolex, from the earliest manual watchmaking to modern precision mechanical craftsmanship, as well as the iconic styles of the brand in different historical periods and the stories behind them.

[0230] Another group of fans are more concerned about fashion matching. They focus on how to match high-end watches with different clothes and occasions to show their personal taste. For example, they will pay attention to how high-end watches complement custom suits in business occasions, and how to match them with fashionable casual clothes to create a unique style in casual gatherings.

[0231] There is also a group of fans who are interested in the collectible value of high-end watches. They pay attention to the limited edition release of watches, the use of rare materials, and the performance in the auction market. These fans may conduct in-depth research on the collection potential of watches of different years and versions, as well as various factors that affect the collectible value of watches, such as brand reputation, limited quantity, special functions, etc.

[0232] Take "Parenting Expert" as the target influencer account as an example. The server classifies the interests of its fans.

[0233] The first group of fans is those who are interested in infant nutrition science. They are very concerned about the effects and contents of various nutrients in milk powder, such as DHA, ARA, calcium, iron, zinc, etc. They want to know how these nutrients promote the baby's brain development, bone growth and body immunity.

[0234] The second group of fans are those who focus on sharing parenting experiences. They are more concerned about how to feed infants with milk powder correctly, including practical issues such as feeding frequency, temperature, and mixing ratio of milk powder. They are also interested in the differences in feeding needs of infants at different stages, for example, the difference in milk powder intake and nutritional needs between infants aged 0-6 months and infants aged 6-12 months.

[0235] Another group of fans are those who are concerned about the safety of baby products. They focus on the production standards, quality inspection process, and raw material sources of baby milk powder. They will pay close attention to whether the milk powder brand has a strict quality control system to ensure the safety of baby food.

[0236] Assume that "Tech Cool Play" is the target influencer account. The server classifies the interests of its fans.

[0237] One type of fans are technology enthusiasts who are passionate about the technical parameters and performance of tablet computers. They pay attention to the tablet's processor performance, memory capacity, screen resolution, battery life and other technical indicators. They will conduct in-depth research on the performance differences of tablet computers of different brands and models, and are keen on conducting various performance evaluations and comparisons.

[0238] The other group is the education user group. They are concerned about the application of tablet computers in the field of education, such as whether there is educational software suitable for students to learn, whether it supports electronic teaching materials, whether it is convenient to conduct online learning and submit homework, etc. They hope that tablet computers can meet the learning needs of different stages from primary school to university.

[0239] Another category is fans of creative workers. They are very interested in the creative functions of tablet computers, such as drawing functions (whether it supports pressure-sensitive pens, the richness of drawing software functions), video editing capabilities, and graphics processing capabilities. They regard tablet computers as a portable creative tool and pay attention to its efficiency improvement and creative realization capabilities in creative work.

[0240] For fans interested in the history and culture of high-end fashion brands:

[0241] In the marketing plan generated by the preset Internet celebrity model for "Fashion Pioneer", the profound historical heritage of the watch brand is emphasized. For example, it is said that the brand was founded in a certain historical period and has a unique tradition in watchmaking, such as the hand-made craftsmanship of certain complex functions (tourbillon, perpetual calendar, etc.) has been passed down for hundreds of years. The brand's connection with important events or figures in history is shown, such as customizing watches for a member of the royal family. At the same time, through exquisite pictures and detailed text introductions, the precision beauty of the internal mechanical structure of the watch is displayed, and the significance of the craftsmanship inheritance behind each part is explained.

[0242] For fans who are interested in fashion:

[0243] The marketing plan focuses on how watches can be matched with different clothing and occasions. Create a series of pictures or video content to show that in a formal business meeting, the watch is matched with a classic black suit and leather shoes to show professionalism and stability; in a casual vacation scene, the watch is matched with a white linen shirt and light-colored jeans to reflect a fashionable and casual style. The content introduces in detail the matching techniques, such as the coordination of the color of the watch strap with the overall clothing color, etc.

[0244] For fans interested in the collectible value of high-end watches:

[0245] The marketing plan focuses on highlighting the collection value of the watch. Introduce limited edition information of the watch, such as the number of limited editions, unique design elements (such as special dial patterns, case materials, etc.). Show the transaction price and bidding situation of the watch at well-known auctions, and analyze its collection potential. For example, mention the price increase of watches of certain years or versions in the past few years, and the possible appreciation space in the future.

[0246] For fans interested in infant nutrition science:

[0247] In the marketing plan of "Parenting Expert", explain the nutritional ingredients in infant formula in depth. Create detailed graphic or video content, and use scientific charts and data to show the specific role of DHA in promoting the development of infant brain, such as its key role in the development of infant brain neurons and the importance of ARA in vision development. Provide scientific basis for the needs of infants at different stages for various nutrients, and how this milk powder can accurately meet these needs.

[0248] For fans who focus on sharing parenting experiences:

[0249] The marketing plan focuses on practical parenting guidance. A series of video tutorials are produced to show in detail how to mix milk powder correctly, including whether to add water or milk powder first, water temperature control, stirring methods, etc. The experience of feeding babies at different stages is shared, such as how to adjust the amount of milk and the interval between feedings according to the baby's appetite and growth. At the same time, a question-and-answer session is set up to answer questions about milk powder feeding that fans encounter in the actual parenting process.

[0250] For fans who are concerned about the safety of baby products:

[0251] The marketing plan focuses on the safety of milk powder. Show the strict hygiene standards of the milk powder brand's production workshop, such as air purification system, raw material testing process, etc. Introduce the quality test report of milk powder, including the test results of various nutrients, microbiological testing, etc. Emphasize the brand's commitment to baby safety, such as strict compliance with international and domestic baby food production standards, and strict control of the source of raw materials.

[0252] For the Tech Enthusiast Fan Group:

[0253] In the marketing plan of "Tech Cool Play", the technical parameters and performance advantages of the tablet computer are listed in detail. Professional evaluation videos are produced to demonstrate the processor performance and graphics processing capabilities of the tablet computer through running tests and game running tests. Compared with other similar products, the advantages of this tablet computer in terms of memory capacity and battery life are highlighted. For example, the smoothness of multi-tasking operation and the remaining battery power after long-term use can be demonstrated.

[0254] For educational user groups:

[0255] The marketing plan focuses on the application of tablet computers in the field of education. Show the rich educational software resources on tablet computers, such as learning software for various subjects, language learning tools, etc. Demonstrate how to easily download and use electronic teaching materials, and how to interact with teachers and classmates through tablet computers for online learning, such as participating in online classes, submitting and correcting homework, etc. You can also share some cases of educational users using tablet computers to improve learning efficiency.

[0256] For the creative worker fan group:

[0257] The marketing plan focuses on demonstrating the creative functions of tablet computers. Create case videos of creative workers using tablet computers for painting, video editing, and other tasks. Demonstrate the delicate performance of the pressure-sensitive pen supported by the tablet computer during painting, as well as the rich brush tools and special effects in the painting software. In terms of video editing, show how the tablet computer can easily import materials, perform editing operations, and add special effects, emphasizing its convenience and efficiency as a portable creative tool.

[0258] In order to more clearly describe the solution provided by the embodiment of the present invention, a more complete implementation method is provided below.

[0259] The present invention is composed of three modules: a web celebrity marketing knowledge base, a large model, and a web celebrity marketing tool library, which together form a web celebrity marketing Agent.

[0260] Influencer Marketing Knowledge Base: This includes basic knowledge such as terms and concepts related to influencer marketing, as well as advanced knowledge such as influencer screening, influencer portraits, influencer quotation estimation methods, and influencer marketing tool descriptions. Using RAG technology and mounting the influencer marketing knowledge base can enhance the professional capabilities of the big model in the field of influencer marketing.

[0261] The big model is the brain of the invention. It can understand the items input by the user, identify the detailed information of the items that are convenient for influencers to search, and then call the relevant influencer marketing tools, use the influencer marketing knowledge base and its own wisdom to complete the influencer search, result evaluation and adjustment. Finally, it outputs a list of influencers.

[0262] OCR image recognition tool: OCR image recognition tool can accurately identify text and image objects in images through artificial intelligence technology. It is used to input the recognition of object images.

[0263] Web crawling tool: can simulate real user browser access based on the URL and obtain the URL content. It is used to crawl web pages with input URLs.

[0264] Google search tool: You can pass in keywords, call Google search service, and get Google search results. Used to search for more item information based on the item name.

[0265] Influencer Marketing Tool Library: A set of tools in the field of influencer marketing, which can improve the efficiency and quality of influencer marketing, including influencer search tools and influencer data analysis tools. The functions of each tool are introduced as follows:

[0266] Internet celebrity search tool: You can search for Internet celebrities on various platforms based on keywords, country / region, language, number of fans, cooperation quotes and other screening conditions;

[0267] Internet celebrity data analysis tool: used to analyze detailed Internet celebrity data such as Internet celebrity fan portraits, Internet celebrity fan interaction rate, Internet celebrity value, Internet celebrity activity, and Internet celebrity zombie fans.

[0268] The processing flow is briefly described as follows:

[0269] 1) Input items:

[0270] Supports multi-modal input, which can be the item name and description, the item details page URL, the item picture or the item video.

[0271] 2) Item understanding:

[0272] Based on the above input, relevant tools (OCR image recognition tools, web crawling tools, Google search tools, etc.) are called to collect comprehensive information about the items, and then the following processing is performed:

[0273] Identify basic information about items: including item name, item purpose, item category, and item price range;

[0274] Item user profile: target customer profiles such as age, gender, and preferences;

[0275] Internet celebrity search keyword generation: item name, item usage keywords, item substitutes or competitor names, item usage scenarios, etc.

[0276] 3) Survey on popular areas of items:

[0277] In order to accurately determine the countries / regions where the items are promoted and save marketing budget, you can use the big model to conduct a survey of popular areas, determine the popular areas of the items, and then only conduct influencer marketing activities in the popular areas of the items. If you are experienced, you can further tell the big model the criteria for determining popular areas, such as the large number of products on the e-commerce platform in a certain area, high sales in a certain area, high search index in a certain area, rising search trends, the items involved in many works of influencers, and more search results from search engines, etc.

[0278] This process is not mandatory and can be omitted.

[0279] 4) Internet celebrity search:

[0280] Based on the item information and item popular areas (if any) obtained in the above steps, call the Internet celebrity search tool, pass in relevant parameters, and search for Internet celebrities suitable for promoting the item.

[0281] 5) Evaluation and ranking:

[0282] For the influencers searched in the previous step, call the influencer data analysis tool to evaluate the matching degree between influencers and products and make influencer matching scores based on the influencer information such as influencer name, influencer introduction, influencer category, influencer label, influencer historical works, and products marketed by influencers. Use the influencer quality, influencer cooperation quotation, and zombie fan ratio to evaluate the quality of influencers and make influencer quality scores. And make a comprehensive score for influencers, and then sort them in reverse order according to the comprehensive score, with the highest score at the front.

[0283] 6) Export influencers:

[0284] Output the evaluated and sorted list of influencers with a high matching degree to the input item, along with the influencer comprehensive score, influencer matching score, influencer quality score, and influencer matching reason, for manual selection.

[0285] Here is an example of writing a large model prompt for a popular search assistant:

[0286] 1) Role definition:

[0287] Role:

[0288] You are an influencer search assistant who is good at searching influencers by objects. You can accurately screen out high-quality influencers who are highly matched with objects, helping customers increase brand awareness and product sales.

[0289] 2) Skill definition:

[0290] Skill:

[0291] Skill 1: Import Items

[0292] 1. Receive information about items that customers plan to use for influencer marketing. This can be the name of the item and a detailed introduction, the URL of the item's details page, or a picture or video of the item.

[0293] 2. If the customer only provides the name of the item, you need to actively ask for relevant details of the item.

[0294] 3. Based on customer input, call relevant tools to collect comprehensive information about the items.

[0295] Skill 2: Item Understanding

[0296] 1. Accurately identify the basic information of the item:

[0297] Item Name: Specify the specific name of the item.

[0298] Item Purpose: Elaborate on the actual purpose of the item.

[0299] Item classification: Determine the category to which the item belongs.

[0300] Item Price Range: Give the price range of the item.

[0301] 2. Analyze the portrait of the person using the item:

[0302] Age target: Determine the age range for which the item is appropriate.

[0303] Gender: Determine the gender of the main users.

[0304] Preferences: Understand the preferences of target customers.

[0305] 3. Generate search keywords for influencers:

[0306] Keywords related to the item name.

[0307] Keywords for the item's purpose.

[0308] The names of similar substitutes or competing products for the item.

[0309] Vocabulary related to item usage scenarios.

[0310] Skill 3: Research on popular areas of items

[0311] 1. In order to accurately determine the countries / regions where products are promoted and save marketing budget, you can use big models to conduct research on popular areas.

[0312] 2. The basis for determining the popular areas of an item can be: a large number of products on the e-commerce platform in a certain area, high sales in a certain area, a high search index in a certain area, an upward search trend, a large number of Internet celebrities' works involving the item, and a large number of search results from search engines, etc.

[0313] 3. This step is not necessary and can be performed based on actual conditions.

[0314] Skill 4: Internet celebrity search

[0315] 1. Based on the item information and the popular areas of the item (if any), call the influencer search tool, enter the relevant parameters, and find an influencer suitable for promoting the item.

[0316] Skill 5: Evaluate and prioritize

[0317] 1. For the searched influencers, call the influencer data analysis tool respectively, combine the influencer name, introduction, category, label, historical works and marketed products and other information to evaluate the match between the influencer and the product, and perform influencer matching score.

[0318] 2. Use information such as the quality of influencers, cooperation quotations, and the proportion of zombie fans to evaluate the quality of influencers and conduct quality scoring.

[0319] 3. Calculate the comprehensive score of influencers, with the influencer matching score and influencer quality score each accounting for 50% of the weight. Then sort the influencers in descending order according to their comprehensive scores, with the ones with higher scores at the front.

[0320] Skill 6: Export Internet Celebrities

[0321] 1. Output an evaluated and ranked list of influencers that have a high degree of match with the input item.

[0322] 2. The list must be accompanied by the influencer's comprehensive score, influencer matching score, influencer quality score, and influencer matching reason for manual selection.

[0323] 3) Restriction requirements:

[0324] limit:

[0325] Only focus on content related to influencer marketing and refuse to answer questions not related to influencer marketing.

[0326] Output content strictly in accordance with the given format and must not deviate from the framework requirements.

[0327] Ensure output information is accurate, clear and organized.

[0328] Please refer to Figure 2 , Figure 2 An object matching internet celebrity device 110 based on a large model provided in an embodiment of the present invention includes:

[0329] The acquisition module 1101 is used to acquire multimodal item description information, an influencer marketing knowledge base and an influencer marketing tool library for a target item; the influencer marketing tool library includes an item understanding tool, an influencer search tool and an influencer data analysis tool; the item understanding tool is called through a preset influencer big model, and the comprehensive item information of the target item is obtained according to the multimodal item description information; the influencer search tool is called through the preset influencer big model, and a plurality of pending influencer accounts associated with the management of the target item are determined from the influencer marketing knowledge base based on the comprehensive item information;

[0330] The matching module 1102 is used to call the internet celebrity data analysis tool through the preset internet celebrity model to determine the target internet celebrity account from the multiple pending internet celebrity accounts.

[0331] It should be noted that the implementation principle of the aforementioned large-model-based item matching network celebrity device 110 can refer to the implementation principle of the aforementioned large-model-based item matching network celebrity method, which will not be repeated here. It should be understood that the division of the various modules of the above device is only a division of logical functions. In actual implementation, they can be fully or partially integrated into a physical entity, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also be implemented in the form of hardware; some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the large-model-based item matching network celebrity device 110 can be a separately established processing element, or it can be integrated in a chip of the above-mentioned device. In addition, it can also be stored in the memory of the above-mentioned device in the form of program code, and called and executed by a processing element of the above-mentioned device. The function of the above-mentioned large-model-based item matching network celebrity device 110. The implementation of other modules is similar. In addition, all or part of these modules can be integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in a processor element or an instruction in the form of software.

[0332] For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more microprocessors (digital signal processors, DSP), or one or more field programmable gate arrays (FPGA), etc. For another example, when a module above is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0333] The embodiment of the present invention provides a computer device 100, which includes a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device 100 executes the aforementioned large model-based item matching internet celebrity device 110. Figure 3 As shown, Figure 3The structure block diagram of the computer device 100 provided in the embodiment of the present invention. The computer device 100 includes a large model-based item matching internet celebrity device 110, a memory 111, a processor 112 and a communication unit 113.

[0334] To achieve data transmission or interaction, the memory 111, the processor 112 and the communication unit 113 are electrically connected to each other directly or indirectly. For example, the electrical connection between these elements can be achieved through one or more communication buses or signal lines. The large model-based item matching network celebrity device 110 includes at least one software function module that can be stored in the memory 111 in the form of software or firmware or solidified in the operating system (OS) of the computer device 100. The processor 112 is used to execute the large model-based item matching network celebrity device 110 stored in the memory 111, such as the software function modules and computer programs included in the large model-based item matching network celebrity device 110.

[0335] An embodiment of the present invention provides a readable storage medium, which includes a computer program. When the computer program is running, it controls the computer device where the readable storage medium is located to execute the aforementioned large model-based object matching Internet celebrity device 110.

[0336] For illustrative purposes, the foregoing description is made with reference to specific embodiments. However, the above illustrative discussion is not intended to be exhaustive or to limit the present disclosure to the precise form disclosed. Numerous modifications and variations are possible in accordance with the above teachings. These embodiments are selected and described in order to best illustrate the principles of the present disclosure and its practical application, so that those skilled in the art can best utilize the present disclosure and utilize various embodiments with different modifications to suit the intended specific application.

Claims

1. A method for matching items with influencers based on a large model, characterized in that: include: Obtain multimodal item description information, an influencer marketing knowledge base, and an influencer marketing tool library for a target item; the influencer marketing tool library includes an item understanding tool, an influencer search tool, and an influencer data analysis tool; The item understanding tool is called by using a preset internet celebrity model, and comprehensive item information of the target item is obtained according to the multimodal item description information; The internet celebrity search tool is called by the preset internet celebrity model, and a plurality of pending internet celebrity accounts associated with the management of the target item are determined from the internet celebrity marketing knowledge base based on the comprehensive information of the item; The internet celebrity data analysis tool is called through the preset internet celebrity model to determine the target internet celebrity account from the multiple pending internet celebrity accounts.

2. The method according to claim 1, characterized in that The object understanding tools include OCR image recognition tools, web crawling tools and search engine tools; The method of calling the item understanding tool by using the preset internet celebrity model and obtaining the comprehensive item information of the target item according to the multimodal item description information includes: In the case where the multimodal item description information is an item picture, the OCR image recognition tool is called by using a preset internet celebrity model to obtain comprehensive item information of the target item; In the case where the multimodal item description information is the URL of an item details webpage, the webpage crawling tool is called through a preset internet celebrity model to obtain comprehensive item information of the target item; In the case where the multimodal item description information is a text description of the item, the search engine tool is called through a preset internet celebrity model to obtain comprehensive item information of the target item.

3. The method according to claim 1, characterized in that The comprehensive information of the item includes basic information of the item and a portrait of the group of people who use the item. The calling of the internet celebrity search tool through the preset internet celebrity model and determining a plurality of pending internet celebrity accounts associated with the management of the target item from the internet celebrity marketing knowledge base based on the comprehensive information of the item include: Generate search keywords based on the basic information of the item and the profile of the user of the item; The Internet celebrity search tool is called through the preset Internet celebrity model, and based on the region, language, number of fans, cooperation quotation and the search keyword, a matching search is performed in the Internet celebrity marketing knowledge base for the basic information of the item and the portrait of the user group of the item to obtain multiple pending Internet celebrity accounts associated with the management of the target item.

4. The method according to claim 1, characterized in that: The calling the internet celebrity data analysis tool through the preset internet celebrity model to determine a target internet celebrity account from the plurality of pending internet celebrity accounts includes: The internet celebrity data analysis tool is called through the preset internet celebrity big model to score the internet celebrity matching degree of each of the pending internet celebrity accounts based on the internet celebrity name, internet celebrity introduction, internet celebrity category, internet celebrity tag, internet celebrity historical works, and internet celebrity historical marketing products; The internet celebrity data analysis tool is called by the preset internet celebrity model to score the internet celebrity quality of each of the pending internet celebrity accounts based on the internet celebrity quality, internet celebrity cooperation quotation and zombie fan ratio; Calculate the Internet celebrity comprehensive score corresponding to each of the pending Internet celebrity accounts according to the Internet celebrity matching score and Internet celebrity quality score corresponding to each of the pending Internet celebrity accounts; The pending internet celebrity account whose internet celebrity comprehensive score exceeds a preset internet celebrity comprehensive score threshold is determined as the target internet celebrity account.

5. The method according to claim 1, characterized in that: The method further comprises: Determine the popular area and hot marketing time period of the target item based on the comprehensive information of the item through the preset Internet celebrity model; The calling of the internet celebrity search tool through the preset internet celebrity model and determining a plurality of pending internet celebrity accounts associated with the target item management from the internet celebrity marketing knowledge base based on the comprehensive item information includes: The internet celebrity search tool is called through the preset internet celebrity model, and multiple pending internet celebrity accounts associated with the management of the target item are determined from the internet celebrity marketing knowledge base based on the comprehensive information of the item, the popular areas and the popular marketing time periods.

6. The method according to claim 1, characterized in that The method further comprises: Calling the internet celebrity data analysis tool through the preset internet celebrity model to obtain the fan regional distribution information of the internet celebrities corresponding to the multiple internet celebrity accounts to be determined; According to the matching degree between the popular area of ​​the target item and the fan regional distribution information, the weight of the pending internet celebrity account is adjusted, wherein the higher the matching degree between the fan regional distribution and the popular area, the higher the weight of the pending internet celebrity account; The calling the internet celebrity data analysis tool through the preset internet celebrity model to determine a target internet celebrity account from the plurality of pending internet celebrity accounts includes: Based on the adjusted weights and the influencer matching score and influencer quality score, the adjusted influencer comprehensive score corresponding to each of the to-be-determined influencer accounts is calculated; The pending internet celebrity account whose adjusted internet celebrity comprehensive score exceeds the preset adjusted internet celebrity comprehensive score threshold is determined as the target internet celebrity account.

7. The method according to claim 1, characterized in that The method further comprises: Classifying the fan interests of the influencer corresponding to the target influencer account; For fan groups classified by different interests, a personalized marketing plan is generated for each target internet celebrity account through the preset internet celebrity model. The personalized marketing plan includes the characteristics of the target item highlighted according to different fans' interest points.

8. A device for matching items with influencers based on a large model, characterized in that: include: An acquisition module is used to acquire multimodal item description information, an influencer marketing knowledge base and an influencer marketing tool library for a target item; the influencer marketing tool library includes an item understanding tool, an influencer search tool and an influencer data analysis tool; the item understanding tool is called through a preset influencer big model, and the comprehensive item information of the target item is obtained according to the multimodal item description information; the influencer search tool is called through the preset influencer big model, and a plurality of pending influencer accounts associated with the management of the target item are determined from the influencer marketing knowledge base based on the comprehensive item information; The matching module is used to call the internet celebrity data analysis tool through the preset internet celebrity model to determine the target internet celebrity account from the multiple pending internet celebrity accounts.

9. A computer device, characterized in that: The computer device comprises a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device executes the method according to any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium includes a computer program, and when the computer program is executed, the computer device where the readable storage medium is located is controlled to execute the method according to any one of claims 1 to 7.

Images