A service recommendation method and device, electronic equipment, medium and product

Abstract

This disclosure provides a service recommendation method, apparatus, electronic device, medium, and product, relating to the field of computer technology. The method includes: filtering available services based on customer needs to generate a candidate service list; for each candidate service in the candidate service list, constructing an input sequence of the candidate service based on the customer needs and the candidate service; inputting the input sequence of the candidate service into an analysis model to obtain the matching degree between the candidate service and the customer needs; and obtaining a service recommendation result based on the matching degree between all candidate services in the candidate service list and the customer needs. This method can accurately identify the complex interaction relationships between customer needs and services, improving recommendation accuracy.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to a service recommendation method, service recommendation apparatus, electronic device, computer-readable storage medium, and computer program product. Background Technology

[0002] With the rapid development of communication services and the increasing diversification of customer needs, subscribing to a single service is no longer sufficient to meet customer demands. Operators have launched various types of services, and there are multiple business rules between these services.

[0003] In related technologies, the service package ordering process follows this procedure: customers submit their service requests or changes through online or offline channels, expressing their needs to a salesperson; the salesperson, based on their business experience and recommended promotional activities, recommends different service packages to the customer, who then chooses. However, because the recommendations are based on the salesperson's experience and promotional activities, the recommendations may not meet the customer's needs, impacting the customer experience. Summary of the Invention

[0004] The purpose of this disclosure is to provide a service recommendation method, service recommendation device, electronic device, computer-readable storage medium, and computer program product, which at least partially solves the problems existing in the related technologies.

[0005] According to a first aspect of this disclosure, a service recommendation method is provided, comprising: filtering available services according to customer needs to generate a candidate service list; for each candidate service in the candidate service list, constructing an input sequence for the candidate service based on the customer needs and the candidate service; inputting the input sequence of the candidate service into an analysis model to obtain a matching degree between the candidate service and the customer needs; and obtaining a service recommendation result for the customer needs based on the matching degree between all candidate services in the candidate service list and the customer needs.

[0006] In some embodiments of this disclosure, the candidate service is a single service or a combination of services.

[0007] In some embodiments of this disclosure, the step of filtering the available services according to customer needs to generate a candidate service list includes: filtering the available services according to the customer needs to obtain a service list that meets the customer needs; and pruning the service list according to the customer needs to generate the candidate service list.

[0008] In some embodiments of this disclosure, the method further includes: performing constraint relationship verification on the service combinations included in the service list based on preset inter-service constraint relationships, and removing service combinations that do not satisfy the inter-service constraint relationships.

[0009] In some embodiments of this disclosure, the service constraint relationship includes at least one of service mutual exclusion relationship, service quantity limit constraint, and service dependency relationship.

[0010] In some embodiments of this disclosure, the step of constructing the input sequence of the candidate services based on the customer needs and the candidate services includes: vectorizing the customer needs to obtain a customer needs vector; vectorizing the candidate services to obtain a service vector of the candidate services; and concatenating the customer needs vector and the service vector of the candidate services to construct the input sequence of the candidate services.

[0011] In some embodiments of this disclosure, the customer demand vector includes a customer demand feature vector and a customer demand embedding vector; wherein, the step of vectorizing the customer demand to obtain the customer demand vector includes: classifying and standardizing the customer demand to obtain the customer demand feature vector; and performing semantic encoding processing on the descriptive information of the customer demand based on a text embedding model to obtain the customer demand embedding vector.

[0012] In some embodiments of this disclosure, the service vector of the candidate service includes the service feature vector of the candidate service and the service embedding vector of the candidate service; wherein, the step of vectorizing the candidate service to obtain the service vector of the candidate service includes: classifying and standardizing the parameter information of the candidate service to obtain the service feature vector of the candidate service; and performing semantic encoding processing on the description information of the candidate service based on a text embedding model to obtain the service embedding vector of the candidate service.

[0013] In some embodiments of this disclosure, the method further includes: inserting a special identifier vector between the customer demand vector and the service vector of the candidate service; the special identifier vector is a preset fixed vector used to separate the customer demand semantics and the candidate service semantics.

[0014] In some embodiments of this disclosure, the method further includes: determining position information of each vector in the input sequence of the candidate service; encoding the position information of the vector to obtain a position code of the vector; the position code of the vector is used to identify the order of the vector in the input sequence of the candidate service.

[0015] In some embodiments of this disclosure, the analysis model includes an encoder and a scoring layer; wherein, the step of inputting the input sequence of the candidate service into the analysis model to obtain the matching degree between the candidate service and the customer demand includes: encoding the input sequence of the candidate service through the encoder to obtain a global feature vector of the candidate service; the global feature vector of the candidate service is used to characterize the global dependency relationship between the customer demand and the candidate service; and the scoring layer calculates the matching degree between the global feature vector of the candidate service to obtain the matching degree between the candidate service and the customer demand.

[0016] In some embodiments of this disclosure, obtaining the service recommendation result for the customer's needs based on the matching degree between all candidate services in the candidate service list and the customer's needs includes: sorting the matching degree between all candidate services and the customer's needs, filtering out a preset number of candidate services with the highest matching degree; and using the filtered candidate services as the service recommendation result for the customer's needs.

[0017] According to a second aspect of this disclosure, a service recommendation apparatus is provided, comprising: a candidate service generation module configured to filter available services according to customer needs and generate a candidate service list; a sequence construction module configured to construct an input sequence of the candidate service for each candidate service in the candidate service list, based on the customer needs and the candidate service; a matching degree determination module configured to input the input sequence of the candidate service into an analysis model to obtain a matching degree between the candidate service and the customer needs; and a recommendation module configured to obtain a service recommendation result for the customer needs based on the matching degree between all candidate services in the candidate service list and the customer needs.

[0018] According to a third aspect of this disclosure, an electronic device is provided, including a processor and a memory, the memory being used to store executable instructions of the processor; wherein the processor is configured to perform the above-described service recommendation method by executing the executable instructions.

[0019] According to a fourth aspect of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described service recommendation method.

[0020] According to a fifth aspect of this disclosure, a computer program product is provided, the computer program product storing instructions that, when executed by a computer, cause the computer to implement the above-described service recommendation method.

[0021] The service recommendation method provided in this disclosure filters available services based on customer needs to generate a candidate service list, constructs an input sequence combining customer needs for each candidate service in the candidate service list, and then uses an analysis model to quantify the matching degree between each candidate service and the needs, thereby objectively ranking and outputting recommendation results based on the matching degree. This method can accurately identify the complex interaction relationship between customer needs and services, and improve recommendation accuracy.

[0022] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0023] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure. It is obvious that the drawings described below are merely some embodiments of this disclosure, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0024] Figure 1 A flowchart of a service recommendation method according to an embodiment of this disclosure is shown.

[0025] Figure 2 A flowchart illustrating the construction of an input sequence based on customer needs and candidate services is shown in this embodiment of the disclosure.

[0026] Figure 3 A flowchart illustrating the matching degree between candidate services and customer needs based on an analytical model is shown in this embodiment of the present disclosure.

[0027] Figure 4 A diagram illustrating the overall architecture of the service recommendation method according to an embodiment of this disclosure is shown.

[0028] Figure 5 A schematic diagram of a service recommendation device according to an embodiment of the present disclosure is shown.

[0029] Figure 6 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown. Detailed Implementation

[0030] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that this disclosure will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0031] Furthermore, the accompanying drawings are merely illustrative of this disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0032] Figure 1 A flowchart of a service recommendation method according to an embodiment of this disclosure is shown. Figure 1 The method provided can be implemented by any electronic device, such as a terminal device, a server, or a service recommendation method jointly implemented by a server and a terminal device, but this disclosure is not limited to these. See reference. Figure 1 The recommended method for this service may include the following steps.

[0033] Step S110: Filter the available services according to customer needs and generate a candidate service list.

[0034] Step S120: For each candidate service in the candidate service list, construct the input sequence of the candidate service based on customer needs and the candidate service.

[0035] Step S130: Input the input sequence of candidate services into the analysis model to obtain the matching degree between candidate services and customer needs.

[0036] Step S140: Based on the matching degree between all candidate services in the candidate service list and the customer's needs, obtain the service recommendation result for the customer's needs.

[0037] In this embodiment, a candidate service list is generated by filtering available services according to customer needs. For each candidate service in the list, an input sequence combining customer needs is constructed. Then, an analysis model is used to quantify the matching degree between each candidate service and the needs. Based on the matching degree, the services are objectively sorted and recommendation results are output. This can accurately identify the complex interaction relationship between customer needs and services and improve the accuracy of recommendations.

[0038] The solutions provided by the present disclosure will be further described below with reference to exemplary embodiments.

[0039] In step S110, the available services are filtered according to customer needs to generate a candidate service list.

[0040] Customer needs refer to the explicit or implicit business expectations expressed by customers when applying for or changing services. These expectations can include, but are not limited to, multiple dimensions such as budget, data usage requirements, preferred benefits, and additional perks. For example, a customer might want "monthly fees not exceeding 80 yuan," "at least 30GB of general data per month," "includes memberships to major video platforms," ​​and "supports 5G networks."

[0041] Available services refer to all services that the service provider can currently offer for subscription. For example, a certain operator's available services may include "79 yuan 5G unlimited package", "10 yuan 30GB targeted data package", "video platform gold membership monthly card" and other services.

[0042] In this step, based on attributes such as tariff caps, data usage requirements, benefit preferences, and additional benefits in the customer's needs, all services currently available for subscription by the service provider (i.e., selectable services) are filtered and combined to obtain a candidate service list. This candidate service list consists of one or more candidate services, and each candidate service is subsequently analyzed to obtain service recommendations tailored to the customer's needs.

[0043] In some embodiments of this disclosure, each candidate service in the candidate service list is a service or a combination of services.

[0044] A service refers to a service product that can be subscribed to independently, such as a basic communication package. A service package is formed by combining multiple services. In one possible implementation, a service package can consist of a main service and one or more supplementary services. For example, a service package may include an 89 yuan basic package, a 10 yuan 30GB targeted data package, and a video platform premium membership.

[0045] In this embodiment, the candidate service is a single service or a combination of services. That is, when generating candidate solutions, it is not limited to recommending a single service, but can also recommend service combinations that can be packaged into a whole and ordered as a whole. This can more comprehensively cover the complex and multi-dimensional needs of customers, avoid missing better combination solutions due to only considering a single product, and improve the flexibility and practicality of the recommendations.

[0046] In some embodiments of this disclosure, filtering available services based on customer needs to generate a candidate service list includes: filtering available services based on customer needs to obtain a service list that meets customer needs; and pruning the service list based on customer needs to generate a candidate service list.

[0047] The service list includes at least one service, each of which can be a single service or a combination of services. Based on attributes such as tariff caps, data usage requirements, preferred benefits, and additional benefits in the customer's needs, a preliminary filter is performed on all services currently available for subscription by the service provider (i.e., selectable services). Services that clearly do not meet the customer's needs are excluded, resulting in a service list that meets those needs. One possible implementation is to use numerical constraint filtering, such as setting a rule to "exclude services with monthly fees exceeding 1.5 times the customer's needs"; or to use category matching filtering, such as setting a rule to "retain only services that support 5G networks."

[0048] In some embodiments of this disclosure, the method further includes: performing constraint relationship verification on the service combinations included in the service list based on preset inter-service constraint relationships, and removing service combinations that do not meet the inter-service constraint relationships.

[0049] In this implementation, each item in the service list can be a single service or a combination of services. After initial filtering to obtain a service list that meets customer needs, the service combinations included in the service list can be further validated based on preset inter-service constraints, thereby eliminating service combinations that do not meet the inter-service constraints. Inter-service constraints refer to the logical relationships of mutual restriction or dependence between different services during service ordering, combination, or use due to business rules, technical limitations, or operational strategies. These relationships determine which services can be ordered simultaneously, which cannot coexist, and which must be combined according to specific conditions, serving as a key basis for ensuring that service combinations are executable and meet customer expectations.

[0050] In some embodiments of this disclosure, the service constraint relationship includes at least one of the following: service mutual exclusion relationship, service quantity limit constraint, and service dependency relationship.

[0051] Service mutual exclusion refers to the inability to subscribe to two or more services simultaneously. For example, a "Campus Discount Package" and a "Corporate Exclusive Package" cannot be subscribed to at the same time; a customer can only choose one. Service quantity restrictions refer to setting an upper or lower limit on the number of services that can be subscribed to, such as "a maximum of two targeted data packages can be stacked" or "at least one basic communication service must be selected." Service dependency refers to the condition that the subscription of one service is contingent upon the existence of another service. For example, subscribing to an "International Roaming Add-on Package" requires that a "Basic Voice Package" has already been activated. It should be noted that in addition to service mutual exclusion, service quantity restrictions, and service dependency, other service constraints also include relationships such as service substitution and service synergy / discount relationships, which are not limited in this disclosure.

[0052] In this embodiment, pruning refers to further streamlining the service list obtained from the initial filtering based on preset rules and thresholds, actively eliminating redundant, suboptimal, or low-value candidate solutions, so as to effectively alleviate the problem of service or service combination explosion and improve the processing efficiency of subsequent recommendation models.

[0053] One possible implementation is to prune based on the similarity between services or service combinations. Specifically, when different services or service combinations are highly similar in dimensions such as functional composition, resource quotas, or additional benefits, only the more economical combination is retained, such as the service or service combination with lower total cost, better unit resource cost, or higher overall cost-effectiveness, while the remaining similar services or service combinations are eliminated, thereby reducing the number of services or service combinations.

[0054] The method provided in this embodiment performs initial filtering of available services to obtain a service list that meets customer needs. This service list may include single services or service combinations. For service combinations, constraint relationships are verified based on the relationships between services to eliminate those that do not meet the constraints. Then, pruning is performed to remove redundancies or suboptimal options, ultimately generating a candidate service list. By introducing a dual screening mechanism of constraint verification and pruning, the size of the candidate list is effectively reduced, and redundancies and suboptimal options are eliminated. Subsequent processing based on this candidate service list can improve recommendation efficiency and customer satisfaction.

[0055] In step S120, for each candidate service in the candidate service list, an input sequence for the candidate service is constructed based on customer needs and the candidate service.

[0056] In this step, each candidate service is processed, and the input sequence of the candidate service is constructed by combining the customer's needs and the candidate service's own attributes to characterize the candidate service's feature representation in the context of the customer's needs.

[0057] Figure 2 A flowchart illustrating the construction of an input sequence based on customer needs and candidate services is shown in this embodiment of the disclosure. (Refer to...) Figure 2 This may include the following steps.

[0058] Step S210: Vectorize the customer requirements to obtain the customer requirement vector.

[0059] In this step, customer requirements are vectorized to transform them into a structured input sequence that the model can process, thus obtaining a customer requirement vector.

[0060] In some embodiments of this disclosure, the customer demand vector includes a customer demand feature vector and a customer demand embedding vector. Specifically, the process of vectorizing customer demands to obtain the customer demand vector includes: classifying and standardizing customer demands to obtain a customer demand feature vector; and semantically encoding the descriptive information of customer demands based on a text embedding model to obtain a customer demand embedding vector.

[0061] In this embodiment, the customer demand vector consists of two parts: one part is the customer demand feature vector, which can be generated based on the structured business parameters of the customer demand; the other part is the customer demand embedding vector, which can be generated based on the customer's demand description expressed in natural language.

[0062] In one possible implementation, customer needs can be categorized into numerical (e.g., monthly fee cap, general data allowance) and categorical (e.g., whether video membership is included, supported network types). For numerical needs, normalization or linear scaling can be used for numerical standardization; for categorical needs, one-hot encoding or tag encoding can be used to convert them into numerical vectors, thereby forming a structured customer need feature vector.

[0063] Text embedding models are pre-trained semantic encoding models, such as Bidirectional Encoder Representations from Transformers (BERT) or other deep learning models suitable for text representation. Customers can express their needs in natural language through online customer service dialogues, search box input, speech-to-text conversion, etc., for example, "I want a data plan with a lot of data and a low price, preferably with video playback." This text is input into the text embedding model, which outputs a customer need embedding vector. This vector can effectively capture semantic information and contextual relationships in the text, supplementing the intent details that structured features fail to cover.

[0064] Step S220: Vectorize the candidate services to obtain the service vectors of the candidate services.

[0065] In this step, the candidate services are vectorized to transform them into a structured input sequence that the model can process, thus obtaining the service vectors of the candidate services.

[0066] In some embodiments of this disclosure, the service vector of a candidate service includes a service feature vector and a service embedding vector. Specifically, the process of vectorizing a candidate service to obtain its service vector includes: classifying and standardizing the parameter information of the candidate service to obtain its service feature vector; and semantically encoding the descriptive information of the candidate service based on a text embedding model to obtain its service embedding vector.

[0067] In this embodiment, the service vector of a candidate service consists of two parts: one part is the service feature vector, which can be generated based on the structured business parameters of the candidate service; the other part is the service embedding vector, which can be generated based on the natural language description of the service.

[0068] In one possible implementation, the parameter information of candidate services can be divided into numerical (such as monthly price, included general data allowance, and call duration) and categorical (such as whether 5G is supported, the type of data plan, and the included benefits platform). For numerical parameters, normalization or linear scaling can be used for numerical standardization; for categorical parameters, one-hot encoding or label encoding can be used to convert them into numerical vectors, thereby forming a structured service feature vector for candidate services.

[0069] The text embedding model has already been explained above and will not be repeated here. Candidate services are usually accompanied by descriptive information in natural language, such as a product description like "This package includes 40GB of high-speed data, supports 5G networks, and includes a free video membership." This descriptive text is input into the text embedding model, which outputs the corresponding service embedding vector. This vector effectively captures the semantic information and contextual relationships in the service description, supplementing service characteristics that structured parameters fail to reflect.

[0070] As explained above, a candidate service can be a single service or a combination of multiple services. If the candidate service is a single service, its parameter information and description text are vectorized as described above to generate service feature vectors and service embedding vectors, which are then merged into the service vector of the candidate service. If the candidate service is a combination of services, i.e., it contains multiple services, each service is vectorized as described above to obtain a service vector for each service, and then the service vectors of each service are used to form the service vector of the candidate service.

[0071] Step S230: Concatenate the customer demand vector and the service vector of the candidate service to construct the input sequence of the candidate service.

[0072] The customer demand vector includes a customer demand feature vector and a customer demand embedding vector. The candidate service vector includes a candidate service feature vector and a candidate service embedding vector. In this step, the customer demand vector and the candidate service vector are concatenated to obtain the input sequence of the candidate service. For example, if the candidate service is a service combination consisting of 3 services, its input sequence is [customer demand feature vector, customer demand embedding vector, service feature vector 1, service embedding vector 1, service feature vector 2, service embedding vector 2, service feature vector 3, service embedding vector 3].

[0073] Step S210 combines structured processing of customer needs with natural language semantic encoding to comprehensively represent the customer's combined intent in terms of business dimensions (such as tariff caps and data usage requirements) and linguistic descriptions (such as "lots of data at low prices"). Step S220 extracts the structured parameter vectors and semantic vectors of the service descriptions from candidate services (whether single services or service combinations). Step S230 then concatenates the customer need vectors with the service vectors of the candidate services to construct an input sequence. This input sequence contains quantifiable and comparable business attribute information and semantically understandable textual description information, enabling subsequent analysis models to perform accurate matching based on a complete and aligned feature space, thereby improving the accuracy of recommendation results and user satisfaction.

[0074] In some embodiments of this disclosure, the method further includes inserting a special identifier vector between the customer demand vector and the service vector of the candidate service.

[0075] The special marker vector is a pre-defined fixed vector used to separate customer requirement semantics from candidate service semantics. In one possible implementation, the special marker vector is an all-zero vector.

[0076] For example, a candidate service is a service combination consisting of three services, and its input sequence is [Customer demand feature vector, Customer demand embedding vector, Special label vector, Service feature vector 1, Service embedding vector 1, Service feature vector 2, Service embedding vector 2, Service feature vector 3, Service embedding vector 3]. It can be seen that a special label vector is inserted between the customer demand vector and the service vector of the candidate service in the input sequence. Subsequent input of this sequence into the model enables the model to better identify the interaction relationship between customer needs and services.

[0077] In some embodiments of this disclosure, the method further includes: determining the position information of each vector in the input sequence of the candidate service; and encoding the position information of the vector to obtain the position code of the vector. The position code of the vector is used to identify the order of the vectors in the input sequence of the candidate service.

[0078] In this implementation, for each vector in the input sequence, its sequential position within the sequence can be determined. For example, the 0th position is the "customer demand feature vector," the 1st position is the "customer demand embedding vector," the 2nd position is the "service feature vector of the first service among candidate services," the 3rd position is the "service embedding vector of the first service among candidate services," and so on. After determining the positional information of each vector in the input sequence, the positional information is converted into a numerical representation matching the vector dimension, i.e., a positional code. By adding a corresponding positional code to each vector, the model can identify the position of each vector in the overall input sequence, thereby accurately distinguishing the roles and order of different components. This enables correct parsing of the input structure and modeling of the logical relationships between services.

[0079] In step S130, the input sequence of candidate services is input into the analysis model to obtain the matching degree between the candidate services and customer needs.

[0080] The analytical model is a machine learning model used to evaluate the match between candidate services and customer needs. It can be pre-trained using user historical order data, interaction behavior logs, etc. In one possible implementation, the analytical model is trained based on the Transformer architecture. Of course, other types of models can also be used, such as multilayer perceptrons, gradient boosting trees, graph neural networks, or recurrent neural networks. This disclosure does not limit this approach.

[0081] In some embodiments of this disclosure, the analysis model includes an encoder and a scoring layer.

[0082] The encoder is used to perform deep semantic modeling on the input sequence containing customer needs and candidate service information, and extract the contextual associations and global dependencies between each element. The scoring layer calculates a scalar value based on the high-order feature representation output by the encoder through a mapping function, which serves as the matching score between the candidate service and the customer needs.

[0083] Figure 3 A flowchart illustrating the matching degree between candidate services and customer needs based on an analytical model is shown in an embodiment of this disclosure. (Refer to...) Figure 3 This may include the following steps.

[0084] Step S310: The input sequence of the candidate service is encoded by the encoder to obtain the global feature vector of the candidate service; the global feature vector of the candidate service is used to characterize the global dependency relationship between customer needs and candidate services.

[0085] In this step, the encoder receives the input sequence of candidate services, maps it to different subspaces through a multi-head self-attention mechanism to identify the global dependency between customer needs and candidate services, and generates a global feature vector. This vector comprehensively represents the degree to which the candidate services meet the customer needs as a whole.

[0086] Step S320: The matching degree of the global feature vector of the candidate service is calculated by the scoring layer to obtain the matching degree between the candidate service and the customer's needs.

[0087] In this step, the scoring layer takes the global feature vector obtained in step S310 as input and maps it to a score, or matching degree, through one or more fully connected neural network layers. This score represents the probability or preference strength of a candidate service in meeting customer needs. The higher the matching degree between a candidate service and customer needs, the better the candidate service meets customer expectations in terms of pricing rationality, resource coverage, benefit matching, and semantic consistency.

[0088] By introducing an analytical model that includes an encoder and a scoring layer, and by utilizing the model's self-attention mechanism and information about customer needs and candidate services in the input sequence, the matching relationship between customer needs and candidate services can be accurately characterized, thereby improving the accuracy of recommendation results.

[0089] In step S140, based on the matching degree between all candidate services in the candidate service list and the customer's needs, a service recommendation result for the customer's needs is obtained.

[0090] In some embodiments of this disclosure, a service recommendation result for the customer's needs is obtained based on the matching degree between all candidate services in the candidate service list and the customer's needs, including: sorting all candidate services and the customer's needs according to their matching degree, and selecting a preset number of candidate services with the highest matching degree; and using the selected candidate services as the service recommendation result for the customer's needs.

[0091] Sort all candidate services in the candidate service list according to their matching degree with customer needs from high to low; then set a threshold or fixed number, such as Top 3 or Top 5, based on business needs or interaction design, and select the top-ranked candidate services; finally, these selected services constitute the service recommendation results presented to the customer.

[0092] By ranking and filtering candidate services based on matching degree, the best recommendations that meet customer needs can be selected. This ensures the accuracy of recommendations while optimizing the user decision-making experience, avoiding interference from redundant information, and improving customer satisfaction.

[0093] Figure 4 A diagram illustrating the overall architecture of the service recommendation method according to an embodiment of this disclosure is shown. Figure 4The following flowchart illustrates the service recommendation method of this disclosure: First, based on customer demand information and service information of available services, a candidate service list is generated, where each candidate service in the list can be a single service or a combination of multiple services. Then, the customer demand information and each candidate service in the candidate service list are vectorized to obtain a corresponding customer demand vector and a service vector for each candidate service, thereby generating an input sequence for each candidate service. Next, the input sequence for each candidate service is input into an analysis model, which outputs the matching degree between each candidate service and the customer demand. Finally, all candidate services in the candidate service list are sorted according to the matching degree, and a recommendation list containing the Top-n candidate services is output, where n is a positive integer.

[0094] By constructing an input sequence that integrates structured features and semantic descriptions, and combining it with an analytical model to match candidate services (including single services or service combinations) with customer needs, high-precision and interpretable personalized recommendations are achieved. The entire process, from generating the candidate service list, vectorizing its representation, calculating the matching degree, to ranking and outputting the results, forms an end-to-end recommendation mechanism that effectively captures the interaction between customer needs and services, supporting flexible responses to diverse business scenarios. Furthermore, this method does not rely on historical user behavior data; for new customers or services without historical interaction data, it can also make effective recommendations based on their feature attributes and textual descriptions, thereby alleviating the cold start problem in service recommendation methods and improving the generalization ability and applicability of this method.

[0095] Figure 5 A schematic diagram of a service recommendation device according to an embodiment of the present disclosure is shown. Figure 5 The service recommendation device 500 shown may include a candidate service generation module 510, a sequence construction module 520, a matching degree determination module 530, and a recommendation module 540.

[0096] The candidate service generation module 510 is configured to filter available services based on customer needs and generate a candidate service list. The sequence construction module 520 is configured to construct an input sequence for each candidate service in the candidate service list, based on customer needs and the candidate service itself. The matching degree determination module 530 is configured to input the input sequence of the candidate services into the analysis model to obtain the matching degree between the candidate services and customer needs. The recommendation module 540 is configured to obtain service recommendation results based on the matching degree between all candidate services in the candidate service list and customer needs.

[0097] In some embodiments of this disclosure, a candidate service is a single service or a combination of services.

[0098] In some embodiments of this disclosure, the candidate service generation module 510 is further configured to: filter the available services according to customer requirements to obtain a service list that meets customer requirements; and prune the service list according to customer requirements to generate a candidate service list.

[0099] In some embodiments of this disclosure, the candidate service generation module 510 is further configured to: perform constraint relationship verification on the service combinations included in the service list based on preset inter-service constraint relationships, and remove service combinations that do not meet the inter-service constraint relationships.

[0100] In some embodiments of this disclosure, the service constraint relationship includes at least one of the following: service mutual exclusion relationship, service quantity limit constraint, and service dependency relationship.

[0101] In some embodiments of this disclosure, the sequence construction module 520 is further configured to: vectorize customer requirements to obtain a customer requirement vector; vectorize candidate services to obtain a service vector of the candidate services; and concatenate the customer requirement vector and the service vector of the candidate services to construct an input sequence of the candidate services.

[0102] In some embodiments of this disclosure, the customer demand vector includes a customer demand feature vector and a customer demand embedding vector. The sequence construction module 520 is further configured to: classify and standardize customer demands to obtain a customer demand feature vector; and perform semantic encoding processing on the descriptive information of customer demands based on a text embedding model to obtain a customer demand embedding vector.

[0103] In some embodiments of this disclosure, the service vector of a candidate service includes a service feature vector and a service embedding vector. The sequence construction module 520 is further configured to: classify and standardize the parameter information of the candidate service to obtain the service feature vector; and perform semantic encoding on the description information of the candidate service based on a text embedding model to obtain the service embedding vector.

[0104] In some embodiments of this disclosure, the sequence construction module 520 is further configured to: insert a special identifier vector between the customer demand vector and the service vector of the candidate service; the special identifier vector is a preset fixed vector used to separate the customer demand semantics and the candidate service semantics.

[0105] In some embodiments of this disclosure, the sequence construction module 520 is further configured to: determine the position information of each vector in the input sequence of the candidate service; encode the position information of the vector to obtain the position code of the vector; and use the position code of the vector to identify the order of the vector in the input sequence of the candidate service.

[0106] In some embodiments of this disclosure, the analysis model includes an encoder and a scoring layer. The matching degree determination module 530 is further configured to: encode the input sequence of candidate services using the encoder to obtain a global feature vector of the candidate services; the global feature vector of the candidate services is used to characterize the global dependency relationship between customer needs and candidate services; and the scoring layer calculates the matching degree between the global feature vector of the candidate services and the customer needs to obtain the matching degree between the candidate services and customer needs.

[0107] In some embodiments of this disclosure, the recommendation module 540 is further configured to: sort all candidate services and their matching degree with customer needs, filter out a preset number of candidate services with the highest matching degree, and use the filtered candidate services as the service recommendation results for customer needs.

[0108] The principle of the service recommendation device embodiment provided in this disclosure is similar to that of the method embodiment described above. Therefore, the real-time implementation of this service recommendation device embodiment can be found in the implementation of the method embodiment described above, and repeated details will not be repeated.

[0109] Figure 6 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown. It should be noted that... Figure 6 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0110] like Figure 6 As shown, the electronic device 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage section 608 into a random access memory (RAM) 603. The RAM 603 also stores various programs and data required for the operation of the electronic device 600. The CPU 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0111] The following components are connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. A removable medium 611, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 610 as needed so that computer programs read from it can be installed into storage section 608 as needed.

[0112] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 609, and / or installed from removable medium 611. When the computer program is executed by central processing unit (CPU) 601, it performs the functions defined above in the system of this disclosure.

[0113] It should be noted that the computer-readable medium disclosed herein may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, terminal device, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, terminal device, or device. In this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit programs for use by or in connection with an instruction execution system, terminal device, or apparatus. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0114] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0115] The modules described in the embodiments of this disclosure can be implemented in software or hardware. The described modules can also be housed in a processor; for example, a processor may be described as including a candidate service generation module, a sequence construction module, a matching degree determination module, and a recommendation module. The names of these modules do not necessarily limit the module itself; for example, the candidate service generation module may also be described as "a module that filters available services according to customer needs to generate a candidate service list."

[0116] In another aspect, this disclosure also provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The computer-readable storage medium carries one or more programs that, when executed by the electronic device, cause the electronic device to perform the methods described in the following embodiments. For example, the electronic device may perform... Figure 1 The steps shown.

[0117] According to one aspect of this disclosure, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in various alternative implementations of the above embodiments.

[0118] It should be understood that any number of elements in the accompanying drawings is for illustrative purposes only and not for limitation, and any naming is for distinction only and has no limiting meaning.

[0119] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0120] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A service recommendation method, characterized in that, The method includes: The available services are filtered based on customer needs to generate a list of candidate services; For each candidate service in the candidate service list, construct the input sequence of the candidate service based on the customer requirements and the candidate service; The input sequence of the candidate services is input into the analysis model to obtain the matching degree between the candidate services and the customer needs; Based on the matching degree between all candidate services in the candidate service list and the customer's needs, a service recommendation result for the customer's needs is obtained.

2. The method according to claim 1, characterized in that, The candidate service is a single service or a combination of services.

3. The method according to claim 1, characterized in that, The step of filtering available services based on customer needs to generate a candidate service list includes: The available services are filtered according to the customer's needs to obtain a list of services that meet the customer's needs. The service list is pruned according to the customer's requirements to generate the candidate service list.

4. The method according to claim 3, characterized in that, The method further includes: Based on the preset inter-service constraints, the service combinations included in the service list are validated for constraints, and service combinations that do not meet the inter-service constraints are removed.

5. The method according to claim 4, characterized in that, The service constraints include at least one of the following: service mutual exclusion, service quantity limit constraint, and service dependency.

6. The method according to claim 1, characterized in that, The step of constructing the input sequence of candidate services based on the customer needs and the candidate services includes: The customer requirements are vectorized to obtain a customer requirement vector; The candidate services are vectorized to obtain the service vectors of the candidate services; The customer demand vector and the service vector of the candidate service are concatenated to construct the input sequence of the candidate service.

7. The method according to claim 6, characterized in that, The customer demand vector includes a customer demand feature vector and a customer demand embedding vector; wherein, the process of vectorizing the customer demand to obtain the customer demand vector includes: The customer needs are classified and standardized to obtain the customer needs feature vector. The semantic encoding of the customer demand description information is performed based on a text embedding model to obtain the customer demand embedding vector.

8. The method according to claim 6, characterized in that, The service vector of the candidate service includes the service feature vector of the candidate service and the service embedding vector of the candidate service. The step of vectorizing the candidate services to obtain the service vectors of the candidate services includes: The parameter information of the candidate services is classified and standardized to obtain the service feature vector of the candidate services. The description information of the candidate services is semantically encoded based on a text embedding model to obtain the service embedding vector of the candidate services.

9. The method according to any one of claims 6 to 8, characterized in that, The method further includes: A special identifier vector is inserted between the customer demand vector and the service vector of the candidate service; the special identifier vector is a preset fixed vector used to separate the customer demand semantics from the candidate service semantics.

10. The method according to claim 9, characterized in that, The method further includes: For each vector in the input sequence of the candidate service, determine the position information of the vector; The position information of the vector is encoded to obtain the position code of the vector; the position code of the vector is used to identify the order of the vector in the input sequence of the candidate service.

11. The method according to claim 1, characterized in that, The analysis model includes an encoder and a scoring layer; wherein, the step of inputting the input sequence of the candidate service into the analysis model to obtain the matching degree between the candidate service and the customer's needs includes: The encoder encodes the input sequence of the candidate service to obtain a global feature vector of the candidate service; the global feature vector of the candidate service is used to characterize the global dependency relationship between the customer demand and the candidate service. The matching degree between the candidate service and the customer's needs is obtained by calculating the matching degree of the global feature vector of the candidate service through the scoring layer.

12. The method according to claim 1, characterized in that, The process of obtaining service recommendation results for the customer's needs based on the matching degree between all candidate services in the candidate service list and the customer's needs includes: Sort all candidate services by their matching degree with the customer's needs, and select a preset number of candidate services with the highest matching degree. The selected candidate services will be used as the service recommendations for the customer's needs.

13. A service recommendation device, characterized in that, The device includes: The candidate service generation module is configured to filter available services based on customer needs and generate a candidate service list. The sequence construction module is configured to construct an input sequence of the candidate services for each candidate service in the candidate service list, based on the customer requirements and the candidate services. The matching degree determination module is configured to input the input sequence of the candidate service into the analysis model to obtain the matching degree between the candidate service and the customer demand. The recommendation module is configured to obtain service recommendation results for the customer's needs based on the matching degree between all candidate services in the candidate service list and the customer's needs.

14. An electronic device, characterized in that, include: processor; as well as Memory for storing the executable instructions of the processor; The processor is configured to execute the method of any one of claims 1-12 by executing the executable instructions.

15. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1-12.

16. A computer program product, characterized in that, The computer program product stores instructions that, when executed by a computer, cause the computer to perform the method described in any one of claims 1-12.

Images