A marketing corpus generation method, device and readable storage medium based on a large model and knowledge base

Abstract

The present invention relates to the field of data processing, and specifically to a marketing corpus generation method, device and readable storage medium based on a large model and a knowledge base. The marketing corpus generation method includes the following steps: S1. When a marketing staff inputs a marketing query, the most likely top 3 marketing categories are obtained through a marketing classification model; S2. According to the marketing category obtained in step S1, the corresponding user group is retrieved from the user database using a fixed SQL template; S3. After the user inputs the marketing query, the marketing query is input into the large model, and the marketing query is rewritten using the large model, while the user query is expanded; S4. The data is input into the large model, and the large model generates a preliminary draft of the marketing copy. The preliminary generated marketing corpus is then input into the large model again for content review. The present invention improves the efficiency of marketing corpus generation, reduces the cost of marketing corpus generation, and can generate accurate and targeted marketing corpus.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a marketing corpus generation method, device and readable storage medium based on a large model and a knowledge base. Background Art

[0002] Marketing corpus generation technology aims to help companies or individuals automatically generate text content that meets marketing needs. These text contents can be used in various forms of marketing activities such as advertising, social media, emails, web pages, etc.

[0003] Traditional marketing corpus creation and generation methods mainly include the following three types: 1. Manual writing: marketers manually write marketing copy based on product characteristics, target audiences and market trends; 2. Template generation method: It is one of the earliest automated marketing corpus generation technologies. It relies on pre-designed text templates. Marketers only need to fill in the variable part of the template to generate the required marketing content; 3. Text generation technology based on traditional machine learning: Generate diverse text content through data collection, data preprocessing, feature extraction, model training and text generation steps.

[0004] However, the above method has the following shortcomings:

[0005] 1. Manual writing is time-consuming, labor-intensive and inefficient;

[0006] 2. The text content generated by the template generation method lacks diversity and innovation. All generated content follows the same structure and language pattern, which easily leads to users getting bored with marketing content. In addition, the template generation method has poor adaptability to different scenarios and needs. Once the marketing needs change, the template needs to be manually updated and adjusted, which increases maintenance costs.

[0007] 3. The defects of traditional machine learning methods are strong data dependence, complex feature engineering, limited model generalization ability and insufficient ability to cope with complex scenarios; obtaining and labeling large amounts of high-quality data is time-consuming and expensive; feature engineering requires rich domain knowledge, a large number of experiments and debugging, and model development is difficult and costly; the model is prone to overfitting problems, that is, it performs well on training data but does not perform well in actual applications. Summary of the invention

[0008] The purpose of the present invention is to address the problems existing in the background technology and to propose a marketing corpus generation method, device and readable storage medium based on a large model and a knowledge base, which improves the efficiency of marketing corpus generation, reduces the cost of marketing corpus generation, and can generate accurate and targeted marketing corpus.

[0009] In a first aspect, the present invention proposes a method for generating marketing corpus based on a large model and a knowledge base, comprising the following steps:

[0010] S1. When a marketing person inputs a marketing query, the marketing classification model obtains the top 3 most likely marketing categories.

[0011] S2. According to the marketing category obtained in step S1, the corresponding user group is retrieved from the user database using a fixed SQL template;

[0012] S3. After the user inputs a marketing query, the marketing query is input into the big model, and the big model is used to rewrite the marketing query, standardize the user query expression, and expand the user query to generate several similar queries;

[0013] S4. Input the data obtained in steps S1-S3 into the big model. The big model generates a preliminary draft of the marketing copy by learning the existing marketing corpus and knowledge base. The preliminary generated marketing corpus is then input into the big model again for content review. Based on the review results, the big model proposes modification suggestions or automatically makes adjustments.

[0014] Preferably, the marketing classification model is a classification model obtained through supervised training using pre-labeled marketing data.

[0015] Preferably, in step S3, user group information, marketing keywords, product information and specific marketing queries are integrated.

[0016] Preferably, in step S3, when rewriting the marketing query, the big model first normalizes the original query, including correcting spelling errors, using standard grammatical structures, and replacing uncommon words; the big model generates a series of new queries similar to the original query, the new queries use synonyms or slightly change the query structure;

[0017] The rewritten marketing query is converted into a semantic vector through a vector generation model. The semantic vector represents the semantic information of the text. The marketing knowledge most similar to these semantic vectors is searched in the vector knowledge base. The marketing knowledge is converted into semantic vectors in advance. The marketing knowledge that best matches the user query is found by calculating the similarity between semantic vectors.

[0018] In a second aspect, the present invention proposes a marketing corpus generation device based on a large model and a knowledge base, comprising a user marketing intention recognition module, a user group matching module, a marketing information acquisition module, and a marketing corpus generation and review module;

[0019] The user marketing intention recognition module is used to obtain the most likely top 3 marketing categories through the marketing classification model when the marketer inputs a marketing query;

[0020] The user group matching module is used to retrieve the corresponding user group from the user database using a fixed SQL template according to the marketing category obtained by the user marketing intention identification module;

[0021] The marketing information acquisition module is used to input the marketing query into the big model after the user inputs the marketing query, and use the big model to rewrite the marketing query, standardize the user query expression, and expand the user query to generate several similar queries;

[0022] The marketing corpus generation and review module is used to input the data obtained through the user marketing intention identification module, the user group matching module and the marketing information acquisition module into the big model. The big model generates a preliminary draft of the marketing copy by learning the existing marketing corpus and knowledge base. The preliminary generated marketing corpus is then input into the big model again for content review. Based on the review results, the big model proposes modification suggestions or makes automatic adjustments.

[0023] Preferably, the marketing classification model is a classification model obtained through supervised training using pre-labeled marketing data.

[0024] Preferably, when processing data, the marketing information acquisition module integrates user group information, marketing keywords, product information and specific marketing queries.

[0025] Preferably, when the marketing information acquisition module rewrites the marketing query, the big model first normalizes the original query, including correcting spelling errors, using standard grammatical structures, and replacing uncommon words; the big model generates a series of new queries similar to the original query, the new queries use synonyms or slightly change the query structure;

[0026] The rewritten marketing query is converted into a semantic vector through a vector generation model. The semantic vector represents the semantic information of the text. The marketing knowledge most similar to these semantic vectors is searched in the vector knowledge base. The marketing knowledge is converted into semantic vectors in advance. The marketing knowledge that best matches the user query is found by calculating the similarity between semantic vectors.

[0027] In a third aspect, the present invention proposes a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned marketing corpus generation method based on a large model and a knowledge base.

[0028] Compared with the prior art, the present invention has the following beneficial technical effects:

[0029] The present invention utilizes the powerful capabilities of artificial intelligence and machine learning. By combining a large model with a knowledge base, the present invention automatically identifies the marketing intentions of users, matches user groups, obtains marketing information, and generates more innovative and personalized marketing corpus, which greatly improves the efficiency and accuracy of marketing corpus generation and improves marketing effectiveness. At the same time, by optimizing the method of combining a large model with a knowledge base, the cost of generating marketing corpus is reduced. In addition, by utilizing the powerful processing and learning capabilities of the large model and knowledge base, marketing knowledge is continuously learned and updated to adapt to market changes and maintain the freshness and effectiveness of marketing corpus. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a flow chart of a method for generating marketing corpus based on a large model and a knowledge base according to Embodiment 1 of the present invention;

[0031] Figure 2 Matching flowcharts to user groups;

[0032] Figure 3 Get a flow chart for marketing knowledge. DETAILED DESCRIPTION

[0033] Embodiment 1

[0034] like Figure 1-Figure 3 As shown, the present embodiment proposes a method for generating marketing corpus based on a large model and a knowledge base, comprising the following steps S1-S4:

[0035] S1. When a marketing person inputs a marketing query, the marketing classification model obtains the most likely top 3 marketing categories. The marketing classification model is a classification model obtained after supervised training using pre-labeled marketing data, such as the BERT model. Query refers to the search phrase or sentence input by the user to find relevant web pages or other information, which can be understood as a way of asking questions or requesting information.

[0036] S2. According to the marketing category obtained in step S1, the corresponding user group is retrieved from the user database using a fixed SQL template; a SQL template is a predefined SQL query used to perform specific database operations; these templates contain parameter placeholders, which are replaced by actual values ​​when the query is executed.

[0037] S3. After the user inputs a marketing query, the marketing query is input into the big model, and the big model is used to rewrite the marketing query, standardize the user query expression, and expand the user query to generate several similar queries, so as to facilitate the subsequent extraction of more diverse marketing-related information from the marketing knowledge base; user group information, marketing keywords, product information and specific marketing queries are integrated. This information is the basis for the big model to generate marketing corpus, ensuring that the model can understand the target audience, product features and marketing goals.

[0038] When rewriting marketing queries, the big model first normalizes the original queries, including correcting spelling errors, using standard grammatical structures, and replacing uncommon words to ensure the quality of the queries; the big model generates a series of new queries that are similar to the original queries. The new queries use synonyms or slightly change the query structure to increase the diversity of the queries.

[0039] The rewritten marketing query is converted into a semantic vector through a vector generation model. The vector generation model is usually based on deep learning, such as Word2Vec, BERT, etc., which can convert text data into a high-dimensional vector. The semantic vector can represent the semantic information of the text. The marketing knowledge most similar to these semantic vectors is searched in the vector knowledge base. The vector knowledge base is pre-established and contains a large amount of marketing knowledge. This marketing knowledge has also been pre-converted into semantic vectors. By calculating the similarity between semantic vectors (such as cosine similarity), the marketing knowledge that best matches the user query is found. This marketing knowledge includes marketing keywords, marketing precautions, product information, etc.

[0040] S4. Input the data obtained in steps S1-S3 into the big model. The big model generates a preliminary draft of the marketing copy by learning the existing marketing corpus and knowledge base. This step utilizes the generation ability of artificial intelligence to propose multiple corpus solutions in an efficient manner. The initially generated marketing corpus is then input into the big model again for content review. At this stage, the big model checks the consistency, accuracy, relevance of the copy, and its fit with the marketing goals. Based on the review results, the big model proposes modification suggestions or automatically makes adjustments to improve the quality of the copy. Modifications include changing the wording, adjusting the structure, or enhancing the attractiveness of the copy.

[0041] This embodiment utilizes the powerful capabilities of artificial intelligence and machine learning. By combining a large model with a knowledge base, it automatically identifies the user's marketing intention, matches user groups, obtains marketing information, and generates more innovative and personalized marketing corpus, which greatly improves the efficiency and accuracy of marketing corpus generation and improves marketing effectiveness. At the same time, by optimizing the method of combining a large model with a knowledge base, the cost of marketing corpus generation is reduced. In addition, by adding user group stratification information and marketing precautions, the generated marketing corpus is made more personalized. Furthermore, by adding a large model review mechanism, potential text errors and paragraph structure errors in the generation process are modified and optimized to ensure the quality and consistency of marketing corpus, further improve the success rate of marketing, and use the powerful processing and learning capabilities of the large model and knowledge base to continuously learn and update marketing knowledge, adapt to market changes, and maintain the freshness and effectiveness of marketing corpus.

[0042] Embodiment 2

[0043] like Figure 1-Figure 3 As shown, the present embodiment proposes a marketing corpus generation device based on a big model and a knowledge base, which corresponds to the marketing corpus generation method based on a big model and a knowledge base in the first embodiment. The marketing corpus generation device includes a user marketing intention recognition module, a user group matching module, a marketing information acquisition module and a marketing corpus generation and review module.

[0044] The user marketing intention identification module is used to obtain the most likely top 3 marketing categories through the marketing classification model when the marketer inputs a marketing query. The marketing classification model is a classification model obtained after supervised training using pre-labeled marketing data.

[0045] The user group matching module is used to retrieve the corresponding user group from the user database using a fixed SQL template based on the marketing category obtained by the user marketing intention identification module.

[0046] The marketing information acquisition module is used to input the marketing query into the big model after the user inputs the marketing query, and use the big model to rewrite the marketing query, standardize the user query expression, and expand the user query to generate several similar queries. When the marketing information acquisition module processes data, it integrates user group information, marketing keywords, product information, and specific marketing queries. This information is the basis for the big model to generate marketing corpus, ensuring that the model can understand the target audience, product features, and marketing goals.

[0047] When the marketing information acquisition module rewrites the marketing query, the big model first normalizes the original query, including correcting spelling errors, using standard grammatical structures, and replacing uncommon words to ensure the quality of the query. The big model generates a series of new queries similar to the original query, which use synonyms or slightly change the query structure to increase the diversity of the query.

[0048] The rewritten marketing query is converted into a semantic vector through a vector generation model. The semantic vector represents the semantic information of the text. The marketing knowledge most similar to these semantic vectors is searched in the vector knowledge base. The marketing knowledge is converted into semantic vectors in advance. The marketing knowledge that best matches the user query is found by calculating the similarity between semantic vectors (such as cosine similarity).

[0049] The marketing corpus generation and review module is used to input the data obtained through the user marketing intention identification module, the user group matching module and the marketing information acquisition module into the big model. The big model generates a preliminary draft of the marketing copy by learning the existing marketing corpus and knowledge base. The preliminary generated marketing corpus is then input into the big model again for content review. The big model checks the consistency, accuracy, relevance and fit of the copy with the marketing goals. Based on the review results, the big model proposes modification suggestions or automatically makes adjustments to improve the quality of the copy. Modifications may include changing the wording, adjusting the structure or enhancing the attractiveness of the copy.

[0050] This embodiment has the following advantages:

[0051] 1. Improve the efficiency and quality of marketing corpus generation: By combining a large model and a knowledge base, this embodiment can quickly and accurately identify the user's marketing intention, match the corresponding user group, and obtain relevant marketing information. Compared with the traditional method of manually writing marketing corpus, this embodiment can greatly reduce labor costs and time costs, while improving the efficiency and quality of marketing corpus generation.

[0052] 2. Personalized marketing corpus: This embodiment can generate personalized marketing corpus that meets user needs based on the user's marketing query and user group information. This personalized customization method can better attract the user's attention and improve marketing effects.

[0053] 3. Use artificial intelligence technology to review and optimize content: This embodiment uses a large model to review and optimize the content of the initially generated marketing corpus to ensure the consistency, accuracy, relevance and fit with marketing goals. This automated content review and optimization method can further improve the quality of marketing corpus and enhance user experience.

[0054] Embodiment 3

[0055] This embodiment provides a readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for generating marketing corpus based on a large model and a knowledge base in the first embodiment are implemented.

[0056] The readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program modules or other data). The readable storage medium includes, but is not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable readonly memory), flash memory or other memory technology, CD-ROM (Compact Disc Read-Only Memory), digital versatile disk (DVD) or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage device, or any other medium that can be used to store the desired information and can be accessed by a computer.

[0057] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A marketing corpus generation method based on a large model and a knowledge base, characterized in that: The following steps are involved: S1. When a marketing person inputs a marketing query, the marketing classification model obtains the most likely top 3 marketing categories. S2. According to the marketing category obtained in step S1, the corresponding user group is retrieved from the user database using a fixed SQL template; S3. After the user inputs a marketing query, the marketing query is input into the big model, and the big model is used to rewrite the marketing query, standardize the user query expression, and expand the user query to generate several similar queries; S4. Input the data obtained in steps S1-S3 into the big model. The big model generates a preliminary draft of the marketing copy by learning the existing marketing corpus and knowledge base. The preliminary generated marketing corpus is then input into the big model again for content review. Based on the review results, the big model proposes modification suggestions or automatically makes adjustments.

2. The method for generating marketing corpus based on a large model and a knowledge base according to claim 1, characterized in that: The marketing classification model is a classification model obtained after supervised training using pre-labeled marketing data.

3. The method for generating marketing corpus based on a large model and a knowledge base according to claim 1, characterized in that: In step S3, user group information, marketing keywords, product information and specific marketing queries are integrated.

4. The method for generating marketing corpus based on a large model and a knowledge base according to claim 1, characterized in that: In step S3, when rewriting the marketing query, the big model first normalizes the original query, including correcting spelling errors, using standard grammatical structures, and replacing uncommon words; the big model generates a series of new queries that are similar to the original query, using synonyms or slightly changing the query structure; The rewritten marketing query is converted into a semantic vector through a vector generation model. The semantic vector represents the semantic information of the text. The marketing knowledge most similar to these semantic vectors is searched in the vector knowledge base. The marketing knowledge is converted into semantic vectors in advance. The marketing knowledge that best matches the user query is found by calculating the similarity between semantic vectors.

5. A marketing corpus generation device based on a large model and a knowledge base, characterized in that: include: User marketing intention recognition module, which is used to obtain the most likely top 3 marketing categories through the marketing classification model when the marketer inputs a marketing query; The user group matching module is used to extract the corresponding user group from the user database using a fixed SQL template according to the marketing category obtained by the user marketing intention identification module; The marketing information acquisition module is used to input the marketing query into the big model after the user inputs the marketing query, rewrite the marketing query using the big model, standardize the user query expression, and expand the user query to generate several similar queries; The marketing corpus generation and review module is used to input the data obtained through the user marketing intention identification module, the user group matching module and the marketing information acquisition module into the big model. The big model generates a preliminary draft of the marketing copy by learning the existing marketing corpus and knowledge base. The preliminary generated marketing corpus is then input into the big model again for content review. Based on the review results, the big model proposes modification suggestions or makes automatic adjustments.

6. The marketing corpus generation device based on a large model and a knowledge base according to claim 5, characterized in that: The marketing classification model is a classification model obtained after supervised training using pre-labeled marketing data.

7. The marketing corpus generation device based on a large model and a knowledge base according to claim 5, characterized in that: When the marketing information acquisition module processes data, it integrates user group information, marketing keywords, product information and specific marketing queries.

8. The marketing corpus generation device based on a large model and a knowledge base according to claim 5, characterized in that: When the marketing information acquisition module rewrites the marketing query, the big model first normalizes the original query, including correcting spelling errors, using standard grammatical structures, and replacing uncommon words; the big model generates a series of new queries that are similar to the original query, using synonyms or slightly changing the query structure; The rewritten marketing query is converted into a semantic vector through a vector generation model. The semantic vector represents the semantic information of the text. The marketing knowledge most similar to these semantic vectors is searched in the vector knowledge base. The marketing knowledge is converted into semantic vectors in advance. The marketing knowledge that best matches the user query is found by calculating the similarity between semantic vectors.

9. A readable storage medium, characterized in that: A computer program is stored on the readable storage medium, and when the computer program is executed by the processor, the steps of the method for generating marketing corpus based on a large model and a knowledge base as described in any one of claims 1 to 4 are implemented.

Images