Marketing copywriting iteration generation method and device based on potential semantic optimization and medium

Abstract

The invention provides a marketing copywriting iteration generation method and device based on potential semantic optimization and a medium. The method comprises the following steps: generating a plurality of initial marketing copywriting by a pre-training language model based on a marketing demand, and storing the initial marketing copywriting as an initial copywriting set; according to the preference sequence, constructing creative increasing partial order case pairs for the adjacent copywriting; for each partial order case pair, calling the to-be-trained language model to extract a corresponding statement-level potential semantic vector, and calculating preference discrimination loss and iterative prediction loss according to the potential semantic vector; inputting an actual marketing demand to the trained marketing copywriting model, generating a first version of draft copywriting and extracting a potential semantic vector, and feeding back the potential semantic vector to the marketing copywriting model for iterative generation to obtain a candidate copywriting set; and performing confidence evaluation on the candidate copywriting set based on the marketing copywriting model, and selecting a target marketing copywriting with the highest confidence. According to the method and the device, high-creativity, high-efficiency, robust and convergent marketing copywriting automatic iteration generation can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular relates to a marketing copy iterative generation method and device based on latent semantic optimization and a medium. BACKGROUND

[0002] With the rapid development of e-commerce and social media, the demand for high-quality marketing copy by enterprises has surged. Large language models have been widely used in advertising creative, content operation and other automatic copy generation scenarios due to their large-scale corpus learning ability. Marketing copy not only needs to accurately convey product selling points and marketing goals, but also emphasizes "creativity" to attract audience attention and promote conversion. The evaluation of creativity usually depends on the comprehensive judgment of the whole sentence or even the context, which is difficult to measure directly through word-level indicators.

[0003] The current mainstream approach mainly adopts the following two types: one is based on supervised fine-tuning or instruction fine-tuning, which minimizes the cross-entropy loss at the word level to make the model reproduce the phrases with higher frequency in the training set; the other is based on reinforcement learning human preference optimization (RLHF), which uses overall scores or ranking to update the model strategy gradient. The former can improve grammar and semantic correctness, but has limited improvement on creativity; the latter usually compares different creative level copies and uses word-level hidden state differences as optimization basis, which can easily lead to excessive noise in gradient signals, and lacks consistency with the "step-by-step revision" process in subsequent deployment scenarios.

[0004] The existing technology still has the following problems: the word-level optimization target cannot effectively capture the creative differences at the sentence or paragraph level, resulting in the convergence of the model-generated copy and the lack of novelty; the reinforcement learning based on global scores simultaneously processes multiple levels of copy, which is easy to confuse and difficult to converge to the real high-creativity direction; the existing solutions generally lack an iterative rewriting mechanism for the latent semantic space, and the model is difficult to improve the quality of the copy step by step according to user feedback during the reasoning phase. The above defects collectively result in insufficient adaptation of the model to marketing needs, making it difficult to meet the actual needs of the industry for high-creativity and high-efficiency copy production. SUMMARY

[0005] Therefore, the embodiments of the present application provide a marketing copy iterative generation method and device based on latent semantic optimization and a medium to solve the problems of inaccurate creativity capture, unstable model convergence and low efficiency of iterative rewriting in the prior art.

[0006] In a first aspect, the embodiment of the present application provides a marketing copy iterative generation method based on latent semantic optimization, comprising: obtaining marketing demand containing product information, audience information and marketing target; generating a plurality of initial marketing copies based on the marketing demand by a pre-trained language model and storing the initial marketing copies as an initial copy set; receiving preference input of an artificial expert on the initial copy set, and constructing a partial order copy pair with increasing creativity for adjacent copies according to a preference order; for each partial order copy pair, calling a to-be-trained language model to extract a corresponding sentence-level latent semantic vector, calculating a preference discrimination loss and an iterative prediction loss according to the latent semantic vector, and jointly updating parameters of the to-be-trained language model to obtain a trained marketing copy model; inputting an actual marketing demand into the trained marketing copy model to generate a first draft copy and extract a latent semantic vector, feeding back the latent semantic vector to the marketing copy model for iterative generation until a distance between latent semantic vectors of adjacent two copies does not exceed a preset threshold or a preset iteration number is reached, and obtaining a candidate copy set; performing confidence evaluation on the candidate copy set based on the marketing copy model, and selecting a target marketing copy with the highest confidence as a final output.

[0007] In a second aspect, the embodiment of the present application provides a marketing copy iterative generation device based on latent semantic optimization, comprising: an obtaining module configured to obtain marketing demand containing product information, audience information and marketing target; a storage module configured to generate a plurality of initial marketing copies based on the marketing demand by a pre-trained language model and store the initial marketing copies as an initial copy set; a constructing module configured to receive preference input of an artificial expert on the initial copy set, and construct a partial order copy pair with increasing creativity for adjacent copies according to a preference order; a training module configured to, for each partial order copy pair, call a to-be-trained language model to extract a corresponding sentence-level latent semantic vector, calculate a preference discrimination loss and an iterative prediction loss according to the latent semantic vector, and jointly update parameters of the to-be-trained language model to obtain a trained marketing copy model; a generating module configured to input an actual marketing demand into the trained marketing copy model to generate a first draft copy and extract a latent semantic vector, feed back the latent semantic vector to the marketing copy model for iterative generation until a distance between latent semantic vectors of adjacent two copies does not exceed a preset threshold or a preset iteration number is reached, and obtain a candidate copy set; and an output module configured to perform confidence evaluation on the candidate copy set based on the marketing copy model, select a target marketing copy with the highest confidence as a final output.

[0008] In a third aspect, the embodiment of the present application provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0009] In a fourth aspect, the present application provides a computer readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above method.

[0010] The above at least one technical solution adopted by the embodiments of the present application can achieve the following beneficial effects: By obtaining marketing demand containing product information, audience information and marketing target; generating a plurality of initial marketing scripts from a pre-trained language model based on the marketing demand and storing them as an initial script set; receiving preference input of artificial experts on the initial script set, and constructing a creative incremental partial order script pair for adjacent scripts according to the preference order; for each partial order script pair, calling a to-be-trained language model to extract a corresponding sentence-level latent semantic vector, calculating a preference discriminant loss and an iterative prediction loss according to the latent semantic vector, and jointly updating the parameters of the to-be-trained language model to obtain a trained marketing script model; inputting an actual marketing demand into the trained marketing script model to generate a first draft script and extract a latent semantic vector, and feeding the latent semantic vector back to the marketing script model for iterative generation until the latent semantic vector distance between adjacent two versions of scripts does not exceed a preset threshold or reaches a preset iteration number, obtaining a candidate script set; performing confidence evaluation on the candidate script set based on the marketing script model, and selecting a target marketing script with the highest confidence as the final output. The present application can realize automatic iterative generation of marketing scripts with high creativity, high efficiency and stable convergence. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0012] Figure 1 is a flowchart of the marketing script iterative generation method based on latent semantic optimization provided by the embodiments of the present application; Figure 2 is a structural schematic diagram of the marketing script iterative generation device based on latent semantic optimization provided by the embodiments of the present application; Figure 3 is a structural schematic diagram of the electronic device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0013] In the following description, for purposes of explanation and not limitation, specific details are set forth such as particular architectures, technologies, techniques, etc. in order to provide a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application can be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.

[0014] Marketing copy usually needs to attract readers to watch through "creativity". On the one hand, the current language modeling optimizes the output of the model in units of tokens. On the other hand, creativity is a relatively coarse-grained fuzzy concept, and usually needs to be jointly analyzed with the context to evaluate the creativity. Therefore, it is difficult to train creativity through token-level optimization goals, and it is necessary to introduce more suitable algorithms to improve the ability of the language model in this regard, and thus improve its ability to generate marketing copy.

[0015] In view of the defects of the existing solutions, the technical scheme of the present application proposes the following solution ideas: 1. For the marketing and promotion needs of specific scenarios, use a large language model to sample multiple copies for each scenario. Then, select the copy closest to the user's preference as the base copy by a marketing expert, and then rewrite it step by step, appropriately improving the creativity and alignment of human preferences of the copy each time, and finally accumulating multiple copies with gradually improved creativity for each scenario; 2. According to the obtained copy partial order data, fine-tune the model in the latent semantic space based on reinforcement learning. When fine-tuning, calculate the loss by comparing the copies of adjacent creativity levels, and require the model to align the latent semantic vectors at the sentence level, rather than aligning the hidden space states at the token level; 3. In actual generation, the same iterative improvement generation scheme is adopted. The model first generates a most basic copy according to the user's required scenario, and then inputs the latent semantic vector of this copy into the language model for iteration until the representation distance of the latent semantic vectors of the two iterations is lower than the threshold value designed by the user, and the iteration is stopped and the final result is output.

[0016] The technical scheme of the present application has at least the following advantages: 1. The reinforcement learning scheme for large language models in the past usually compares all creativity levels of the copy jointly, which is easy to cause confusion. The reinforcement learning in the present scheme emphasizes the comparison of the relationship between adjacent creativity levels of the copy, which is closer to the scenario of iterative generation of the copy after the deployment of the model, and makes the model make full use of the existing data and expert preferences; 2. Past large language model fine-tuning algorithms usually focus on the alignment of word-level latent space states, which does not conform to the true characteristics of creative evaluation. The fine-tuning scheme in this solution requires the model to adapt to the sentence-level latent semantic vector, which is more suitable for creative copy generation scenarios and also helps improve the efficiency and effectiveness of copy iteration generation after model deployment.

[0017] The technical solutions of the present application will be described in detail below in combination with the drawings and specific embodiments.

[0018] Figure 1 is a flowchart of a marketing copy iterative generation method based on latent semantic optimization provided by an embodiment of the present application. As shown in Figure 1 , the marketing copy iterative generation method based on latent semantic optimization can specifically include: S101, obtaining marketing demand containing product information, audience information and marketing target; S102, generating a plurality of initial marketing copies based on the marketing demand by a pre-trained language model and storing them as an initial copy set; S103, receiving preference input of artificial experts on the initial copy set, and constructing a creative incremental partial order copy pair for adjacent copies according to the preference order; S104, for each partial order copy pair, calling a to-be-trained language model to extract the corresponding sentence-level latent semantic vector, calculating the preference discriminant loss and the iterative prediction loss according to the latent semantic vector, and jointly updating the to-be-trained language model parameters to obtain a trained marketing copy model; S105, inputting actual marketing demand to the trained marketing copy model to generate a first draft copy and extract a latent semantic vector, feeding back the latent semantic vector to the marketing copy model for iterative generation until the latent semantic vector distance of adjacent two versions of the copy does not exceed a preset threshold or reaches a preset iteration number, obtaining a candidate copy set; S106, performing confidence evaluation on the candidate copy set based on the marketing copy model, selecting the target marketing copy with the highest confidence as the final output.

[0019] In some embodiments, generating a plurality of initial marketing copies based on the marketing demand by a pre-trained language model and storing them as an initial copy set includes: analyzing the marketing demand to form a model input prompt; under the set temperature parameters, sampling strategy and maximum output length, calling the pre-trained language model for multiple text sampling based on different random seeds to obtain multiple candidate marketing copies; performing duplicate detection and format standardization processing on the candidate marketing copy to obtain a deduplicated marketing copy; The deduplicated marketing copy is written into a preset storage area together with the corresponding generation probability or score to constitute an initial copy set.

[0020] Specifically, in the present embodiment, the system first receives marketing requirements containing product information, audience information and marketing targets, parses them into structured fields and splices to generate model input prompts; then, based on multiple sets of random seeds, the pre-trained language model is called multiple times under the generation control parameters such as set temperature parameters, sampling strategies and maximum output length, to obtain a corresponding number of candidate marketing copies; then, the text similarity calculation and normalization rules are used to perform duplicate detection and format uniform processing on the candidate marketing copies, and the highly repeated or format-unconformed entries are deleted to obtain the deduplicated marketing copies; finally, the deduplicated marketing copies together with their respective generation probabilities or comprehensive scores are written into a preset storage area to form an initial copy set, laying a data foundation for subsequent artificial preference sorting and model reinforcement learning.

[0021] The important technical terms involved in the present embodiment are explained as follows: Temperature parameter: a real number that controls the entropy value of the language model output distribution. When the temperature increases, the generation result is more diverse and creative, and when the temperature decreases, it tends to be more deterministic.

[0022] Sampling strategy: a truncation or rearrangement method for probability distribution, including but not limited to kernel sampling, top sampling, etc., to balance creativity and readability.

[0023] Random seed: the initial value set for the model random number generator. Different random seeds can trigger different decoding paths under the same prompt, improving the diversity of candidate copies.

[0024] Duplicate detection: based on edit distance, Tok-N intersection ratio score or sentence vector cosine similarity to deduplicate candidate copies, and the threshold can be adjusted according to the scene.

[0025] Format normalization: unify punctuation, case, line breaks and special symbols to the preset standard to ensure consistency for subsequent manual review and vectorization processing.

[0026] Generation probability or comprehensive score: the cumulative log probability of the model during the decoding process or the post-calculated language fluency, emotional intensity and other indicators, used as an initial quantitative reference for copy quality.

[0027] In one specific scenario example, take "smart fitness bracelet summer promotion" as a scenario example. The system first parses the marketing requirements, and embeds the fields such as "product name: smart fitness bracelet", "function highlights: 24-hour heart rate monitoring, real-time temperature recording", "target audience: urban white-collar", "marketing goal: improve the attention of new products and promote the conversion of the first batch of pre-sales" into a unified template to form the following model input prompt: "Please write a summer promotion copy for the smart fitness bracelet suitable for the urban white-collar group, emphasizing the 24-hour heart rate monitoring and real-time temperature recording functions, with a casual but professional and credible tone, and the word count is controlled within 50 words." The system sets the temperature parameter to 0.9, adopts core sampling, and sets the top-p value to 0.92, and fixes the maximum output length to 80 characters; under random seeds 101, 204, 317, 423 and 566, a total of 5 candidate marketing texts are generated by calling the pre-trained language model. After obtaining the text, the system first removes two approximately repeated texts with a sentence vector cosine similarity threshold of <0.85, and then unifies the Chinese and English punctuation, removes extra spaces, and ensures that the description is in simplified Chinese, finally obtains 3 de-duplicated marketing texts. For example: "Wrist health assistant, 24-hour heart rate and temperature monitoring, let summer exercise more secure!" "Urban pace fast, bracelet protection faster, real-time monitoring of heart rate and temperature, give you professional health management." "Lightweight bracelet, all-weather temperature and heart rate monitoring, accompany you to burn every step of summer!" The system also retains the generation probability of each text, and writes it into the specified database table or vector repository in the form of text-probability key-value pairs, forming the initial text set of this scenario. This set will be directly called in the subsequent manual preference sorting step, realizing seamless connection with the subsequent training process.

[0028] In some embodiments, the preference input of the artificial expert for the initial text set is received, and the adjacent texts are paired according to the preference order to construct a creative increasing partial order text pair, including: Receive the preference order information of the artificial expert for each marketing text in the initial text set, and generate a preference order list; According to the preference order list, the first marketing text and the second marketing text are paired, forming a partial order text pair, and recording the preference relationship that the first marketing text is better than the second marketing text in the partial order text pair; The pairing and recording operations are performed in a loop to build all partial order text pairs covering the preference order list; Write all partial order text pairs and corresponding preference relationships to a preset training data storage area for subsequent model training step calls.

[0029] Specifically, the embodiment focuses on constructing creative incremental partial order pairs of marketing copy. First, an initial set of marketing copy is displayed to an artificial expert through an interactive interface, and the overall ordering is completed in a drag-and-drop or numbered manner. The system receives and analyzes the expert's preference ordering information in real time, generating a unique preference order list. Then, the system extracts the first and second marketing copy in order according to the adjacent relationship of the list, automatically labels the one-way preference relationship of "the first is better than the second", and generates partial order pairs of marketing copy one by one. When the list is traversed, the system serializes all partial order pairs of marketing copy and their corresponding preference relationships into the training data storage area, providing complete data input for subsequent reinforcement learning training based on latent semantic vectors.

[0030] Important technical terms involved in the embodiment are explained as follows: Preference ordering information: The order data formed after the artificial expert performs overall or partial ordering on the initial set of marketing copy, containing absolute position and relative superiority and inferiority relationship.

[0031] Preference order list: The ordered marketing copy index sequence obtained by sorting the preference ordering information, which is the direct basis for constructing partial order pairs of marketing copy.

[0032] Partial order pair of marketing copy: The ordered binary tuple derived from the adjacent two marketing copy in the preference order list, explicitly recording the one-way preference of the former over the latter.

[0033] Preference relationship: The "better than" label added by the system to the partial order pair of marketing copy, used to guide the model to converge towards a higher creative level during training.

[0034] Training data storage area: A data structure dedicated to persisting partial order pairs of marketing copy and their preference relationships, which can be a database table, a key-value store, or a vector repository, and needs to ensure high-speed retrieval during subsequent training.

[0035] In some examples, the "smart fitness bracelet summer promotion" scenario of the preceding embodiment is continued. The initial set of marketing copy contains three de-duplicated marketing copy, which is displayed in the form of a list in the expert workstation: Number C1: "Wrist health assistant, 24-hour heart rate and body temperature at one's fingertips, let summer exercise more secure!" Number C2: "Urban rhythm is fast, bracelet protection is faster, real-time monitoring of heart rate and body temperature, giving you professional-level health management." Number C3: "Lightweight bracelet, all-weather heart rate and body temperature monitoring, accompany you to burn every step of summer!" After reading, the expert considers that C2 is the best, C1 is the second, and C3 is relatively general, and thus gives the preference order information "C2>C1>C3". The system immediately parses this information and generates a preference order list [C2, C1, C3]. According to the adjacent pairing rule, the system first pairs C2 and C1 into the first group of partial order pairs 〈C2, C1〉, and marks "C2 is better than C1"; then pairs C1 and C3 into the second group of partial order pairs 〈C1, C3〉, and marks "C1 is better than C3". After traversal, two partial order pairs and the corresponding preference relations are obtained. The system generates a unique key for each pair and writes it into the training data storage area, for example, in JSON line storage: {"pair_id":"P001","better_id":"C2","worse_id":"C1"} {"pair_id":"P002","better_id":"C1","worse_id":"C3"} After this storage process is completed, the partial order pairs can be directly called by the potential semantic reinforcement learning module for calculating the preference discrimination loss and the iterative prediction loss, thereby driving the model to continuously optimize in the creativity dimension.

[0036] In some embodiments, for each partial order pair, the to-be-trained language model is called to extract the corresponding sentence-level potential semantic vector, including: Each marketing copy in the partial order pair is input into the to-be-trained language model to obtain the word-level hidden state of the preset intermediate layer; Perform a pooling operation on the word-level hidden state to obtain the corresponding sentence-level initial semantic embedding; The sentence-level initial semantic embedding is transformed into a fixed-dimensional potential semantic vector via a trainable projection layer.

[0037] Specifically, in this embodiment, the system performs the potential semantic vector extraction process for the partial order pairs 〈C2, C1〉 and 〈C1, C3〉 one by one. First, the copy text is sent into the to-be-trained language model, and the complete word-level hidden state is intercepted at the output of the second-to-last transformation module; then the hidden state is pooled to obtain the sentence-level initial semantic embedding; finally, the sentence-level embedding is mapped to a uniform-dimensional potential semantic vector through a trainable projection layer, and the standardization cache is completed, providing input for subsequent double-loss calculation.

[0038] The important technical terms involved in this embodiment are explained as follows: Word-level hidden state: In the Transformer structure, each word will get a new vector representation after each layer of self-attention and feed-forward network. These hierarchical vectors are the word-level hidden states. The closer to the end of the network, the more abstract and complete the semantic information represented.

[0039] Pooling operation: A simplified method to compress a set of sequence-level vectors into a single vector. In this embodiment, mean pooling is used to take the arithmetic mean of all word vectors in the same text according to the dimension, thereby forming the initial sentence-level semantic embedding, ensuring that the length difference of the sentence does not affect the semantic alignment.

[0040] Initial sentence-level semantic embedding: The overall text representation obtained after pooling has a condensed representation of the whole sentence's intent and style, which is the basis for subsequent projection and distance calculation.

[0041] Trainable projection layer: The mapping layer, usually composed of one or more fully connected networks, is responsible for converting high-dimensional embeddings into a pre-set uniform dimension for direct comparison between different texts. This projection layer is updated together with the main model parameters, allowing the latent semantic space to be continuously optimized during training.

[0042] In some examples, for the partial order pair 〈C2, C1〉, the system first inputs the C2 text into the language model to be trained after word-by-word tokenization and locates it to the 11th layer output to obtain an L×H word-level hidden state matrix, where L is the number of words and H is the hidden dimension; the average of the matrix along the word dimension is taken to obtain a 1×H sentence-level initial semantic embedding v1; v2 is sent to a trainable projection layer with a dimension of H×256 to obtain a 256-dimensional latent semantic vector z2. The same process is performed on the C1 text to obtain the latent semantic vector z1. The system then writes z2, z1 into the cache and attaches the text identification and preference label for subsequent loss calculation. z1 and z3 are extracted in the same way for 〈C1, C3〉 and cached.

[0043] By pooling the word-level hidden state and uniformly mapping it through the projection layer, the system achieves cross-document semantic alignment of texts of different lengths and expressions, ensuring that the latent semantic vector retains the overall creative feature while maintaining consistent dimensions, which can be directly used for vector distance calculation and preference discrimination, significantly improving the convergence speed and creative recognition accuracy in the subsequent reinforcement learning phase.

[0044] In some embodiments, the preference discrimination loss and the iterative prediction loss are calculated based on the latent semantic vector, and are used together to update the parameters of the language model to be trained to obtain a trained marketing text model, including: Based on the latent semantic vectors in the partial order text pair, the first marketing text is output by the discriminator head to be better than the second marketing text, and compared with the artificial preference label to calculate the preference discrimination loss; input the potential semantic vector of the second marketing script and the corresponding text as a conditional input to the language model to be trained, generate a predicted marketing script and extract the potential semantic vector of the predicted marketing script; calculate an iterative prediction loss according to the vector distance between the potential semantic vector of the predicted marketing script and the potential semantic vector of the first marketing script; weight the preference discrimination loss and the iterative prediction loss according to a preset weight to obtain a total loss, and perform back propagation based on the total loss to update the parameters of the language model to be trained.

[0045] Specifically, in the present embodiment, the system performs double loss calculation using the cached potential semantic vectors z2, z1, z3 (corresponding to scripts C2, C1, C3, respectively). First, the vector pair 〈z2, z1〉, 〈z1, z3〉 is received by the discrimination head, which outputs the preference probability that "the first script is better than the second script"; then, the vector z1 and its original script C1 are input as a conditional input to the language model to be trained to generate a predicted script C and extract its potential semantic vector z; finally, the system calculates the preference discrimination loss and the iterative prediction loss, respectively, and combines the total loss according to the weight to update the model parameters.

[0046] The important technical terms involved in the present embodiment are explained as follows: Discrimination head: a lightweight binary classification subnetwork attached to the top of the language model, which only accepts a pair of potential semantic vectors as input and outputs a probability score p that the first script is better than the second script, used to measure the consistency of the model in creative ranking.

[0047] Preference discrimination loss: the consistency of the model output and the artificial preference label is measured using the cross-entropy formula log p, and the smaller the p is, the smaller the loss is.

[0048] Iterative prediction loss: the difference between the predicted script vector z and the better script vector z2 is measured using the cosine distance or Euclidean distance, and the smaller the distance is, the closer the model is to the target along the creative gradient.

[0049] Preset weight: used to balance the importance of the two types of loss, in the present embodiment, λ1=0.6 is assigned to the discrimination loss with a higher weight, and λ2=0.4 is assigned to the iterative loss with a lower weight.

[0050] Back propagation: the gradient of the model parameters is calculated according to the total loss and gradient descent update is performed, so that the model gradually improves the ranking consistency and iterative improvement ability.

[0051] For example, in some examples, the system reads the partial order pair 〈C2, C1〉, inputs z2, z1 into the discriminator head to obtain the probability p = 0.88, and the cross-entropy loss L ce ≈ 0.13 after comparing with the human label "C2 is better than C1". Then the model decodes the z1 and the script C1 to generate the predicted script C: "Wrist assistant guards the pace, heart rate and body temperature double monitoring helps you cool down and burn fat", and extracts the latent vector z; the cosine distance d ≈ 0.27 between z and z2 is calculated, and the iterative prediction loss L iter ≈ 0.27 is obtained. The system weights the total loss L total = λ1L ce + λ2L iter ≈ 0.19, and performs one backward propagation and weight update on all trainable parameters of the model. Repeat the above process for the partial order pair 〈C1, C3〉 and accumulate the gradient to form a complete training batch. After multiple rounds of batch iteration, the model gradually forms a directional representation of "moving from a suboptimal to a more optimal script" in the sentence-level latent semantic space.

[0052] By simultaneously minimizing the preference discrimination loss and the iterative prediction loss, the embodiment enables the language model to not only learn to accurately determine the creativity level, but also to generate more creative rewritten texts along the latent semantic gradient, significantly improving the consistency of the model's creativity ranking and the convergence speed of iterative optimization, laying a solid foundation for fast iteration of high-quality scripts in the subsequent reasoning stage.

[0053] In some embodiments, the latent semantic vector is fed back to the marketing script model for iterative generation until the distance between the latent semantic vectors of the adjacent two versions of the script does not exceed a preset threshold or reaches a preset number of iterations, to obtain a candidate script set, including: The latent semantic vector of the first version of the draft script and the actual marketing demand are jointly input as additional context into the marketing script model to generate a second version of the marketing script; The latent semantic vector of the second version of the marketing script is extracted, and the vector distance between the latent semantic vector and the latent semantic vector of the previous version of the marketing script is calculated; When the vector distance is greater than the preset threshold and the number of iterations performed is less than the preset number of iterations, the second version of the marketing script and the latent semantic vector are taken as new previous version data, and the foregoing generation and distance calculation steps are repeated; When the vector distance is not greater than the preset threshold or the number of iterations performed reaches the preset number of iterations, the marketing scripts and the corresponding latent semantic vectors in the iterative process are collected and stored as a candidate script set.

[0054] Specifically, the embodiment describes an automatic iteration mechanism in the deployment stage. The system first obtains a first draft and its potential semantic vector, and inputs the vector and the structured marketing demand as additional context into the marketing copy model to generate a second version. Then, the system extracts the potential semantic vector of the second version and calculates the distance with the previous version. If the distance is higher than the threshold and the maximum iteration number is not reached, the second version is considered as the new previous version for further iteration. When any termination condition is triggered, the system collects all historical versions along with the vectors to form a candidate copy set, which provides input for subsequent confidence evaluation.

[0055] The important technical terms involved in the embodiment are explained as follows: Iteration number: refers to the count of consecutive calls to the model to rewrite the draft. The embodiment sets an upper limit of 4 to prevent the generation process from being too long.

[0056] Additional context: adds the potential semantic vector label of the previous version after the original marketing demand prompt, prompting the model to rewrite while maintaining the same scenario and moving towards a new semantic target.

[0057] Vector distance threshold: a numerical boundary to measure the creative difference between adjacent versions. A cosine distance of 0.20 is used as the stopping criterion. If the distance is less than or equal to the threshold, it is determined that the rewriting is converging.

[0058] Candidate copy set: a dataset composed of all version copies and their respective potential semantic vectors generated during the iteration process, which is the direct input for confidence evaluation and final draft screening.

[0059] For example, in a specific example, in the "smart fitness bracelet summer promotion" scenario, the user allows a maximum of 4 iterations and sets the convergence threshold to 0.20. The system first generates the first draft V0: "Wrist helper guards heart rate, summer fat-burning more secure". After extracting the potential semantic vector z0, it is input into the model along with the marketing demand field as additional context to obtain the second version V1: "Lightweight bracelet, 24-hour heart rate and body temperature monitoring, accompany you to enjoy a summer". Extract z1 and calculate d(z0, z1)=0.34>0.20, iteration count is 1<4, continue the loop. With z1 and V1 as new context, generate the third version V2: "Urban rhythm fast, bracelet real-time guardian, heart rate and body temperature at your fingertips", the distance d(z1, z2)=0.26 is still greater than the threshold, and the iteration count is 2<4. Continue to generate the fourth version V3: "Bracelet on wrist, 24-hour heart rate and body temperature dual monitoring, let every step burn fat more efficiently", d(z2, z3)=0.18≤0.20, meet the convergence condition, the loop ends. The system serializes V0 to V3 and their respective vectors to form a candidate copy set.

[0060] By quantifying the potential semantic vector distance in real time after each round of generation and comparing it with the threshold value, the embodiment can automatically converge while ensuring the progress of creativity, avoiding resource waste caused by invalid iterations, and providing multiple versions of alternatives for subsequent confidence screening, significantly improving the creativity level and generation efficiency of the final marketing copy.

[0061] In some embodiments, the candidate copy set is subjected to confidence evaluation based on the marketing copy model, and the target marketing copy with the highest confidence is selected, including: The potential semantic vector of each marketing copy in the candidate copy set and the potential semantic vector of the actual marketing demand are input into the confidence evaluation submodule of the marketing copy model to generate a corresponding confidence score; The confidence scores are compared to determine the marketing copy with the highest confidence score as the target marketing copy.

[0062] Specifically, the embodiment describes the confidence evaluation mechanism in the candidate copy screening stage. The system first extracts the potential semantic vector q of the actual marketing demand and inputs it together with the potential semantic vector zi of each marketing copy in the candidate copy set into the confidence evaluation submodule; the submodule generates a confidence score si based on vector similarity, language fluency, and emotional fit; the system then compares all si, selects the marketing copy with the highest confidence as the target marketing copy, and outputs it.

[0063] The important technical terms involved in the embodiment are explained as follows: Confidence evaluation submodule: an index fusion component attached to the marketing copy model, which uses a multilayer perceptron structure to receive (q, zi) vector pairs and outputs a confidence score in the 0-1 interval to measure the matching degree and communicability of the copy to the demand.

[0064] Similarity branch: based on cosine similarity to measure the fit of the copy and the demand in the potential semantic space, which is the core contribution to confidence evaluation.

[0065] Language fluency branch: uses the language model's perplexity or syntax score to evaluate the smoothness of the copy to ensure easy readability.

[0066] Emotional fit branch: detects the consistency of the copy's emotional color and the expected marketing tone (such as motivation, warmth, professionalism) through a fine-grained sentiment classifier.

[0067] Integrated weight: the scores of the three branches are weighted and summed according to the empirical weights α, β, and γ to form the final confidence score si, and the sum of the weights is 1.

[0068] For example, in some examples, continuing the candidate text set {V0, V1, V2, V4} in the preceding embodiment, the system first performs homologous coding on the marketing demand prompt to obtain a potential semantic vector q. Then the potential semantic vector zi of each version of the text is read one by one and input into the confidence assessment submodule together with q: The cosine similarity of V0 is 0.76, the fluency score is 0.79, the emotional fit score is 0.74, and the confidence s0 weighted by a=0.5, b=0.3, and g=0.2 is approximately 0.76; The similarity of V1 is 0.82, the fluency is 0.83, the emotional fit is 0.80, and the confidence s1 is approximately 0.82; The similarity of V2 is 0.78, the fluency is 0.81, the emotional fit is 0.86, and the confidence s2 is approximately 0.80; The similarity of V3 is 0.85, the fluency is 0.86, the emotional fit is 0.84, and the confidence s3 is approximately 0.85.

[0069] The system compares s0 to s3 and finds that s3 is the highest, so it determines that V3: "Wristband, 24-hour heart rate and body temperature monitoring, make every step burn fat more efficiently" is the target marketing text and pushes it to the user.

[0070] By fusing the similarity, language fluency, and emotional fit in the potential semantic space through the confidence assessment submodule, this embodiment can quickly identify the text that best matches the demand and has the best reading experience among multiple versions of candidate text, significantly improving the marketing effect and user satisfaction of the final output.

[0071] The present application adjusts the marketing model through the marketing model fine-tuning method of "creative incremental partial order data construction + potential semantic optimization based on reinforcement learning + text iterative generation based on potential semantic vector". By introducing sentence-level potential semantic alignment, the language model can more efficiently master the text creation skills of creative progressive iteration improvement. This solution not only better adapts to the quantitative evaluation characteristics of creativity, but also optimizes the model's progressive improvement ability based on potential semantics during training, making it more suitable for the actual marketing text generation scenario. At the same time, this approach also allows users to supervise the model's real-time adjustment of its text iteration method during reasoning through timely feedback to meet the user's text production needs.

[0072] The following is an apparatus embodiment of the present application, which can be used to execute the method embodiments of the present application. For details not disclosed in the apparatus embodiments of the present application, please refer to the method embodiments of the present application.

[0073] Figure 2 is a structural schematic diagram of the marketing text iterative generation apparatus based on potential semantic optimization provided by an embodiment of the present application. As shown in Figure 2As shown, the marketing copy iterative generation device based on latent semantic optimization includes: The acquisition module 201 is configured to acquire marketing requirements containing product information, audience information, and marketing targets. The storage module 202 is configured to generate a plurality of initial marketing copies based on the marketing requirements from a pre-trained language model and store the initial marketing copies as an initial copy set. The construction module 203 is configured to receive preference inputs of artificial experts on the initial copy set, and construct a partial order copy pair with increasing creativity according to the adjacent copies in the preference order. The training module 204 is configured to call a to-be-trained language model to extract a corresponding sentence-level latent semantic vector for each partial order copy pair, calculate a preference discriminant loss and an iterative prediction loss according to the latent semantic vector, and jointly update the to-be-trained language model parameters to obtain a trained marketing copy model. The generation module 205 is configured to input actual marketing requirements to the trained marketing copy model, generate a first draft copy, extract a latent semantic vector, and feed the latent semantic vector back to the marketing copy model for iterative generation until the latent semantic vector distance between adjacent two versions of the copy does not exceed a preset threshold or reaches a preset iteration number, to obtain a candidate copy set. The output module 206 is configured to perform confidence evaluation on the candidate copy set based on the marketing copy model, select a target marketing copy with the highest confidence as the final output.

[0074] In some embodiments, Figure 2 The storage module 202 of the marketing copy iterative generation device based on latent semantic optimization parses a model input prompt from the marketing requirements; under a set temperature parameter, sampling strategy, and maximum output length, the pre-trained language model is called based on different random seeds to perform multiple text samplings, and a plurality of candidate marketing copies are obtained; the candidate marketing copies are subjected to duplication detection and format standardization processing to obtain de-duplicated marketing copies; and the de-duplicated marketing copies are written into a preset storage area together with corresponding generation probabilities or scores to constitute an initial copy set.

[0075] In some embodiments, Figure 2 The construction module 203 of the marketing copy iterative generation device based on latent semantic optimization receives preference order information of artificial experts on each marketing copy in the initial copy set, generates a preference order list, pairs a first marketing copy and a second marketing copy adjacent in ranking according to the preference order list to form a partial order copy pair, and records a preference relationship that the first marketing copy is better than the second marketing copy in the partial order copy pair; the pairing and recording operations are cyclically performed to construct all partial order copy pairs covering the preference order list; and all partial order copy pairs and the corresponding preference relationships are written into a preset training data storage area for subsequent model training steps.

[0076] In some embodiments, Figure 2The training module 204 inputs each marketing script in the partial order script pair into the language model to be trained to obtain the wordpiece-level hidden state of the preset intermediate layer; performs a pooling operation on the wordpiece-level hidden state to obtain the initial sentence-level semantic embedding; and transforms the initial sentence-level semantic embedding into a fixed-dimension latent semantic vector via a trainable projection layer.

[0077] In some embodiments, Figure 2 The training module 204 inputs each marketing script in the partial order script pair into the language model to be trained to obtain the wordpiece-level hidden state of the preset intermediate layer; performs a pooling operation on the wordpiece-level hidden state to obtain the initial sentence-level semantic embedding; and transforms the initial sentence-level semantic embedding into a fixed-dimension latent semantic vector via a trainable projection layer.

[0078] In some embodiments, Figure 2 The generation module 205 inputs the latent semantic vector of the first draft script and the actual marketing demand as additional context into the marketing script model to generate a second version of the marketing script; extracts the latent semantic vector of the second version of the marketing script and calculates the vector distance between the latent semantic vector and the latent semantic vector of the previous version of the marketing script; when the vector distance is greater than a preset threshold and the number of iterations performed is less than a preset number of iterations, takes the second version of the marketing script and the latent semantic vector as new previous version of data, and repeats the foregoing generation and distance calculation steps; when the vector distance is not greater than the preset threshold or the number of iterations performed reaches the preset number of iterations, collects each version of the marketing script and the corresponding latent semantic vector in the iteration process, and stores them as a candidate script set.

[0079] In some embodiments, Figure 2 The output module 206 inputs the latent semantic vector of each marketing script in the candidate script set and the latent semantic vector of the actual marketing demand into the confidence evaluation submodule of the marketing script model to generate a corresponding confidence score; compares the confidence scores to determine the marketing script with the highest confidence score as the target marketing script.

[0080] It should be understood that the size of the serial number of each step in the above embodiments does not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0081] Figure 3 is a structural schematic diagram of an electronic device 3 provided by the embodiments of the present application. As shown inFigure 3 As shown, the electronic device 3 of this embodiment includes a processor 301, a memory 302, and a computer program 303 stored in the memory 302 and executable on the processor 301. The processor 301 implements the steps in each of the above method embodiments when executing the computer program 303. Alternatively, the processor 301 implements the functions of each module / unit in each of the above apparatus embodiments when executing the computer program 303.

[0082] By way of example, the computer program 303 can be divided into one or more modules / units, which are stored in the memory 302 and executed by the processor 301 to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program 303 in the electronic device 3.

[0083] The electronic device 3 can be a desktop computer, a notebook computer, a palm computer, a cloud server, and the like. The electronic device 3 can include but is not limited to the processor 301 and the memory 302. Those skilled in the art can understand that the electronic device 3 can include more or fewer components, or combine certain components, or different components, for example, the electronic device can also include an input / output device, a network access device, a bus, and the like. Figure 3 The electronic device 3 is merely an example and does not constitute a limitation on the electronic device 3, which can include more or fewer components than those shown, or combine certain components, or different components, for example, the electronic device can also include an input / output device, a network access device, a bus, and the like.

[0084] The processor 301 can be a central processing unit (CPU), or other general purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic components, discrete hardware components, etc. The general purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0085] The memory 302 can be an internal storage unit of the electronic device 3, for example, a hard disk or a memory of the electronic device 3. The memory 302 can also be an external storage device of the electronic device 3, for example, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the electronic device 3. Further, the memory 302 can include both the internal storage unit and the external storage device of the electronic device 3. The memory 302 is used to store computer programs and other programs and data required by the electronic device. The memory 302 can also be used to temporarily store data that has been output or will be output.

[0086] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of each functional unit and module is exemplified, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be realized in the form of hardware or software function unit. In addition, the specific name of each functional unit and module is only for easy distinction, and does not limit the protection scope of the present application. The specific working process of the unit and module in the above system can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0087] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described or recorded in detail in a certain embodiment can refer to the relevant description of other embodiments.

[0088] Those of ordinary skill in the art can appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0089] In the embodiments of the present application, it should be understood that the disclosed apparatus / computer device and method can be implemented in other manners. For example, the described apparatus / computer device embodiments are merely schematic. For example, the division of the modules or units can be different, and each can include multiple sub-modules or units. Some or all of the modules or units can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0090] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0091] In addition, each functional unit in the various embodiments of the present application can be integrated into a processing unit, or each unit can be a physically independent unit, or two or more units can be integrated into a unit. The integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0092] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, all or part of the flow of the above-mentioned embodiment methods can be implemented by a computer program instructing related hardware to complete, and the computer program can be stored in a computer readable storage medium. When the processor executes the computer program, the steps of each method embodiment described above can be implemented. The computer program can include computer program code, which can be in the form of source code, object code, executable file or some intermediate form. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0093] The above examples are only used to illustrate the technical solutions of the present application, but not limit the same; although the technical solutions of the present application are described in detail with reference to the foregoing examples, those skilled in the art should understand that the technical solutions recorded in the foregoing examples can still be modified, or some technical features thereof can be replaced by equivalent ones; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for iteratively generating marketing copy based on latent semantic optimization, characterized in that: include: Obtain marketing needs including product information, audience information and marketing goals; Based on the marketing needs, a pre-trained language model generates a number of initial marketing texts and stores them as an initial text set; receiving preference input from human experts on the initial copy set, and constructing partially ordered copy pairs with increasing creativity for adjacent copies according to the preference order; For each partially ordered copy pair, the language model to be trained is called to extract the corresponding sentence-level latent semantic vector. The preference discrimination loss and iterative prediction loss are calculated based on the latent semantic vector, and are used together to update the parameters of the language model to be trained to obtain a trained marketing copy model. Input actual marketing needs into the trained marketing copy model to generate a first draft copy and extract latent semantic vectors. The latent semantic vectors are fed back into the marketing copy model for iterative generation until the distance between the latent semantic vectors of two adjacent versions of the copy does not exceed a preset threshold or reaches a preset number of iterations, thereby obtaining a set of candidate copies. A confidence evaluation is performed on the candidate copy set based on the marketing copy model, and a target marketing copy with the highest confidence is selected, and the target marketing copy is used as the final output.

2. The method according to claim 1, characterized in that The generating of a plurality of initial marketing texts by the pre-trained language model based on the marketing demand and storing the generated and stored initial texts as an initial text set includes: Analyze the marketing needs to form model input prompts; Under the set temperature parameters, sampling strategy and maximum output length, the pre-trained language model is called based on different random seeds to perform multiple text samplings to obtain multiple candidate marketing copywriting; Perform duplication detection and format standardization on the candidate marketing copy to obtain a deduplicated marketing copy; The deduplicated marketing copy together with the corresponding generation probability or score is written into a preset storage area to form the initial copy set.

3. The method according to claim 1, characterized in that The receiving of the preference input of the manual expert on the initial copy set and constructing partial order copy pairs of increasing creativity for adjacent copies according to the preference order includes: receiving preference ranking information of each marketing copy in the initial copy set from a human expert, and generating a preference order list; According to the preference order list, pairing the first marketing copy and the second marketing copy that are adjacent in ranking to form a partial order copy pair, and recording a preference relationship that the first marketing copy is better than the second marketing copy in the partial order copy pair; cyclically performing pairing and recording operations to construct all partial order copy pairs covering the preference order list; All the partially ordered text pairs and the corresponding preference relationships are written into a preset training data storage area for subsequent model training steps to call.

4. The method according to claim 1, wherein For each partially ordered text pair, calling the language model to be trained to extract the corresponding sentence-level latent semantic vector includes: Input each marketing copy in the partially ordered copy pair into the language model to be trained to obtain a word-level hidden state of a preset intermediate layer; Performing a pooling operation on the word-level hidden state to obtain the corresponding sentence-level initial semantic embedding; The sentence-level initial semantic embedding is transformed into a latent semantic vector of fixed dimension via a trainable projection layer.

5. The method according to claim 4, characterized in that The calculation of the preference discrimination loss and the iterative prediction loss based on the latent semantic vector and the use of both for updating the parameters of the language model to be trained to obtain a trained marketing copy model includes: Based on the latent semantic vectors in the partially ordered copy pairs, the discriminant head outputs the preference probability that the first marketing copy is better than the second marketing copy, and compares it with the manual preference labels to calculate the preference discrimination loss; Using the latent semantic vector and corresponding text of the second marketing copy as conditions, inputting the language model to be trained, generating a predicted marketing copy, and extracting the latent semantic vector of the predicted marketing copy; Calculating an iterative prediction loss based on a vector distance between the latent semantic vector of the predicted marketing copy and the latent semantic vector of the first marketing copy; The preference discrimination loss and the iterative prediction loss are weightedly summed according to preset weights to obtain a total loss, and back propagation is performed based on the total loss to update the parameters of the language model to be trained.

6. The method according to claim 1, characterized in that Feeding the latent semantic vector back to the marketing copy model for iterative generation until the distance between the latent semantic vectors of two adjacent versions of the copy does not exceed a preset threshold or reaches a preset number of iterations, thereby obtaining a candidate copy set, including: Inputting the latent semantic vector of the first draft copy and the actual marketing demand as additional context into the marketing copy model to generate a second version of the marketing copy; Extracting a latent semantic vector from the second version of the marketing copy, and calculating a vector distance between the latent semantic vector and the latent semantic vector of the previous version of the marketing copy; When the vector distance is greater than the preset threshold and the number of iterations executed is less than the preset number of iterations, the second version of the marketing copy and the latent semantic vector are used as the new previous version of data, and the aforementioned generation and distance calculation steps are repeated; When the vector distance is not greater than the preset threshold or the number of executed iterations reaches the preset number of iterations, each version of the marketing copy and the corresponding latent semantic vector in the iteration process are collected and stored as the candidate copy set.

7. The method according to claim 1, characterized in that The step of performing confidence evaluation on the candidate copy set based on the marketing copy model and selecting the target marketing copy with the highest confidence includes: Inputting the latent semantic vector of each marketing copy in the candidate copy set and the latent semantic vector of the actual marketing demand into the confidence evaluation submodule of the marketing copy model to generate a corresponding confidence score; The confidence scores are compared to determine the marketing copy with the highest confidence score as the target marketing copy.

8. A device for iteratively generating marketing copy based on latent semantic optimization, characterized in that: include: The acquisition module is used to obtain marketing needs including product information, audience information and marketing goals; A storage module, configured to generate a plurality of initial marketing texts based on the marketing needs using a pre-trained language model and store the generated texts as an initial text set; A construction module is configured to receive the preference input of the human expert on the initial copy set, and construct partial order copy pairs with increasing creativity for adjacent copies according to the preference order; A training module is configured to, for each partially ordered copy pair, call the language model to be trained to extract the corresponding sentence-level latent semantic vector, calculate the preference discrimination loss and iterative prediction loss based on the latent semantic vector, and use these together to update the parameters of the language model to be trained to obtain a trained marketing copy model; A generation module is used to input actual marketing needs into the trained marketing copy model, generate a first draft copy, extract latent semantic vectors, and feed the latent semantic vectors back to the marketing copy model for iterative generation until the distance between the latent semantic vectors of two adjacent versions of the copy does not exceed a preset threshold or reaches a preset number of iterations, thereby obtaining a set of candidate copies; The output module is used to perform confidence evaluation on the candidate copy set based on the marketing copy model, select the target marketing copy with the highest confidence, and use the target marketing copy as the final output.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images