Marketing knowledge-based dynamic cue word engineering system and readable medium

Abstract

The marketing knowledge-based dynamic cue word engineering method comprises the following steps of: subdividing marketing industries and constructing an industry parameter library; establishing a general cue word database; creating an industry cue word database; constructing a feedback prompt word database; obtaining marketing task information; matching general and industry specific cue words; fusing the matched cue words to construct a new text cue word; and inputting the new text cue word to the language large model to generate a marketing document.

Description

Technical Field

[0001] The invention relates to the field of marketing, and in particular to a dynamic prompt word engineering system and a readable medium based on marketing knowledge. Background Art

[0002] With the rapid development of natural language processing and generative artificial intelligence technologies, the marketing industry has gradually introduced intelligent means to optimize marketing content by improving the performance of large language model AI (such as deepseek, Wenxin Yiyan, iFlytek Spark, etc.) in marketing and promotion tasks through prompt word engineering.

[0003] However, currently, AI mainly uses the marketing needs input by users as prompts to generate content based on the prompts, lacking the ability to dynamically adjust and customize marketing needs. In addition, these methods lack the guidance of professional marketing knowledge for specific industries, resulting in low quality and low efficiency of the generated marketing copy, which cannot fully meet the needs of different marketing scenarios and users. Summary of the invention

[0004] The present invention provides a dynamic prompt word engineering system and method based on marketing knowledge, and the technical problems to be solved are:

[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0006] The method of dynamic prompt word engineering based on marketing knowledge includes the following steps:

[0007] Step S100: Segment the marketing industry according to the industry vertical classification level; and construct industry parameter libraries according to the above different industries;

[0008] Step S200: construct a database of general prompt words; the general prompt words include general parameters, basic prompts, a wide range of marketing scenarios and general strategies, providing a general basic framework for generating marketing copy;

[0009] Step S300, constructing a database of industry prompt words; industry prompt words are a supplement to general prompt words; industry prompt words are a supplement to general prompt words, and industry prompt words provide more targeted prompt words according to specific content and marketing strategies of different industries;

[0010] Step S400, constructing a database of feedback prompt words;

[0011] Step S500, obtaining marketing brief information and / or task information such as demand parameters;

[0012] Step S600, matching corresponding general prompt words from a database of general prompt words according to the above task information;

[0013] Step S700, matching corresponding industry prompt words from an industry prompt word database according to the industry-related parameters in the task information;

[0014] Step S800, integrating the above-mentioned matched general prompt words and industry prompt words to construct new text prompt words;

[0015] Step S900: input the AI ​​text prompt words into the language model, and the language model generates text.

[0016] Further, in step S200, the general prompt words used to generate the marketing copy include but are not limited to:

[0017] Content types: product recommendations, reviews, check-ins, and knowledge popularization;

[0018] Title elements: Use keywords such as "how to", "recommendation", "reveal" to attract attention;

[0019] Personality style: Generate copy in the voice of professional bloggers, mothers, students, and housewives;

[0020] Content length: Shorter - 70 words, Medium - 200 words, Longer - 400 words to suit different marketing needs.

[0021] Further, in step S400,

[0022] Feedback prompt words are prompt words constructed based on user feedback, which are used to dynamically adjust and optimize prompt words and generate personalized prompt words. By collecting user feedback data in actual use, the system can continuously optimize and improve prompt words to improve their effectiveness and practicality.

[0023] Further, step S500 specifically includes:

[0024] Step S510: parse these marketing briefs, extract key information and convert it into colloquial and scenario-based prompt words;

[0025] Step S520: Manual review and modification of the parsed information may also be performed.

[0026] Further, in step S1000, if the user is not satisfied with the above-generated text, and the user evaluates the text, the evaluation content is adjusted and entered into the database of feedback prompt words, and the last feedback information is supplemented to construct new text prompt words, and step S900 is performed; otherwise, the language model accepted by the user generates the text to obtain the final text.

[0027] Furthermore, the specific steps of forming the feedback prompt words are as follows:

[0028] 1. Collect user feedback

[0029] After the system generates the copy using the original text prompt words, it provides scoring and evaluation functions. Users can rate the copy and describe specific problems or improvement suggestions through the open evaluation box;

[0030] 2. Adjust the prompt words for new texts

[0031] The collected user feedback is input into the big model, and the model generates corresponding feedback prompt words based on the feedback;

[0032] 3. Online AB testing

[0033] The effectiveness of the improved prompt words is judged through online AB testing: the original prompt words are group A and the improved prompt words are group B; feedback data on the copywriting from the two groups of customers is collected. If the data of group B is significantly better than that of group A, the prompt words of group B will be automatically adopted; otherwise, the prompt words will be continuously optimized.

[0034] A document generation device comprises a processor and a non-transitory memory having instructions thereon, wherein the instructions, when executed by the processor, enable the processor to implement the method according to any one of claims 1 to 6.

[0035] A non-transitory computer-readable medium storing a code, wherein when the code is executed by a processor, the processor executes the method according to any one of claims 1 to 6. The beneficial effects of the present invention are:

[0036] 1. Improve the pertinence and effectiveness of prompt words in the marketing field:

[0037] Structuring prompt words by industry and level can provide professional prompt words for different marketing scenarios and improve the quality and effectiveness of marketing copy generation.

[0038] 2. Realize dynamic adjustment and personalized customization of prompt words:

[0039] Through hierarchical management and user feedback adjustment, the system can continuously optimize and personalize prompt words to improve user satisfaction.

[0040] 3. Simplify the marketing br ief parsing process and improve the accuracy of generated content:

[0041] The marketing br ief analysis function enables users to quickly convert marketing goals into specific action plans and improve work efficiency. At the same time, by accurately analyzing the marketing br ief, the system can improve the effect of generated content, make the generated content more in line with the user's marketing needs, and ensure the effectiveness and relevance of the copy in practical applications.

[0042] 4. Enhance the diversity and creativity of marketing strategies:

[0043] The system combines the generation results of multiple important marketing parameters and multiple large models to provide rich marketing strategy suggestions and promote creativity and diversification. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of the steps of the present invention;

[0045] Figure 2 The process from inputting marketing demand information to outputting copywriting for this application;

[0046] Figure 3 This is the multi-level prompt word utilization process in the present invention. DETAILED DESCRIPTION

[0047] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings, wherein the same components are represented by the same reference numerals.

[0048] It should be noted that the words "front", "rear", "left", "right", "up" and "down" used in the following description refer to directions in the accompanying drawings, and the words "inside" and "outside" refer to directions toward or away from the geometric center of a specific component, respectively.

[0049] In order to make the contents of the present invention more clearly understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0050] The method of dynamic prompt word engineering based on marketing knowledge includes the following steps:

[0051] Step S100: subdivide the marketing industry according to the industry vertical classification level; and construct an industry parameter library according to the above different industries.

[0052] Step S110, we build a special prompt word library according to different marketing industries (including but not limited to personal care, home, film and television, food, parent-child, etc.). In order to ensure that the prompt words can accurately meet the needs of specific industries and sub-industries, we further subdivide the industry into different levels; first-level classification: such as personal care, home, food, etc.; second-level classification: taking personal care as an example, the second-level classification includes fabric care, mother and baby, skin care, etc.; third-level classification: under the skin care classification, the third-level classification can be subdivided into facial skin care, hand skin care, body care, etc.

[0053] Step S120: We have developed special parameters for different industries and sub-industries to ensure the accuracy and effectiveness of the prompt words. For example:

[0054] Beauty industry: In the skin care category, parameters include applicable skin type, ingredient analysis, clinical trial data, etc. These parameters can help generate marketing copy that is more in line with actual needs.

[0055] Home appliance industry: Parameters include product functions, technical parameters, energy efficiency ratings, user reviews, etc. These parameters help highlight the product's technical advantages and user experience.

[0056] Food industry: Parameters include nutritional content, taste description, food safety certification, etc. These parameters can generate more detailed and credible product descriptions.

[0057] In the prompt word project, we combine industry-specific keywords, phrases and marketing strategies to build an efficient prompt word library. Each prompt word library is optimized by professional marketers to ensure its application effect in specific industries and scenarios. In this way, prompt words can not only provide basic parameter settings, but also generate more accurate and targeted marketing copy in combination with the needs of specific industries. Take the beauty industry as an example:

[0058] Keywords: moisturizing, antioxidant, no additives.

[0059] Phrases: Deeply nourishing, anti-aging formula, all-day hydration.

[0060] Marketing strategy: Emphasize the safety of product ingredients and effectiveness of use.

[0061] Some of the generated copy: "The mask uses an additive-free formula, rich in natural moisturizing ingredients, deeply moisturizes the skin, effectively resists oxidation, and keeps your skin hydrated and smooth all day long."

[0062] Through the above detailed industry classification and specific parameter settings, we can highlight the differences between industries and make prompt words more in line with different market needs.

[0063] For example, the beauty industry focuses on the description of ingredients and effects, while the home appliance industry pays more attention to technical parameters and user experience. Such detailed processing makes the generated copy more persuasive and attractive in their respective industries.

[0064] Step S200: Build a database of common prompt words; common prompt words include common parameters, basic prompts, a wide range of marketing scenarios and common strategies, providing a common basic framework for generating marketing copy.

[0065] General prompts are suitable for general parameter settings and basic prompts, and can meet most basic needs. General prompts cover a wide range of marketing scenarios and general strategies, providing a general basic framework for generating marketing copy. General prompts used to generate marketing copy include but are not limited to:

[0066] Content types: product recommendations, reviews, check-ins, knowledge popularization, etc.

[0067] Title elements: Use keywords such as "how to", "recommended", "revealed" to attract attention.

[0068] Personality style: Generate copy in the voice of different characters such as professional bloggers, mothers, students, housewives, etc.

[0069] Content length: shorter (about 70 words), medium (about 200 words), longer (about 400 words) to meet different marketing needs.

[0070] Step S300: Build a database of industry prompt words; industry prompt words are a supplement to general prompt words. Industry prompt words are a supplement to general prompt words. Industry prompt words provide more targeted prompt words based on specific content and marketing strategies of different industries. The details are as follows:

[0071] Home appliance industry: Industry prompt words include "energy-saving products", "smart home functions", "after-sales service guarantee", etc. These prompt words can generate targeted copywriting, such as emphasizing the energy-saving advantages or intelligent features of the product to attract the attention of specific consumer groups.

[0072] Step S400, constructing a database of feedback prompt words;

[0073] Feedback prompt words are constructed based on user feedback and are used to dynamically adjust and optimize prompt words to generate personalized prompt words. By collecting user feedback data in actual use, the system can continuously optimize and improve prompt words to improve their effectiveness and practicality.

[0074] Step S500, obtaining marketing brief information and / or task information such as demand parameters;

[0075] The specific steps include:

[0076] Step S510: parse the marketing brief and extract the marketing information, and then colloquially translate the marketing information; the marketing brief is a key document used to guide marketing activities, which outlines the core elements of the project, including project background, goals, target audience, and core information. The system can automatically parse these marketing briefs, extract key information, and convert it into colloquial and scenario-based prompts. This process not only improves the accuracy of marketing copy generation, but also facilitates the processing and understanding of large models, so that the generated content is more in line with the actual marketing needs of users. Example: parse "increase product awareness" into specific action prompts, such as "publish attractive product introduction copy on social media."

[0077] Step S520: Manual review and modification of the parsed information may also be performed.

[0078] Step S600, matching corresponding general prompt words from a database of general prompt words according to the above task information;

[0079] Step S700, matching corresponding industry prompt words from an industry prompt word database according to the industry-related parameters in the task information;

[0080] Step S800, integrating the above-mentioned matched general prompt words and industry prompt words to construct new text prompt words;

[0081] Step S900: input the AI ​​text prompt words into the language model, and the language model generates text.

[0082] In summary, building a general prompt word database as the basic framework and providing industry prompt words as a supplement ensures the flexibility and scalability of the method and can adapt to the ever-changing market and marketing needs. By segmenting industries and sub-industries and building a specialized prompt word database, it ensures that marketing copy can accurately meet the needs of specific industries, thereby improving marketing effectiveness.

[0083] Step S1000: If the user is not satisfied with the generated text, and the user evaluates the text, the evaluation content is adjusted and entered into the feedback prompt word database, and the last feedback information is supplemented to construct new text prompt words, and step S900 is performed; otherwise, the language model accepted by the user generates the text to obtain the final text (target text).

[0084] Among them, the specific steps of forming the feedback prompt words are as follows:

[0085] 1. Collect user feedback

[0086] After the system generates the copy using the text prompt words, it provides the scoring and evaluation functions. Users can rate the copy (1-5 stars) and describe specific problems or improvement suggestions through the open evaluation box. For example, a user rated the copy of a home appliance product as 3 stars and pointed out in the open evaluation: "The copy lacks a specific description of the energy-saving function."

[0087] 2. Adjust the prompt words for new texts

[0088] The collected user feedback is input into the big model (such as Wenxinyiyan), and the model generates corresponding feedback prompts based on the feedback. These feedback prompts are analyzed and optimized to better meet user needs. Example: Input user feedback into the big model: "The copy lacks a specific description of the energy-saving function." The big model returns the feedback prompt: "Add a description of the product's energy-saving function in the copy...", and the feedback prompt is combined with the raw text prompt to generate an improved raw text prompt.

[0089] 3. Online AB testing

[0090] The effectiveness of the improved prompt words is judged through online AB testing: the original prompt words are group A, and the improved prompt words are group B. The feedback data of the two groups of customers on the copywriting is collected. If the data of group B is obviously better than that of group A, the prompt words of group B will be automatically adopted; otherwise, the prompt words will be continuously optimized.

[0091] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. The method of dynamic prompt word engineering based on marketing knowledge includes the following steps: Step S100: Segment the marketing industry according to the industry vertical classification level; and construct industry parameter libraries according to the above different industries; Step S200: construct a database of general prompt words; the general prompt words include general parameters, basic prompts, a wide range of marketing scenarios and general strategies, providing a general basic framework for generating marketing copy; Step S300, constructing a database of industry prompt words; industry prompt words are a supplement to general prompt words; industry prompt words are a supplement to general prompt words, and industry prompt words provide more targeted prompt words according to specific content and marketing strategies of different industries; Step S400, constructing a database of feedback prompt words; Step S500, obtaining marketing brief information and / or task information such as demand parameters; Step S600, matching corresponding general prompt words from a database of general prompt words according to the above task information; Step S700, matching corresponding industry prompt words from an industry prompt word database according to the industry-related parameters in the task information; Step S800, integrating the above-mentioned matched general prompt words and industry prompt words to construct new text prompt words; Step S900: input the AI ​​text prompt words into the language model, and the language model generates text.

2. The method of dynamic prompt word engineering based on marketing knowledge according to claim 1, characterized in that: For step S200, the general prompt words used to generate marketing copy include but are not limited to: Content types: product recommendations, reviews, check-ins, and knowledge popularization; Title elements: Use keywords such as "how to", "recommendation", "secrets" to attract attention; Personality style: Generate copy in the voice of professional bloggers, mothers, students, and housewives; Content length: Shorter - 70 words, Medium - 200 words, Longer - 400 words to suit different marketing needs.

3. The method of dynamic prompt word engineering based on marketing knowledge according to claim 1, characterized in that: In step S400, Feedback prompt words are prompt words constructed based on user feedback, which are used to dynamically adjust and optimize prompt words and generate personalized prompt words. By collecting user feedback data in actual use, the system can continuously optimize and improve prompt words to improve their effectiveness and practicality.

4. The method of dynamic prompt word engineering based on marketing knowledge according to claim 1, characterized in that: Step S500 specifically includes: Step S510: parse these marketing briefs, extract key information and convert it into colloquial and scenario-based prompt words; Step S520: Manual review and modification of the parsed information may also be performed.

5. The method of dynamic prompt word engineering based on marketing knowledge according to claim 1, characterized in that: Step S1000: If the user is not satisfied with the generated text, and the user evaluates the text, the evaluation content is adjusted and entered into the feedback prompt word database, and the last feedback information is supplemented to construct new text prompt words, and step S900 is performed; otherwise, the language model accepted by the user generates the text to obtain the final text.

6. The method of dynamic prompt word engineering based on marketing knowledge according to claim 1, It is characterized in that Among them, the specific steps of forming the feedback prompt words are as follows:

1. Collect user feedback After the system generates the copy using the original text prompt words, it provides scoring and evaluation functions. Users can rate the copy and describe specific problems or improvement suggestions through the open evaluation box; 2. Adjust the prompt words for new texts The collected user feedback is input into the big model, and the model generates corresponding feedback prompt words based on the feedback; 3. Online AB testing The effectiveness of the improved prompt words is judged through online AB testing: the original prompt words are group A and the improved prompt words are group B; feedback data on the copywriting from the two groups of customers is collected. If the data of group B is significantly better than that of group A, the prompt words of group B will be automatically adopted; otherwise, the prompt words will be continuously optimized.

7. A copywriting generation device, comprising a processor and a non-transitory memory having instructions thereon, wherein: The instructions, when executed by the processor, cause the processor to implement the method according to any one of claims 1-6.

8. A non-transitory computer readable medium storing codes, which, when executed by a processor, cause the processor to perform the method according to any one of claims 1 to 6.

Images