Information generation method and device, equipment, medium and program product

Abstract

The embodiment of the invention provides an information generation method and device, equipment, a medium and a program product, and is applied to the technical field of information processing. The method comprises the following steps: acquiring demand information of a service; according to the demand information, a first service scheme is determined from preset service schemes, and the first service scheme is a service scheme which has a matching degree with the demand information greater than or equal to a preset matching degree in the preset service schemes; generating prompt information corresponding to the demand information according to the first service scheme; and generating a second service scheme corresponding to the demand information according to the prompt information. According to the information generation method provided by the embodiment of the invention, the whole process from obtaining the demand information to generating the second service scheme is free from manual intervention operation, the cost and the error rate of manually compiling the service scheme are remarkably reduced, the generation efficiency and the quality of the service scheme are improved, and the problems of manually compiling the service scheme and generating the second service scheme are effectively solved. And the quality of the service scheme is poor.

Description

Technical Field

[0001] The present invention belongs to the field of information processing technology, and in particular relates to an information generation method, device, equipment, medium and program product. Background Art

[0002] With the development of Internet technology, servers have provided users with a variety of services. However, service recommendations, i.e., service plans, are often written manually, and some writers have relatively limited experience, resulting in poor quality of service plans. Summary of the Invention

[0003] The embodiments of the present invention provide an information generation method, apparatus, device, medium, and program product, which can solve the problem of poor quality of service plans caused by manual writing of service plans.

[0004] In a first aspect, an embodiment of the present invention provides an information generation method, the information generation method comprising:

[0005] Obtain service demand information;

[0006] Determine a first service plan from the preset service plans based on the demand information, where the first service plan is a service plan among the preset service plans whose matching degree with the demand information is greater than or equal to the preset matching degree;

[0007] generating prompt information corresponding to the demand information according to the first service plan;

[0008] A second service solution corresponding to the demand information is generated according to the prompt information, and the second service solution includes service recommendation information generated by the prompt information.

[0009] In some possible implementations of the embodiments of the present application, the matching degree includes a similarity value, the preset matching degree includes a preset similarity threshold, the preset service solution includes a plurality of reference service solutions, and each of the plurality of reference service solutions includes reference prompt information and reference service recommendation information;

[0010] According to the demand information, a first service plan is determined from the preset service plans, including:

[0011] Calculating a similarity value between each reference prompt information and the demand information based on the semantic features of each reference prompt information and the semantic features of the demand information;

[0012] Filtering target prompt information from multiple reference prompt information based on the similarity value between each reference prompt information and the requirement information, wherein the similarity value between the target prompt information and the requirement information is greater than a preset similarity threshold;

[0013] A first service plan is determined according to the reference service plan corresponding to the target prompt information.

[0014] In some possible implementations of the embodiments of the present application, determining the first service solution according to the reference service solution corresponding to the target prompt information includes:

[0015] When the number of reference service solutions corresponding to the target prompt information is at least two, obtaining a recommendation evaluation index value for each of the at least two reference service solutions, the recommendation evaluation index value being used to reflect the quality of a reference service corresponding to the reference service solution recommended based on the reference service solution;

[0016] A first service solution is selected from at least two reference service solutions according to the recommended evaluation index value of each reference service solution, wherein the recommended evaluation index value of the first service solution is the maximum value of the recommended evaluation index values ​​of the at least two reference service solutions.

[0017] In some possible implementations of the embodiments of the present application, generating prompt information corresponding to the demand information according to the first service solution includes:

[0018] Extract keywords from content related to the prompt information in the first service solution according to the reference prompt information structure corresponding to the demand information, and generate first reference prompt information corresponding to the first service solution;

[0019] Extract keywords from the demand information according to the reference prompt information structure to generate second reference prompt information corresponding to the demand information;

[0020] According to the first reference prompt information, the second reference prompt information is adjusted to obtain prompt information corresponding to the demand information.

[0021] In some possible implementations of the embodiments of the present application, the reference prompt information structure includes product information items and service recommendation group feature information items;

[0022] The first reference prompt information includes reference product information corresponding to the product information item in the reference prompt information of the first service solution and reference service recommendation group characteristic information corresponding to the service recommendation group characteristic information item in the reference prompt information of the first service solution;

[0023] The second reference prompt information includes product information corresponding to the product information item in the demand information and service recommendation group characteristic information corresponding to the service recommendation group characteristic information item in the demand information.

[0024] In some possible implementations of the embodiments of the present application, generating a second service solution corresponding to the demand information according to the prompt information includes:

[0025] Processing the prompt information corresponding to the demand information through the service solution generation model to obtain a second service solution;

[0026] The service solution generation model is obtained through training based on training samples, and the training samples include sample prompt information and sample service solutions corresponding to the sample prompt information.

[0027] In some possible implementations of the embodiments of the present application, the second service solution further includes optimized prompt information generated from the prompt information, and the service solution generation model includes a semantic optimization model and an information expansion model;

[0028] The service solution generation model processes the prompt information corresponding to the demand information to obtain a second service solution, including:

[0029] Perform semantic optimization processing on the prompt information corresponding to the demand information through the semantic optimization model to obtain optimized prompt information;

[0030] The optimization prompt information is expanded through the information expansion model to obtain the initial service recommendation information;

[0031] The initial service recommendation information is semantically optimized through the semantic optimization model to obtain service recommendation information.

[0032] In a second aspect, an embodiment of the present application provides an information generating device, the information generating device comprising:

[0033] Acquisition module, used to obtain service demand information;

[0034] A first determining module is configured to determine a first service solution from the preset service solutions based on the demand information, the first service solution being a service solution in the preset service solutions whose matching degree with the demand information is greater than or equal to a preset matching degree;

[0035] A first generating module, configured to generate prompt information corresponding to the demand information according to the first service solution;

[0036] The second generating module is configured to generate a second service solution corresponding to the demand information according to the prompt information, wherein the second service solution includes service recommendation information generated by the prompt information.

[0037] In a third aspect, an embodiment of the present application provides an electronic device comprising: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the information generation method as described in any one of the first aspects is implemented.

[0038] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having computer program instructions stored thereon. When the computer program instructions are executed by a processor, an information generation method as described in any one of the first aspects is implemented.

[0039] In a fifth aspect, an embodiment of the present application provides a computer program product, which includes a computer program or instructions, and when the computer program or instructions are executed by a processor, it implements the information generation method as any one of the first aspects.

[0040] The information generation method, device, equipment, medium and program product of the embodiments of the present invention can automatically filter out a first service plan from preset service plans based on the service demand information, whose degree of matching with the demand information is greater than or equal to the preset matching degree. After determining the first service plan, prompt information corresponding to the demand information is further generated based on the first service plan. The prompt information can provide a reference basis for the generation of the second service plan and optimize the generation of the second service plan, effectively avoiding problems caused by omissions or deviations in the understanding of the demand information, and ensuring the accuracy and effectiveness of the second service plan. In this way, full automation is achieved from the acquisition of demand information to the generation of the second service plan, and the entire process does not require human intervention. This method of automatically generating the second service plan significantly reduces the cost required for manually writing service plans, while reducing the error rate caused by human factors, significantly improving the generation efficiency and quality of service plans, and effectively solving the problem of poor quality of service plans caused by manually writing service plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0042] Figure 1 A schematic diagram showing a flow chart of an information generation method provided by some embodiments of the present application;

[0043] Figure 2 A flowchart illustrating a specific implementation of step 120 provided in some embodiments of the present application is shown;

[0044] Figure 3 A flowchart illustrating a specific implementation of step 1203 provided in some embodiments of the present application is shown;

[0045] Figure 4 A flowchart illustrating a specific implementation of step 130 provided in some embodiments of the present application is shown;

[0046] Figure 5 A flowchart illustrating a specific implementation of step 140 provided in some embodiments of the present application is shown;

[0047] Figure 6A schematic diagram showing the structure of an information generating device provided in some embodiments of the present application is shown;

[0048] Figure 7 A schematic structural diagram of an electronic device provided in some embodiments of the present application is shown. DETAILED DESCRIPTION

[0049] The features and exemplary embodiments of various aspects of the present invention will be described in detail below. In order to make the objects, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can be implemented without the need for some of these specific details. The following description of the embodiments is merely intended to provide a better understanding of the present invention by illustrating examples of the present invention.

[0050] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, the elements defined by the phrase "comprising..." do not exclude the presence of other identical elements in the process, method, article, or device comprising the elements.

[0051] It should be noted that the acquisition, storage, use and processing of data in the embodiments of this application are in compliance with the relevant provisions of national laws and regulations.

[0052] It should be noted that in the embodiments of the present application, certain software, components, models and other existing solutions in the industry may be mentioned. They should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0053] In order to solve the problems in the aforementioned related technologies, embodiments of the present invention provide an information generation method, apparatus, device, medium, and program product.

[0054] The following is combined with Figure 1 To the attached Figure 6 , the information generation method provided in the embodiment of the present application is described in detail through specific embodiments and their application scenarios.

[0055] Figure 1 Schematic diagram showing the flow of information generation methods provided by some embodiments of the present application. Figure 1 As shown, the information generation method is applied to an electronic device and may specifically include steps 110 to 140.

[0056] Step 110: Obtain service demand information.

[0057] Step 120 : Determine a first service solution from the preset service solutions based on the demand information. The first service solution is a service solution in the preset service solutions whose matching degree with the demand information is greater than or equal to the preset matching degree.

[0058] Step 130: Generate prompt information corresponding to the demand information according to the first service solution.

[0059] Step 140: Generate a second service plan corresponding to the demand information according to the prompt information, where the second service plan includes service recommendation information generated by the prompt information.

[0060] In this way, based on the service demand information, the first service plan with a degree of matching greater than or equal to the preset degree of matching with the demand information can be automatically screened out from the preset service plans. After determining the first service plan, prompt information corresponding to the demand information is further generated based on the first service plan. This prompt information can provide a reference basis for the generation of the second service plan, and optimize the generation of the second service plan, effectively avoiding problems caused by omissions or deviations in the understanding of the demand information, and ensuring the accuracy and effectiveness of the second service plan. In this way, full automation is achieved from the acquisition of demand information to the generation of the second service plan, and the entire process does not require human intervention. This method of automatically generating the second service plan significantly reduces the cost required for manually writing service plans, while reducing the error rate caused by human factors, significantly improving the generation efficiency and quality of service plans, and effectively solving the problem of poor quality of service plans caused by manually writing service plans.

[0061] First, involving step 110, the demand information for the service in the embodiment of the present application refers to the product information of the product involved in the service, the service recommendation group characteristic information of the product, and the service solution generation request information for the service, and the service solution generation request information is used to request the electronic device to generate a service solution corresponding to the service based on the product information of the product involved in the service and the service recommendation group characteristic information of the product. Among them, the product information may include product basic attribute information, product functional characteristic information and product usage attribute information. Specifically, the product basic attribute information includes but is not limited to product specification parameters, product technical performance indicators, and product configuration parameters; the product usage attribute information includes but is not limited to product operating instructions and product application scenario descriptions. The service recommendation group characteristic information may include demographic characteristic information, behavioral characteristic information and demand characteristic information. Specifically. Demographic characteristic information includes but is not limited to age, gender, occupation, income level, usage area, and city level; behavioral characteristic information includes but is not limited to usage habits, consumption preferences, and decision-making patterns; demand characteristic information includes but is not limited to functional requirements and service requirements.

[0062] For example, the original demand information can be obtained from channels such as service instructions, user demand documents, historical service records, real-time conversation logs, etc., and then natural language processing technology can be used to clean and structure the original demand information, extract product information and product service recommendation group characteristic information, and obtain service demand information.

[0063] Secondly, regarding step 120, the preset service plan in the embodiment of the present application includes multiple reference service plans, which are multiple historical service plans that have been pre-established and put into use. Each reference service plan in the multiple reference service plans includes reference prompt information and reference service recommendation information. The reference prompt information may include reference product information corresponding to the product information in the above step 110 and reference service recommendation group characteristic information corresponding to the service recommendation group characteristic information in the above step 110. The reference service recommendation information may include service recommendation information generated based on the reference prompt information. Business personnel can make service recommendations based on the service recommendation information, which may specifically include service recommendation text.

[0064] For example, the context of historical service plans can be modeled using encoders and pre-trained language representation models (Bidirectional Encoder Representation from Transformers, BERT) in deep learning, generating context-related word vectors, and converting historical service plans into vector representations for storage. Compared to structured data or semantic network forms, vector representations can map the text data of historical service plans into a high-dimensional space, saving computing resources while better capturing the semantic information of historical service plans on a global scale, so as to better quantify the degree of match between preset service plans and demand information.

[0065] Next, we will explain the above BERT model in detail.

[0066] The BERT model is a pre-trained model based on the self-attention mechanism. It generates vector representations of text-based reference solutions in two stages: pre-training and fine-tuning. The BERT model is pre-trained on a large-scale corpus, deeply learning the diverse representations and rich semantic information of language.

[0067] First, the BERT model encodes the input text data. This encoding process breaks the text into words and assigns unique identifiers to each word, helping the BERT model parse the different structures within the text. This processing enables the BERT model to capture the connections between words in the text and further learn deeper linguistic patterns.

[0068] Secondly, to more accurately understand the semantics of text, the BERT model introduces a positional embedding mechanism. Positional embedding is an encoding method used to represent the specific position of a word in a text. This encoding method helps the BERT model capture the relative order between words. By combining positional embedding with word embedding, the BERT model can more accurately understand the semantic information in a sentence.

[0069] The BERT model then generates semantic representations of text using multiple Transformer layers composed of a self-attention mechanism and a feedforward neural network. The self-attention mechanism enables the model to flexibly focus on information at different positions when processing sequential data; the feedforward neural network extracts features and performs complex nonlinear transformations. The stacking of these Transformer layers greatly enhances the BERT model's ability to capture semantic information, enabling it to generate highly expressive text representations.

[0070] The BERT model then averages or concatenates the vector representations of all words in the text to obtain a vector representation of the entire text, i.e., the reference service solution.

[0071] In the embodiments of the present application, the degree of match is used to measure the compatibility between the demand information and the preset service solution. The degree of match may include a similarity value. Correspondingly, the preset degree of match may include a preset similarity threshold. The preset degree of match may be a standard value set by the user based on business experience and market research. Only when the degree of match between the preset service solution and the demand information reaches or exceeds the preset similarity threshold is the preset service solution determined as the first service solution.

[0072] In some embodiments of the present application, in order to effectively avoid the influence of interference information in the reference service recommendation information on the matching degree and improve the accuracy of determining the first service solution, such as Figure 2 As shown, the above step 120 may specifically include steps 1201 to 1203.

[0073] Step 1201 : Calculate the similarity value between each piece of reference prompt information and the requirement information based on the semantic features of each piece of reference prompt information and the semantic features of the requirement information.

[0074] The semantic feature of the reference prompt information refers to the semantic vector of the reference prompt information, and the semantic feature of the demand information refers to the semantic vector of the demand information.

[0075] For example, the similarity between the reference prompt information and the requirement information can be evaluated by the cosine similarity between the reference prompt information and the requirement information, that is, the similarity value. similarity Specifically, it can be expressed by the following formula (1):

[0076]

[0077] Where A represents the vector of demand information encoded by the encoder; B represents the vector of reference prompt information; A·B represents the inner product of vector A and vector B, which is the sum of the products of the elements at corresponding positions; ‖A‖ represents the norm of vector A; and ‖B‖ represents the norm of vector B.

[0078] It can be understood that A·B is used to express the similarity between vector A and vector B at corresponding positions. similarityThe value range is [-1, 1], where 1 indicates that the reference prompt information and the requirement information are completely similar, -1 indicates that the reference prompt information and the requirement information are completely dissimilar, and 0 indicates that there is no correlation between the reference prompt information and the requirement information. Norms are used to measure the size of vectors, including but not limited to the L1 norm and L2 norm. The L1 norm, also known as the Manhattan norm, is the sum of the absolute values ​​of the vector elements. The L2 norm, also known as the Euclidean norm, is the square root of the sum of the squares of the vector elements.

[0079] In some non-limiting examples of the present application, the L2 norm can be used to calculate the modulus of vector A, that is, the "length" or "amplitude" of vector A is measured by calculating the square root of the sum of the squares of each element of vector A.

[0080] Specifically, the L2 norm of vector A can be calculated using the following formula (2):

[0081]

[0082] Where a represents an element of vector A.

[0083] It should be noted here that when calculating Cosine similarity In the process, you can use matrices, scientific Python (Scipy), numerical Python (Numpy) libraries, etc. to accelerate the calculation process to improve calculation efficiency and reduce calculation time.

[0084] By using cosine similarity to calculate the similarity of the semantic vectors between the reference prompt information and the requirement information, we can better compare the textual semantic similarity between the reference prompt information and the requirement information. Compared with string matching, this method has the following advantages:

[0085] First, this method transforms complex and tedious string operations into vector and matrix operations in vector space. Compared to character-by-character comparisons and tedious string operations, this transformation significantly improves efficiency and speed, making it much faster to process large amounts of text data.

[0086] Secondly, the similarity calculation method using semantic vectors can more accurately measure the degree of semantic connection between two texts. This method not only deeply explores the meaning and context of the text, but also avoids misjudgments caused by spelling errors, capitalization differences, or synonyms. As a result, this method demonstrates greater accuracy and robustness when processing text data with complex and changing semantics.

[0087] Finally, by converting the text into a vector representation and calculating cosine similarity, we can obtain a quantitative result that clearly measures the similarity between two texts. This cosine similarity calculation method based on semantic vectors is not only efficient but also accurate. It can quickly complete the similarity comparison task between large amounts of text, while also considering the overall semantic characteristics of the text rather than relying solely on the frequency of keyword occurrences, thereby providing more reliable and accurate similarity matching results.

[0088] Step 1202 : Filter target prompt information from multiple reference prompt information based on the similarity value between each reference prompt information and the requirement information, wherein the similarity value between the target prompt information and the requirement information is greater than a preset similarity threshold.

[0089] For example, similarity values ​​calculated between all reference prompt information and the requirement information are traversed, and each similarity value is compared with a preset similarity threshold. Reference prompt information having a similarity value greater than the preset threshold is selected to determine the target prompt information. For example, if the preset similarity threshold is set to 0.6, then step 1202 includes selecting the reference prompt information to determine the target prompt information only when the similarity value between the reference prompt information and the requirement information is greater than 0.6.

[0090] Step 1203: Determine a first service solution based on the reference service solution corresponding to the target prompt information.

[0091] For example, after determining the target prompt information, a pre-established association table in the electronic device may be searched, the association table recording the mapping relationship between each reference prompt information and the corresponding reference service solution. The corresponding reference service solution is found in the association table according to the target prompt information.

[0092] Thus, similarity matching is performed using the reference prompt information in the reference service plan, avoiding interference from unnecessary features in the reference service recommendation information and increasing the accuracy of the similarity value. Next, by filtering the reference prompt information using a preset similarity threshold, it is possible to quickly and accurately locate the target prompt information that is highly relevant to the demand information from multiple reference prompt information. This target prompt information can provide an effective reference for determining the first service plan. Subsequently, the first service plan determined based on the target prompt information is highly semantically consistent with the customer's demand information, effectively improving the accuracy of the first service plan determination.

[0093] In some embodiments of the present application, if the target prompt information corresponds to multiple reference service solutions, a secondary screening can be further performed based on other business rules, such as giving priority to service solutions with lower costs, higher service quality evaluations, or those that are more in line with current market trends. Figure 3As shown, the above step 1203 may specifically include steps 12031 to 12032.

[0094] Step 12031, when the number of reference service solutions corresponding to the target prompt information is at least two, obtain the recommended evaluation index value of each reference service solution in the at least two reference service solutions, and the recommended evaluation index value is used to reflect the quality of the reference service corresponding to the reference service solution recommended based on the reference service solution.

[0095] Among them, recommendation evaluation indicators are quantitative standards used to measure the performance of reference service solutions in recommending their corresponding reference services. Specifically, they may include at least one of the following: recommendation success rate, recommendation response speed, recommendation accuracy, recommendation coverage, user engagement, recommendation influence, and cost-effectiveness. The recommendation success rate refers to the proportion of users who successfully accept and use the reference service after the application of the reference service solution; the recommendation response speed refers to the time interval between the business end recommending the reference service to the user based on the reference service solution and receiving the user's response; the recommendation coverage refers to the scope of the service recommendation group in the reference service solution; the user engagement refers to the degree of user interaction with the reference service under the application of the reference service solution, such as clicking on details, consulting, etc.; the recommendation influence refers to the degree of influence of the reference service on user decision-making and the service market under the application of the reference service solution; and the cost-effectiveness refers to the ratio of the recommendation quality of the reference service to the service cost under the application of the reference service solution. The recommendation evaluation indicator value is the specific quantitative value corresponding to each recommendation evaluation indicator, which is used to intuitively reflect the performance of the reference service solution in terms of the recommendation evaluation indicator.

[0096] For example, if the recommendation success rate, recommendation accuracy, recommendation coverage, user engagement, and recommendation influence all range from 0 to 100%, then a recommendation evaluation index value of over 60% for the reference service solution is considered excellent, and a value below 40% is considered poor. If the recommendation response speed is measured in points, a value below 3 is considered excellent, and a value above 5 is considered poor. If the cost-effectiveness is measured in a range of 0 to 10, a value of 8 or above is considered excellent, and a value below 5 is considered poor.

[0097] In some embodiments of the present application, for each reference service plan, the recommended evaluation index value corresponding to the reference service plan may refer to the comprehensive value of the reference service plan under multiple recommended evaluation indicators. The calculation method of the comprehensive value may set different weights according to the business needs corresponding to the service. For example, if the current business focuses on the recommendation success rate and cost-effectiveness, the weight of the recommended evaluation index value corresponding to the recommendation success rate may be set to 0.6, and the weight of the recommended evaluation index value corresponding to the cost-effectiveness may be set to 0.4.

[0098] Step 12032: Filter a first service solution from at least two reference service solutions based on the recommended evaluation index value of each reference service solution. The recommended evaluation index value of the first service solution is the maximum value of the recommended evaluation index values ​​of the at least two reference service solutions.

[0099] Exemplarily, after the recommended evaluation index value corresponding to each reference service solution in at least two reference service solutions is calculated based on the above step 12031, the reference service solution with the largest recommended evaluation index value is determined as the first service solution.

[0100] Therefore, by recommending the evaluation index value to conduct a secondary screening of the reference service solutions, the one-sidedness of selecting the first service solution based solely on semantic similarity is effectively avoided, and it is helpful to screen out the first service solution whose semantic matching degree meets the preset conditions and meets the business needs. The prompt information corresponding to the demand information is generated based on the first service solution, which can draw on the past experience of the first service solution and improve the quality of the target prompt information, and then generate the second service solution based on the high-quality target prompt information, which can further increase the overall quality and feasibility of the second service solution.

[0101] Furthermore, regarding step 130, the prompt information in the embodiment of the present application is used to generate the second service plan in step 140, which may specifically include part of the information in the demand information and part of the information in the first service plan. By recommending the first service plan that has a high degree of match with the demand information and a good effect, accurate prompt information can be generated.

[0102] In some embodiments of the present application, in order to improve the accuracy of the prompt information corresponding to the demand information, such as Figure 4 As shown, the above step 130 may specifically include steps 1301 to 1303.

[0103] Step 1301 : extract keywords from content related to prompt information in the first service solution according to the reference prompt information structure corresponding to the demand information, and generate first reference prompt information corresponding to the first service solution.

[0104] Among them, the reference prompt information structure is a pre-set framework that stipulates the key information categories that should be included in the prompt information, which can specifically include product information items and service recommendation group characteristic information items, wherein the product information items can include at least one of the following: product basic attribute information items, product function characteristic information items and product usage attribute information items; the service recommendation group characteristic information items can include at least one of the following: demographic characteristic information items, behavioral characteristic information items, and demand characteristic information items. Keyword extraction refers to the process of identifying and extracting words or phrases that can accurately summarize the main content of the text and are representative and of key significance from a piece of text content. The first reference prompt information includes the reference product information corresponding to the product information item in the reference prompt information of the first service plan and the reference service recommendation group characteristic information corresponding to the service recommendation group characteristic information item in the reference prompt information of the first service plan.

[0105] In one example, the reference prompt information structure may include product basic attribute information items and product functional feature information items in the product information items, and demographic feature information items and behavioral feature information items in the service recommendation group feature information items. In another example, the reference prompt information structure may include product functional feature information items in the product information items and demographic feature information items in the service recommendation group feature information items. It is worth noting that the reference prompt information structure can be adjusted based on different business needs and is not specifically limited here.

[0106] Exemplarily, first, the reference prompt information structure corresponding to the current demand information is clarified. For example, if the current demand information is to formulate prompt information for a certain electronic product promotion service, the reference prompt information structure includes product basic attribute information items, product functional feature information items, demographic feature information items, behavioral feature information items, and demand feature information items. Then, the first service plan is input into the pre-trained prompt information generation model, and the prompt information generation model is used to perform keyword extraction processing on the first service plan, and output the first reference prompt information. Among them, the prompt information generation model is trained based on the prompt information training sample, and the prompt information training sample includes a sample service plan and sample prompt information corresponding to the sample service plan. The sample prompt information is obtained by annotating the sample service plan based on the reference prompt information structure. The prompt information training sample is used to train the prompt information generation model until the preset training stop condition is met to obtain the prompt information generation model.

[0107] The prompt information generation model in the embodiment of the present application uses the deep learning model T5 (Text-to-Text Transfer Transformer) model. The T5 model is a multi-purpose natural language processing model based on the Transformer structure, which has demonstrated excellent capabilities in the summary task of text keyword extraction. The Transformer structure consists of multiple "encoder-decoder" layers, each encoder is responsible for encoding and representing the input text, and the decoder is responsible for generating the target text based on the encoded representation. This structure enables the T5 model to capture the contextual dependencies in the input text, thereby improving the accuracy and coherence of the generated result, that is, the generated prompt information. In addition to the Transformer structure, the T5 model also uses a self-attention mechanism to capture the contextual dependencies in the input text. The self-attention mechanism allows the T5 model to focus on keywords in different positions during the generation process, and decides which keywords should be included in the final output based on the relationship between them.

[0108] The self-attention calculation formula can be expressed by the following formula (3):

[0109]

[0110] Among them, Q, K and V represent the query, key and value vectors after linear transformation respectively. Softmax represents the probability normalization function. is the normalization factor, d k Indicates the dimension of the vector.

[0111] As you can understand, the process of calculating the query vector Q, key vector K, and value vector V is a crucial step in the self-attention mechanism. First, the input query vector Q, key vector K, and value vector V are preprocessed by applying linear transformations. This process involves scaling, offsetting, and other operations to make them more suitable for the self-attention computation. The preprocessed Q, K, and V are then fed into a softmax function for probability normalization. The softmax function operates by dividing the exponent of each element by the sum of all the exponents, ultimately outputting a value between 0 and 1. This way, each element of the vector represents its importance in the sequence, or its weight. Next, attention scores are calculated by calculating the dot product of Q and K, and the dot product of Q and V. These scores are further used in weighted summation or other related calculations to produce the final attention output. This attention output is then used to generate prompts in natural language processing. Thanks to this mechanism, the T5 model is able to gain a deeper understanding of the input text, thereby generating more accurate and relevant prompts. To further enhance model training effectiveness, the T5 model also introduces residual connections and layer normalization techniques. Residual connections help the model better learn complex patterns in input data, improving its generalization capabilities; layer normalization accelerates model training and enhances its stability and convergence speed.

[0112] Step 1302 : extract keywords from the demand information according to the reference prompt information structure, and generate second reference prompt information corresponding to the demand information.

[0113] The second reference prompt information includes product information corresponding to the product information item in the demand information and service recommendation group characteristic information corresponding to the service recommendation group characteristic information item in the demand information.

[0114] Exemplarily, the demand information may be input into the prompt information generation model in step 1301 , and the demand information may be processed by the prompt information generation model to obtain the second reference prompt information.

[0115] Step 1303: Adjust the second reference prompt information according to the first reference prompt information to obtain prompt information corresponding to the requirement information.

[0116] Exemplarily, the first reference prompt information is inserted into the corresponding positions of each information item in the reference prompt information structure as a reference example for generating the second service plan. For example, if the reference prompt information structure is "For [demographic characteristic information item], our service / product has [product feature information item] and can solve [demand characteristic information item]," and if the "demographic characteristic information item" in the first reference prompt information is "sports enthusiasts under 25 years old," the "product feature information item" is "fast charging, lightweight and portable," and the "demand characteristic information item" is "sufficient device power requirement," then this information is inserted into the reference prompt information structure. Then, based on the presentation of the first reference prompt information in the reference prompt information structure, the second reference prompt information is adjusted. If the "target customer group" in the second reference prompt information is "young people who enjoy outdoor activities," it can be adjusted to a more precise description similar to the first reference prompt information: "outdoor sports enthusiasts under 25 years old." If the "product feature information item" is "long battery life," it can be adjusted to "long battery life and support for fast charging" in combination with the "fast charging" in the first reference prompt information. Through such adjustment, the second reference prompt information is made to match the presentation of the first reference prompt information in the template in terms of format and content style, and finally the prompt information is obtained.

[0117] Thus, by extracting keywords from the demand information and organizing it according to its structure, the service demand information can be presented in a clear and structured manner, highlighting the key points of the demand information. By using the first reference prompt information as an example to adjust the second reference prompt information, it is possible to fully draw on the effective information and expression methods of the first service solution. The generated prompt information not only contains the core points of the demand information, but also has a more standardized and unified structure and expression style. This provides strong support for the accurate generation of the second service solution corresponding to the demand information, improves the quality and accuracy of the generated second service solution corresponding to the demand information, and enhances the matching degree between the second service solution and the user's needs.

[0118] Then, in step 140, the second service solution in the embodiment of the present application includes prompt information corresponding to the demand information and service recommendation information generated by the prompt information. The service recommendation information can be a service recommendation text used to recommend specific service products or service packages to the user.

[0119] In some embodiments of the present application, the above-mentioned step 140 may include: processing the prompt information corresponding to the demand information through a service solution generation model to obtain a second service solution; wherein the service solution generation model is obtained based on training samples, the training samples include sample prompt information and sample service solutions corresponding to the sample prompt information, the sample prompt information includes first sample product information and first sample service recommendation group characteristic information, and the sample service solution includes second sample product information, second sample service recommendation group characteristic information and sample service recommendation information.

[0120] The prompt information corresponding to the demand information is input into the service solution generation model, and the service solution generation model processes the prompt information of the demand information and outputs a second service solution. The service solution generation model in the embodiment of the present application is a vertical large language model with the function of generating service solutions.

[0121] Exemplarily, the number of training samples is multiple. First, an initial solution generation model and multiple training samples are constructed. Then, the initial solution generation model is trained using the multiple training samples. When the preset training stop condition is met, the above-mentioned service solution generation model is obtained. The initial solution generation model may include the Chat General Language Model 2-6 Billion Parameters (ChatGLM2-6B) model and the Language Model for Autoregressive Modeling of the Web 2-70 Billion Parameters (LLaMA2-70B) model. ChatGLM2-6B is an open source conversational language model built on the General Language Model (GLM) architecture, with 6.2 billion parameters, and optimized for Chinese. LLaMA2-70B is a model built on the Language Model (LM) architecture, with 70 billion parameters, capable of handling various natural language tasks. These two models are optimized for Chinese question answering and text generation, and can generate answers that meet human preferences. Next, to obtain personalized models for the service solution generation scenario, the two models were subjected to efficient parameter fine-tuning using the Low-Rank Adaptation (LORA) technique. Before conducting LORA fine-tuning training, the hyperparameters required for LORA fine-tuning were defined. For example, by evaluating the complexity of the fine-tuning training task, hyperparameters such as lora_rank were reasonably set. Complexity assessment can consider factors such as the size of the training samples, the sample data characteristics, namely the sample prompt information, the diversity of the sample service solutions corresponding to the sample prompt information, and the specific requirements of the service solution generation task. For example, if the size of the training samples is small and the sample data characteristics are relatively simple, lora_rank can be set to a small value; otherwise, it can be appropriately increased. Subsequently, the training process is started, and the training samples are input into the initial solution generation model. During the training process, key training indicators such as the loss function are monitored. By observing the changing trend of the loss function, it is determined whether the model training has converged. If the model training does not converge, for example, the loss function value does not decrease steadily after a long period of training, or there are large fluctuations, you need to adjust the training hyperparameters, such as appropriately adjusting the value of lora_rank, or changing other hyperparameters such as the learning rate, and then retrain until the model training converges and a solution generation model is obtained.

[0122] It is understandable that the lightweight fine-tuning training based on LORA freezes the weights of the pre-trained model during fine-tuning to avoid retraining a large number of parameters. Instead, a trainable rank decomposition matrix is ​​introduced at each layer of the Transformer architecture. In this way, the number of parameters that need to be trained for downstream tasks is significantly reduced, and the fine-tuning efficiency is improved. At the same time, it can flexibly adapt to different downstream task requirements, and on the basis of retaining the powerful language understanding and generation capabilities of the pre-trained model, it performs personalized optimization for the service solution generation task in the business scenario of the business end. By reasonably setting hyperparameters and monitoring the training process, the model can quickly and effectively learn the key knowledge and patterns in the sample service solutions, and improve the performance of the model in this specific field. Compared with the traditional full-parameter fine-tuning method, the LORA method significantly reduces the consumption of training time and computing resources, improves the fine-tuning efficiency, and can flexibly adapt to different business needs and data characteristics to generate service solutions that are more in line with actual application scenarios.

[0123] Thus, the service solution generation model generates a second service solution, efficiently and accurately transforming the user's input demand information into a complete service solution with practical application value. Based on the knowledge and patterns learned from the training samples, the service solution model fully considers various factors, such as the product information corresponding to the service and the user group characteristics in the service recommendation group characteristic information. The generated service recommendation information is more closely aligned with user needs, improving the relevance and effectiveness of the service solution. Compared with manually written service solutions, this improves the efficiency of service solution generation, reduces human errors, and better meets the needs of business end users to quickly generate high-quality service solutions in diverse business scenarios.

[0124] In some embodiments of the present application, the second service solution may further include optimized prompt information generated by the prompt information, and the service solution generation model may include a semantic optimization model and an information expansion model. Figure 5 As shown, the above step 140 may specifically include steps 1401 to 1403.

[0125] Step 1401 : semantically optimize the prompt information corresponding to the demand information using a semantic optimization model to obtain optimized prompt information.

[0126] Step 1402: Perform information expansion processing on the optimization prompt information through the information expansion model to obtain initial service recommendation information.

[0127] Step 1403 : Perform semantic optimization processing on the initial service recommendation information through the semantic optimization model to obtain service recommendation information.

[0128] In one example, the aforementioned ChatGLM2-6B can be used to construct an initial semantic optimization model, which can then be fine-tuned on a specific semantic optimization task dataset to obtain a semantic optimization model. For example, a large amount of first-sample product information and first-sample service recommendation group characteristic information containing semantic errors and ambiguous expressions can be collected, as well as labeled data corresponding to the first-sample prompt information, namely, second-sample product information with clear semantic expressions and second-sample service recommendation group characteristic information. Fine-tuning ChatGLM2-6B is then trained to learn how to correct these issues, thereby achieving semantic optimization functionality.

[0129] Similarly, the above-mentioned LLaMA2-70B model can be used to construct an initial information extension model. For the information extension task, the LLaMA2-70B model is fine-tuned through the second sample product information, the second sample service recommendation group characteristic information and the corresponding sample service plan to obtain an information extension model, which focuses on mining potential information related to the input prompt information to obtain initial service recommendation information.

[0130] In some embodiments of the present application, when prompt information corresponding to demand information needs to be processed, the prompt information is first input into the ChatGLM2-6B model that has been fine-tuned into a semantic optimization model for semantic optimization, and the optimized prompt information is output; then, the optimized prompt information is input into the LLaMA2-70B model that has been fine-tuned into an information expansion model for information expansion, and the initial service recommendation information is output; subsequently, the initial service recommendation information is input into the ChatGLM2-6B model that has been fine-tuned into a semantic optimization model for semantic optimization, and the service recommendation information is output, and then, the optimized prompt information and the service recommendation information are determined as the second service plan, or the service recommendation information is determined as the second service plan.

[0131] As a result, the optimized prompt information processed by the semantic optimization model is more semantically accurate and clear, which helps the information expansion model understand and process the prompt information, avoiding deviations in service solution generation caused by semantic ambiguity or errors. By processing the optimized prompt information with the information expansion model, a more comprehensive initial service recommendation information is generated. Based on this initial service recommendation information, more comprehensive service recommendations can be provided to users. In this way, the coordinated use of the semantic optimization model and the information expansion model can improve the quality and accuracy of the second service solution.

[0132] Based on the information generation method provided in the above embodiment, the present application also provides a specific implementation of an information generation device. Please refer to the following embodiment.

[0133] See first Figure 6 The information generating device 600 provided in this embodiment of the present application includes the following modules:

[0134] Acquisition module 610, used to obtain service demand information;

[0135] A first determining module 620 is configured to determine a first service solution from the preset service solutions based on the demand information, where the first service solution is a service solution in the preset service solutions whose matching degree with the demand information is greater than or equal to a preset matching degree;

[0136] A first generating module 630 is configured to generate prompt information corresponding to the demand information according to the first service solution;

[0137] The second generating module 640 is configured to generate a second service solution corresponding to the demand information according to the prompt information, where the second service solution includes service recommendation information generated from the prompt information.

[0138] Thus, the first determination module 620 can automatically filter out the first service plan whose degree of matching with the demand information is greater than or equal to the preset matching degree from the preset service plans based on the demand information of the service obtained by the acquisition module 610. After determining the first service plan, the first generation module 630 can further generate prompt information corresponding to the demand information based on the first service plan. The prompt information can provide a reference basis for the generation of the second service plan in the second generation module 640, and optimize the generation of the second service plan, effectively avoiding problems caused by omissions or deviations in the understanding of the demand information, and ensuring the accuracy and effectiveness of the second service plan. Thus, full automation is achieved from the acquisition of demand information to the generation of the second service plan, and the entire process does not require human intervention. This method of automatically generating the second service plan significantly reduces the cost required for manually writing service plans, while reducing the error rate caused by human factors, significantly improving the generation efficiency and quality of service plans, and effectively solving the problem of poor quality of service plans caused by manually writing service plans.

[0139] As an implementation of the present application, the first determining module 620 may specifically include:

[0140] a calculation submodule for calculating, when the degree of matching includes a similarity value, the preset degree of matching includes a preset similarity threshold, the preset service solution includes a plurality of reference service solutions, and each of the plurality of reference service solutions includes reference prompt information and reference service recommendation information, a similarity value between each piece of reference prompt information and the requirement information based on a semantic feature of each piece of reference prompt information and a semantic feature of the requirement information;

[0141] a screening submodule, configured to screen target prompt information from the plurality of reference prompt information according to a similarity value between each reference prompt information and the requirement information, wherein the similarity value between the target prompt information and the requirement information is greater than a preset similarity threshold;

[0142] The first determining submodule is configured to determine a first service solution according to a reference service solution corresponding to the target prompt information.

[0143] As an implementation of the present application, the first determination submodule specifically includes:

[0144] an acquiring unit, configured to acquire, when the number of reference service solutions corresponding to the target prompt information is at least two, a recommended evaluation index value for each of the at least two reference service solutions, the recommended evaluation index value being used to reflect the quality of a reference service corresponding to the reference service solution recommended based on the reference service solution;

[0145] The screening unit is used to screen a first service solution from at least two reference service solutions according to the recommended evaluation index value of each reference service solution, wherein the recommended evaluation index value of the first service solution is the maximum value of the recommended evaluation index values ​​of the at least two reference service solutions.

[0146] As an implementation of the present application, the first generating module 630 may specifically include:

[0147] A first generating unit is configured to extract keywords from content related to the prompt information in the first service solution according to a reference prompt information structure corresponding to the demand information, and generate first reference prompt information corresponding to the first service solution;

[0148] A second generating unit is configured to extract keywords from the demand information according to the reference prompt information structure, and generate second reference prompt information corresponding to the demand information;

[0149] The adjusting unit is configured to adjust the second reference prompt information according to the first reference prompt information to obtain prompt information corresponding to the requirement information.

[0150] As an implementation method of the present application, the reference prompt information structure includes product information items and service recommendation group characteristic information items; the first reference prompt information includes reference product information corresponding to the product information items in the reference prompt information of the first service plan and reference service recommendation group characteristic information corresponding to the service recommendation group characteristic information items in the reference prompt information of the first service plan; the second reference prompt information includes product information corresponding to the product information items in the demand information and service recommendation group characteristic information corresponding to the service recommendation group characteristic information items in the demand information.

[0151] As an implementation method of the present application, the above-mentioned second generation module 640 can be specifically used to: process the prompt information corresponding to the demand information through the service solution generation model to obtain a second service solution; wherein, the service solution generation model is obtained based on training samples, and the training samples include sample prompt information and sample service solutions corresponding to the sample prompt information.

[0152] As an implementation of the present application, the second generation module 640 may specifically include:

[0153] a second determining submodule configured to, when the second service solution further includes optimized prompt information generated from the prompt information and the service solution generation model includes a semantic optimization model and an information extension model, perform semantic optimization processing on the prompt information corresponding to the demand information using the semantic optimization model to obtain optimized prompt information;

[0154] The third determination submodule is configured to perform information expansion processing on the optimization prompt information through an information expansion model to obtain initial service recommendation information;

[0155] The fourth determining submodule is configured to perform semantic optimization processing on the initial service recommendation information through a semantic optimization model to obtain service recommendation information.

[0156] Figure 7 A schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present invention is shown.

[0157] The electronic device may include a processor 701 and a memory 702 storing computer program instructions.

[0158] Specifically, the processor 701 may include a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0159] The memory 702 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 702 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 702 may include removable or non-removable (or fixed) media. Where appropriate, the memory 702 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 702 is a non-volatile solid-state memory.

[0160] In certain embodiments, the memory 702 may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical, or other physical / tangible memory storage device. Thus, generally, the memory 702 includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the data processing method according to the first aspect of the present application.

[0161] The processor 701 implements any one of the information generating methods in the above embodiments by reading and executing computer program instructions stored in the memory 702 .

[0162] In one example, the electronic device may further include a communication interface 703 and a bus 710. Figure 7 As shown, the processor 701, the memory 702, and the communication interface 703 are connected via a bus 710 and communicate with each other.

[0163] The communication interface 703 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0164] Bus 710 comprises hardware, software or both, couples the parts of electronic equipment to each other.For example, and not limitation, bus can comprise accelerated graphics port (AGP) or other graphics bus, enhanced industry standard architecture (EISA) bus, front side bus (FSB), hypertransport (HT) interconnection, industry standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations.In suitable cases, bus 710 can comprise one or more buses.Although the present application embodiment describes and shows specific bus, the application considers any suitable bus or interconnection.

[0165] The electronic device can execute the information generation method in the embodiment of the present application, thereby realizing the combination Figures 1 to 6 The described information generation method and device.

[0166] In addition, in combination with the information generation method in the above embodiment, the embodiment of the present application may provide a computer-readable storage medium for implementation. The computer-readable storage medium stores computer program instructions; when the computer program instructions are executed by the processor, any one of the information generation methods in the above embodiment is implemented. Examples of computer-readable storage media include non-transitory computer-readable storage media, such as portable disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), portable compact disk read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, etc.

[0167] In addition, in conjunction with the information generation method in the above-mentioned embodiment, the embodiments of the present application may be implemented by providing a computer program product. The program product is stored in a storage medium and may specifically include a computer program or instructions. When the computer program or instructions are executed by a processor, any of the information generation methods in the above-mentioned embodiments is implemented. The program product is executed by at least one processor to implement the various processes of the above-mentioned data processing method embodiment and can achieve the same technical effects. To avoid repetition, it will not be described here.

[0168] It should be understood that the present invention is not limited to the specific configurations and processes described above and illustrated in the figures. For the sake of brevity, a detailed description of known methods is omitted. In the above embodiments, several specific steps are described and illustrated as examples. However, the method of the present invention is not limited to the specific steps described and illustrated. Those skilled in the art may make various changes, modifications, and additions, or change the order of the steps after understanding the spirit of the present invention.

[0169] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of the present invention are programs or code segments used to perform the required tasks. The programs or code segments can be stored in a machine-readable medium, or transmitted on a transmission medium or communication link via a data signal carried in a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, read-only memories (ROMs), flash memories, erasable read-only memories (EROMs), floppy disks, compact disc read-only memories (CD-ROMs), optical discs, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segments can be downloaded via computer networks such as the Internet and intranets.

[0170] It should also be noted that the exemplary embodiments described herein describe methods or systems based on a series of steps or devices. However, the present invention is not limited to the order of the steps described above. In other words, the steps may be performed in the order described in the embodiments, or in a different order, or several steps may be performed simultaneously.

[0171] Aspects of the present disclosure have been described above with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer or other programmable data processing device to produce a machine so that these instructions executed by the processor of the computer or other programmable data processing device enable the implementation of the function / action specified in one or more boxes of the flowchart and / or block diagram. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor or a field programmable logic circuit. It is also understood that each box in the block diagram and / or flowchart and the combination of the boxes in the block diagram and / or flowchart can also be implemented by dedicated hardware that performs the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0172] The above description is only a specific embodiment of the present invention. Those skilled in the art will clearly understand that for the convenience and brevity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present invention is not limited to this. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention.

Claims

1. A method for generating information, characterized in that: include: Obtain service demand information; Determining a first service solution from the preset service solutions based on the demand information, the first service solution being a service solution among the preset service solutions whose matching degree with the demand information is greater than or equal to a preset matching degree; generating prompt information corresponding to the demand information according to the first service solution; A second service solution corresponding to the demand information is generated according to the prompt information, wherein the second service solution includes service recommendation information generated by the prompt information.

2. The information generation method according to claim 1, characterized in that The matching degree includes a similarity value, the preset matching degree includes a preset similarity threshold, the preset service solution includes a plurality of reference service solutions, and each reference service solution in the plurality of reference service solutions includes reference prompt information and reference service recommendation information; Determining a first service solution from preset service solutions based on the demand information includes: Calculating a similarity value between each reference prompt information and the demand information based on a semantic feature of each reference prompt information and a semantic feature of the demand information; Filtering target prompt information from the plurality of reference prompt information according to a similarity value between each reference prompt information and the requirement information, wherein the similarity value between the target prompt information and the requirement information is greater than a preset similarity threshold; The first service solution is determined according to the reference service solution corresponding to the target prompt information.

3. The information generation method according to claim 2, characterized in that The determining the first service solution according to the reference service solution corresponding to the target prompt information includes: When the number of reference service solutions corresponding to the target prompt information is at least two, obtaining a recommendation evaluation index value for each of the at least two reference service solutions, the recommendation evaluation index value being used to reflect the quality of a reference service corresponding to the reference service solution recommended based on the reference service solution; A first service solution is selected from the at least two reference service solutions according to the recommended evaluation index value of each reference service solution, wherein the recommended evaluation index value of the first service solution is the maximum value of the recommended evaluation index values ​​of the at least two reference service solutions.

4. The information generation method according to any one of claims 1 to 3, characterized in that The generating, according to the first service solution, prompt information corresponding to the demand information includes: extracting keywords from content related to the prompt information in the first service solution according to the reference prompt information structure corresponding to the demand information, and generating first reference prompt information corresponding to the first service solution; extracting keywords from the demand information according to the reference prompt information structure to generate second reference prompt information corresponding to the demand information; According to the first reference prompt information, the second reference prompt information is adjusted to obtain prompt information corresponding to the demand information.

5. The information generation method according to claim 4, characterized in that The reference prompt information structure includes product information items and service recommendation group feature information items; The first reference prompt information includes reference product information corresponding to the product information item in the reference prompt information of the first service solution and reference service recommendation group characteristic information corresponding to the service recommendation group characteristic information item in the reference prompt information of the first service solution; The second reference prompt information includes product information corresponding to the product information item in the demand information and service recommendation group characteristic information corresponding to the service recommendation group characteristic information item in the demand information.

6. The information generation method according to any one of claims 1 to 3, characterized in that: Generating a second service solution corresponding to the demand information according to the prompt information includes: Processing the prompt information corresponding to the demand information through a service solution generation model to obtain the second service solution; The service solution generation model is obtained through training based on training samples, and the training samples include sample prompt information and sample service solutions corresponding to the sample prompt information.

7. The information generating method according to claim 6, characterized in that The second service solution further includes optimized prompt information generated from the prompt information, and the service solution generation model includes a semantic optimization model and an information expansion model; The generating model of the service solution processes the prompt information corresponding to the demand information to obtain the second service solution, including: Performing semantic optimization processing on the prompt information corresponding to the demand information by using the semantic optimization model to obtain optimized prompt information; Performing information expansion processing on the optimization prompt information through the information expansion model to obtain initial service recommendation information; The initial service recommendation information is semantically optimized using the semantic optimization model to obtain the service recommendation information.

8. An information generating device, characterized in that: The information generating device includes: Acquisition module, used to obtain service demand information; a first determining module, configured to determine a first service solution from the preset service solutions based on the demand information, the first service solution being a service solution among the preset service solutions whose matching degree with the demand information is greater than or equal to a preset matching degree; A first generating module, configured to generate prompt information corresponding to the demand information according to the first service solution; The second generating module is configured to generate a second service solution corresponding to the demand information according to the prompt information, wherein the second service solution includes service recommendation information generated by the prompt information.

9. An electronic device, characterized in that: The electronic device includes: a processor and a memory storing computer program instructions; the processor reads and executes the computer program instructions to implement the information generation method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed by a processor, the information generating method according to any one of claims 1 to 7 is implemented.

11. A computer program product, characterized in that When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes the information generating method according to any one of claims 1 to 7.

Images