Marketing crime generation method and system based on marketing brief identification and analysis engineering

Abstract

The invention discloses a marketing document generation method and system based on marketing brief identification and analysis engineering, and relates to the technical field of artificial intelligence (GenAI) and natural language processing (NLP), the marketing document generation method and system based on marketing brief identification and analysis engineering accurately analyzes user demands, automatically identifies and analyzes brief uploaded by a user, generates structured JSON output, and improves the user experience. The accuracy and correlation of copywriting generation are improved; through oral and scene processing, the generated copywriting better fits an actual application scene, and the user experience is improved; and efficient copywriting generation: high-quality marketing copywriting is generated by using a large language model and an automatic cue word project, so that the copywriting generation efficiency and effect are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence (GenAI) and natural language processing (NLP), and in particular to a marketing copy generation method and system based on marketing brief recognition and parsing engineering. Background Art

[0002] With the development of artificial intelligence technology, large language models (such as GPT-4) have performed well in generating content. However, in the generation of marketing copy, existing technologies still have significant limitations, especially in the understanding of marketing needs.

[0003] A marketing brief, also known as a marketing activity brief or work requirements, is a medium for brands or Party A to convey information to service providers and creative producers (such as advertising companies, marketing agencies, KOLs, etc.) during marketing activities. It is a document that details the requirements, goals, strategies, execution details and expected results of a marketing project, including project overview, user portraits, marketing strategies, execution details, precautions, etc. Currently, information extraction in marketing briefs mainly relies on manual work, and the process of analyzing product elements is cumbersome and inefficient. In addition, although a large language model can be used to extract brief content, the product content extracted from the brief by this method usually lacks colloquial and scenario-based processing. Most of the copywriting generated using these unprocessed product materials is difficult to meet actual marketing needs, such as being too empty, stiff, and advertising-like.

[0004] Traditional content generation methods are not only time-consuming and laborious, but also the quality of the generated copy is uneven and cannot effectively meet the diverse needs of different marketing scenarios. In order to solve this problem, a method is needed that can automatically parse brief content and process it in a colloquial and scenario-based manner to improve the efficiency and quality of copy generation. Summary of the invention

[0005] The purpose of the present invention is to provide a marketing copy generation method and system based on marketing brief recognition and analysis engineering, which improves the generation quality and efficiency of marketing copy by introducing colloquial and scenario-based processing into the brief information extraction method, meets the needs of users in different scenarios, extracts text from the brief uploaded by the user, and uses a large model to identify and classify the extracted text with preset parameters, including general parameters and specific industry segmentation parameters; after the above brief recognition is completed, the extracted product introduction parameters and product selling point parameters are parsed point by point using the large model, and this step divides the information contained in the two parameters into multiple point-by-point analysis items; automatically selects relevant content after brief processing according to user needs, and constructs prompt words (prompt) to generate the final copy.

[0006] In order to achieve the above-mentioned purpose, the present invention provides the following technical solutions: a method and system for generating marketing copy based on marketing brief recognition and analysis engineering, including that the marketing brief is a key document for guiding marketing activities, which summarizes the core elements of the project, including project background, goals, target audience and core information; by using a large language model, the system can automatically recognize the marketing brief and extract key product marketing parameters from it;

[0007] After the brief recognition is completed, the extracted product introduction parameters and product selling point parameters are analyzed point by point using the big model. This step divides the information contained in the two parameters into multiple point analysis items.

[0008] The automatic prompt word project automatically selects relevant content after brief processing according to user needs, constructs prompt words and generates the final copy.

[0009] Furthermore, the system uses a preset prompt word template for analyzing content relevance and utilizes a large language model to perform content relevance analysis on user needs and brief content, thereby realizing the function of automatically selecting content related to user needs from the parsed and processed brief information, accurately matching user needs with key information in the brief, and ensuring that the generated copy is highly relevant and more authentic.

[0010] Furthermore, the user demand requires the introduction of brand information in the copy. The system automatically completes the prompt word template as input, calls the large language model to obtain the content corresponding to the "brand introduction" parameter as material, and prepares for the subsequent copy generation.

[0011] Furthermore, after obtaining the material, the system constructs a final prompt word in combination with a dynamic prompt word template, takes the final prompt word as input, and calls the large language model to generate the final marketing copy.

[0012] Furthermore, the text in the brief uploaded by the user is extracted;

[0013] Word file: directly extract the text content in the file;

[0014] There are two methods for processing PDF files: pure image extraction method and text + image extraction method;

[0015] Pure image extraction method: extract each page in the PDF as an image, and then call the large model to extract the text in the image;

[0016] Text + image extraction method: extract the text in the PDF page by page. When a page of text is extracted, extract the image of that page. Then call the large model to extract the text in the image. Combine the two texts to get the text of the current page.

[0017] The large model is used to identify and classify the extracted text according to preset parameters, including general parameters and specific industry segmentation parameters.

[0018] Compared with the prior art, the marketing copy generation method and system based on marketing brief recognition and analysis engineering provided by the present invention have the following beneficial effects:

[0019] The marketing copy generation method and system of the marketing brief recognition and parsing project can accurately analyze user needs: automatically identify and parse the brief uploaded by the user, generate structured JSON output, and improve the accuracy and relevance of copy generation;

[0020] Improve the colloquial and scenario-based effects of copywriting: Through colloquial and scenario-based processing, the generated copywriting is more in line with the actual application scenario, improving the user experience;

[0021] Efficient copywriting generation: Using large language models and automatic prompt word engineering to generate high-quality marketing copywriting, the efficiency and effectiveness of copywriting generation are improved;

[0022] Technological innovation: Combining brief recognition, processing and automatic prompt word engineering, it provides a new marketing copy generation method and promotes the development of intelligent marketing copy generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0024] Figure 1 It is a schematic diagram of the system structure of the present invention;

[0025] Figure 2 It is a schematic diagram of the module structure of the brief recognition function of the present invention;

[0026] Figure 3 A schematic diagram of the complete process of the present invention from brief recognition, spoken language processing, scenario processing to copywriting generation by a large language model;

[0027] Figure 4 The figure is a schematic diagram of a specific process of brief parameter identification of the present invention. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0029] See also Figure 1-4 , a marketing copy generation method and system based on marketing brief recognition and analysis engineering, including the following functions and modules:

[0030] 1. Brief recognition

[0031] The marketing brief is a key document used to guide marketing activities. It outlines the core elements of the project, including project background, goals, target audience, and core information. By using a large language model, the system can automatically identify the marketing brief and extract key product marketing parameters from it. This step includes:

[0032] 1. Extract the text from the brief uploaded by the user (using Python libraries, including pillow, pdfplumber and pdf2image)

[0033] a) Word file: directly extract the text content in the file

[0034] b) The methods for processing PDF files include pure image extraction method and text + image extraction method;

[0035] i) Pure image extraction method: extract each page in the PDF as an image, and then call the large model to extract the text in the image

[0036] ii) Text + image extraction method: extract the text in the PDF page by page. When a page of text is extracted, extract the image of that page. Then call the large model to extract the text in the image, and combine the two pieces of text to get the text of the current page.

[0037] 2. Use the big model to identify and classify the extracted text based on preset parameters, including general parameters and specific industry segmentation parameters.

[0038] General industry parameters: divide the general parameters of each industry, including but not limited to brand name, target population, spokesperson endorsement, core selling points, brand introduction, horizontal comparison, etc.

[0039] Specific industry segmentation parameters: more detailed parameter division for different industries. For example:

[0040] Beauty and skin care: applicable skin types, clinical trial data, ingredient analysis, skin care steps, etc.

[0041] Household appliance industry: technical parameters, energy efficiency rating, etc.

[0042] Food industry: nutritional information, taste description, food safety certification, etc.

[0043] The big model is called to identify the brief text extracted in 1 (including industry identification and identification of specific industry segmentation parameters), and the content corresponding to different preset parameters is extracted from the text as the output of the big model. The output is structured to obtain brief parameter information in JSON format so that it can be used by subsequent modules. For example, the brief text input into the big model contains the content "... multiple massage modes, intelligent adjustment of strength..." After recognition, the big model classifies this part of the text into the preset parameter "product selling point".

[0044] 2. Brief analysis

[0045] After the above brief recognition is completed, the extracted product introduction parameters and product selling point parameters are analyzed point by point using the big model. This step divides the information contained in the two parameters into multiple point analysis items.

[0046] example:

[0047] Product introduction: Invisible massage chair, with global sales of more than 80,000+, foldable and easy to store without taking up space...

[0048] Point analysis: 1. Invisible massage chair; 2. Global sales exceed 80,000+; 3. Foldable and easy to store without taking up space...

[0049] After obtaining multiple sub-point analysis items, you can use the colloquial / scenario-based functions for these analysis items, and use the big model to create multiple sets of colloquial / scenario-based results for each sub-point analysis item for users to choose freely. Colloquial processing makes the generated copy closer to the user's expression habits, increasing naturalness and affinity; scenario-based processing enables the copy to adapt to specific application scenarios, improving its practicality and appeal.

[0050] example:

[0051] Point analysis item 1: Foldable and easy to store without taking up space

[0052] Colloquial / scenario-based results:

[0053] 1. The space at home is limited, but this little thing doesn’t take up any space and can help me relax my body. It’s really convenient.

[0054] 2. I am busy in a crowded environment every day, and this small helper that does not take up space allows me to find a moment of relaxation anytime and anywhere.

[0055] 3. On days when there is a lack of outdoor activities, this small artifact that does not take up much space allows me to relax at home.

[0056] 4. When you are taking care of your baby at home, don’t worry about the tight space. This little thing does not take up space and can be used anytime and anywhere.

[0057] 5. If your home is small, you can use this space-saving item at any time to relax and make your life more comfortable.

[0058] …

[0059] Point analysis item 2: Adapt to a variety of seats, even beds

[0060] Colloquial / scenario-based results:

[0061] 1. Whether you are sitting on a chair or lying on the bed, it can adapt to your situation, allowing you to enjoy a relaxing time in any occasion.

[0062] 2. If you are looking for a product that can adapt to a variety of seats or even beds, this one is definitely your best choice.

[0063] 3. This massage artifact is super flexible. It can fit perfectly on your sofa, office chair or bed at home, allowing you to enjoy professional-level massage relaxation wherever you lie!

[0064] 4. This innovative product has a wide range of seat compatibility. Simply pair it with any chair to immerse yourself in a comfortable massage experience anytime and anywhere.

[0065] …

[0066] 3. Automatic prompt word project

[0067] a) Automatically select relevant content after brief processing according to user needs

[0068] The system uses preset prompt word templates for analyzing content relevance and a large language model to perform content relevance analysis on user needs and brief content. It automatically selects content related to user needs from the parsed and processed brief information, accurately matches user needs with key information in the brief, and ensures that the generated copy is highly relevant and more authentic.

[0069] Example: Prompt word template: "This is the user requirement: (user requirement). Please analyze the requirement and select appropriate parameters to use as material. Parameters: ......".

[0070] The user requires the brand information to be introduced in the copy. The system automatically completes the prompt word template as input, calls the large language model to obtain the content corresponding to the "brand introduction" parameter as material, and prepares for the subsequent copy generation.

[0071] b) Construct prompt words to generate final copy

[0072] After obtaining the material in step a) above, the system constructs the final prompt word in combination with the dynamic prompt word template, takes the final prompt word as input, and calls the large language model to generate the final marketing copy.

[0073] Example: Dynamic prompt word template: "...(user needs), please combine these materials to generate the copy: (material)".

[0074] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. Marketing copy generation method based on marketing brief recognition and analysis engineering, It is characterized in that The marketing brief is a key document used to guide marketing activities. It outlines the core elements of the project, including project background, goals, target audience, and core information. By using a large language model, the system can automatically identify the marketing brief and extract key product marketing parameters from it. After the brief recognition is completed, the extracted product introduction parameters and product selling point parameters are analyzed point by point using the big model. This step divides the information contained in the two parameters into multiple point analysis items. The automatic prompt word project automatically selects relevant content after brief processing according to user needs, constructs prompt words and generates the final copy.

2. The method for generating marketing copy based on marketing brief recognition and analysis engineering according to claim 1, It is characterized in that The system uses a preset prompt word template for analyzing content relevance and a large language model to perform content relevance analysis on user needs and brief content, thereby realizing the function of automatically selecting content related to user needs from the parsed and processed brief information, accurately matching user needs with key information in the brief, and ensuring that the generated copy is highly relevant and more authentic.

3. The method for generating marketing copy based on marketing brief recognition and analysis engineering according to claim 2, It is characterized in that The user demand requires the introduction of brand information in the copy. The system automatically completes the prompt word template as input, calls the large language model to obtain the content corresponding to the "brand introduction" parameter as material, and prepares for the subsequent copy generation.

4. The method for generating marketing copy based on marketing brief recognition and analysis engineering according to claim 3, It is characterized in that After obtaining the material, the system constructs the final prompt word in combination with the dynamic prompt word template, takes the final prompt word as input, and calls the large language model to generate the final marketing copy.

5. The marketing copy generation system based on marketing brief recognition and analysis engineering according to claim 1, It is characterized in that Extracting text from the brief uploaded by the user; Word file: directly extract the text content in the file; There are two methods for processing PDF files: pure image extraction method and text + image extraction method; Pure image extraction method: extract each page in the PDF as an image, and then call the large model to extract the text in the image; Text + image extraction method: extract the text in the PDF page by page. When a page of text is extracted, extract the image of that page. Then call the large model to extract the text in the image. Combine the two texts to get the text of the current page. The large model is used to identify and classify the extracted text according to preset parameters, including general parameters and specific industry segmentation parameters.

Images