Order processing method, device and system

Abstract

The embodiment of the invention provides an order processing method, device and system, and the method comprises the steps: responding to a problem processing request for a target order, and determining the target order data of the target order; according to the target order data, in a historical problem case library, determining a historical problem case having an association relationship with the target order and case information of the historical problem case; determining a problem processing chain corresponding to the historical problem case according to the case information, and generating reference information corresponding to the target order according to the problem processing chain and the target order data; and inputting the reference information into the problem decision model to obtain a problem decision result corresponding to the target order output by the problem decision model. Reference is provided for problem processing through historical problem cases, and the problem processing efficiency and accuracy are improved.

Description

Technical Field

[0001] The embodiments in this specification relate to the field of artificial intelligence technology, and in particular to an order processing method, apparatus and system. Background Technology

[0002] With the continuous development of internet technology, online shopping has become widespread. Sellers can publish product information through a product information service system, while buyers can browse and purchase products through applications provided by the system. During online shopping through this system, buyers may encounter disagreements with sellers regarding purchased products. For example, a buyer might request a refund for a product, but the seller might refuse. Their differing demands regarding the same product can lead to a dispute, and the buyer can then initiate a refund dispute resolution process through the system. However, current technology often relies on manual intervention to handle disputes, resulting in long processing times and poor efficiency. Therefore, accurately and quickly resolving after-sales issues is a pressing problem that needs to be addressed. Summary of the Invention

[0003] In view of this, embodiments of this specification provide an order processing method. One or more embodiments of this specification also relate to an order processing apparatus, a computing device, a computer-readable storage medium, and a computer program product, to address technical deficiencies in the prior art.

[0004] According to a first aspect of the embodiments of this specification, an order processing method is provided, comprising: In response to a problem handling request for a target order, the target order data for the target order is determined; Based on the target order data, identify historical problem cases related to the target order in the historical problem case database, and obtain case information for those historical problem cases; Based on the case information, determine the problem handling chain corresponding to the historical problem case, and generate reference information corresponding to the target order based on the problem handling chain and the target order data; The reference information is input into the problem decision model to obtain the problem decision result corresponding to the target order output by the problem decision model.

[0005] According to a second aspect of the embodiments of this specification, an order processing method is provided, comprising: In response to an after-sales issue handling request for a target order, determine the target order data for the target order; Based on the target order data, identify historical after-sales problem cases related to the target order in the historical after-sales problem case database, and obtain case information for the historical after-sales problem cases; Based on the case information, determine the after-sales problem handling chain corresponding to the historical after-sales problem case, and generate reference information corresponding to the target order based on the after-sales problem handling chain and the target order data; The reference information is input into the problem decision model to obtain the after-sales problem decision result corresponding to the target order output by the problem decision model.

[0006] According to a third aspect of the embodiments of this specification, an order processing apparatus is provided, comprising: The determination module is configured to determine the target order data of the target order in response to a problem handling request for the target order; The recall module is configured to determine, based on the target order data, historical problem cases related to the target order in the historical problem case library, and case information of the historical problem cases; The generation module is configured to determine the problem processing chain corresponding to the historical problem case based on the case information, and generate reference information corresponding to the target order based on the problem processing chain and the target order data; The output module is configured to input the reference information into the problem decision model and obtain the problem decision result corresponding to the target order output by the problem decision model.

[0007] According to a fourth aspect of the embodiments of this specification, an order processing apparatus is provided, comprising: The determination module is configured to determine the target order data of the target order in response to an after-sales problem handling request for the target order; The recall module is configured to determine, based on the target order data, historical after-sales problem cases related to the target order in the historical after-sales problem case library, and case information of the historical after-sales problem cases; The generation module is configured to determine the after-sales problem handling chain corresponding to the historical after-sales problem case based on the case information, and generate reference information corresponding to the target order based on the after-sales problem handling chain and the target order data; The output module is configured to input the reference information into the problem decision model and obtain the after-sales problem decision result corresponding to the target order output by the problem decision model.

[0008] According to a fifth aspect of the embodiments of this specification, an order processing system is provided, the system comprising a client and a server, wherein... The client generates a problem handling request for the target order and sends the problem handling request to the server. The server, in response to the problem handling request, determines the target order data of the target order; based on the target order data, it identifies historical problem cases related to the target order in the historical problem case library, and case information of the historical problem cases; based on the case information, it determines the problem handling chain corresponding to the historical problem cases, and generates reference information corresponding to the target order based on the problem handling chain and the target order data; it inputs the reference information into the problem decision model to obtain the problem decision result corresponding to the target order output by the problem decision model; and it sends the problem decision result to the client.

[0009] According to a sixth aspect of the embodiments of this specification, a computing device is provided, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the above-described order processing method.

[0010] According to a seventh aspect of the embodiments of this specification, a computer-readable storage medium is provided that stores computer-executable instructions that, when executed by a processor, implement the steps of the order processing method described above.

[0011] According to an eighth aspect of the embodiments of this specification, a computer program product is provided, including a computer program or instructions that, when executed by a processor, implement the steps of the order processing method described above.

[0012] One embodiment of this specification implements the process of determining the target order data when handling a problem-solving request for a target order. Based on the target order data, historical problem cases and their related cases are identified in a historical problem case library. Subsequently, the problem-solving chain of the historical problem cases is determined based on the case information, clarifying the processing approach for each historical problem case. This allows reference information for the target order to be generated based on the problem-solving chain and the target order data, achieving the effect of providing a reference for the current problem-solving process using historical problem cases related to the target order. The reference information is then input into a problem decision model, enabling the model to process the target order's problem according to the reference information, improving problem-solving efficiency and accuracy. Attached Figure Description

[0013] Figure 1 A system framework diagram of an order processing system according to one embodiment of this specification is shown; Figure 2 A flowchart of an order processing method according to an embodiment of this specification is shown; Figure 3 This diagram illustrates a case library construction method according to an embodiment of this specification. Figure 4A A schematic diagram of a model decision process according to an embodiment of this specification is shown; Figure 4B A schematic diagram of a model decision logic provided according to an embodiment of this specification is shown; Figure 5 A flowchart illustrating the processing procedure of an order processing method according to an embodiment of this specification is shown. Figure 6 This specification shows a schematic diagram of the structure of an order processing apparatus according to one embodiment; Figure 7 A flowchart of an order processing method according to an embodiment of this specification is shown; Figure 8 This specification shows a schematic diagram of the structure of an order processing apparatus according to one embodiment; Figure 9 A structural block diagram of a computing device provided according to one embodiment of this specification is shown. Detailed Implementation

[0014] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0015] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a,” “described,” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0016] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0017] Furthermore, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0018] The technical solutions provided in this application can employ deep learning models with relatively large parameter scales. However, this large model is merely an example; this application does not limit the number of model parameters supported by the deep learning model used, aiming to meet actual needs. The deep learning models involved in this application can be artificial intelligence-based language models (LM) or multimodal models (MM).

[0019] First, the terms and concepts used in one or more embodiments of this specification will be explained.

[0020] Disputes: When consumers and merchants have disputes over after-sales requests and fail to reach an agreement, the platform is triggered to intervene in the mediation service scenario (e.g., a buyer applies for platform intervention after the merchant refuses to refund).

[0021] The judgment states that the platform, based on after-sales guidelines, evidence from both parties, and intelligent analysis results, provides dispute resolution suggestions, including refunds, compensation, and behavioral correction.

[0022] KBS (Knowledge Base System) is a knowledge base system that stores internal and external knowledge, announcements, rules, and other information about a system or platform. KBS can be applied to e-commerce platforms as an automated decision-making system based on preset rules and structured knowledge.

[0023] Currently, after-sales issues frequently arise between buyers and sellers on e-commerce platforms, requiring platform intervention for adjudication. Common methods include rule-based dispute resolution or manual dispute resolution. Rule-based dispute resolution relies on existing rules in the KBS (Knowledge Base System) for judgment. When a user's dispute meets the requirements of the relevant rules, the system's built-in solution is triggered. However, KBS cannot provide clear and reasonable solutions for complex disputes. This hinders timely resolution of disputes and reduces trust between buyers and sellers in the platform.

[0024] To address the aforementioned technical problems, this specification provides an order processing method. One or more embodiments of this specification also relate to an order processing apparatus, a computing device, a computer-readable storage medium, and a computer program product, which will be described in detail in the following embodiments.

[0025] See Figure 1 , Figure 1 This diagram illustrates a system framework diagram of an order processing system according to an embodiment of this specification, wherein the order processing system includes a client 102 and a server 104, wherein... The client 102 generates a problem handling request for the target order and sends the problem handling request to the server. The server 104, in response to the problem handling request, determines the target order data of the target order; based on the target order data, it determines historical problem cases related to the target order in the historical problem case library, and case information of the historical problem cases; based on the case information, it determines the problem handling chain corresponding to the historical problem cases, and generates reference information corresponding to the target order based on the problem handling chain and the target order data; it inputs the reference information into the problem decision model to obtain the problem decision result corresponding to the target order output by the problem decision model; and it sends the problem decision result to the client.

[0026] In this context, the client refers to the terminal device or application used by the user (buyer or seller) to interact with the system. Examples include e-commerce platform apps on mobile phones and web browsers. The client initiates the dispute resolution process. Users submit dispute requests (such as "return and refund" or "refund only") through the client interface and ultimately receive the system's judgment. The server, on the other hand, refers to a server cluster running in the cloud or data center, carrying the core project logic and data processing functions. The server is the execution core of the entire dispute resolution process. It receives requests from the client, invokes various modules (data processing, case retrieval, model reasoning, etc.) for calculation and decision-making, and returns the final result to the client.

[0027] In practical applications, a user initiates an action command, or problem-solving request, on the client side, marking the start of the process for handling an order issue or dispute. The problem-solving request may include the target order's order ID, the dispute type selected by the user (e.g., "quality issue," "misrepresentation"), a preliminary complaint description, and uploaded evidence (images, videos), etc. Upon receiving the problem-solving request, the server determines the target order data based on the request. The target order data consists of all or key information related to the specific order in question. This target order data may include the order ID, product information (name, category, price), buyer and seller IDs, logistics information, after-sales status, chat history summaries, and evidence text / metadata uploaded by the user (buyer and / or seller). This forms the basis for subsequent analysis and judgment.

[0028] In practice, after the server determines the target order data, it can recall historical dispute cases from the historical dispute case database based on the target order data to identify historical dispute cases related to the target order. The historical dispute case database can be understood as a database storing past dispute cases that have been adjudicated. Each case contains its original data (similar to "target order data") and the final judgment result. Related historical dispute cases can be understood as historical cases similar to the current target order in key characteristics, such as the same product category (both are "mobile phones") or the same refund reason (both are "water damage"). The case information of historical dispute cases can be understood as detailed data of the recalled historical dispute cases, including the order data, judgment result, and even the dispute resolution approach, i.e., the problem resolution chain, of that historical case.

[0029] In one embodiment of this specification, if a processing approach exists in the case information, the problem-solving chain can be directly obtained from the case information. If no processing approach exists in the case information, the case information can be analyzed and summarized to obtain the corresponding problem-solving chain. The problem-solving chain is not a simple judgment result, but rather the reasoning process and decision-making logic that leads to that result, used to guide the dispute resolution of the current target order. Based on the problem-solving chain and target order data, reference information for the target order can be generated. The reference information can be understood as prompts constructed to guide the large language model in reasoning, integrating "target order data" (current case details) and "problem-solving chain" (the thought chain of historical precedents), forming a complete context that includes role settings, reference examples, current descriptions, and task instructions. Subsequently, the reference information can be input into the problem decision model, which can be understood as a large language model that performs the final judgment reasoning; it is an AI model capable of complex reasoning. It receives "reference information" as input, imitates the reasoning methods in the "problem-solving chain" of historical cases, analyzes the current dispute, and generates judgment suggestions. The problem decision model outputs a problem decision result, which is a judgment recommendation on the dispute issue of the target order. The problem decision result can be a specific and actionable solution, such as: "Support the buyer to return the goods and get a refund, with the return postage borne by the seller" or "Support the seller and refuse the refund".

[0030] Based on this, the order processing system provided in this specification solves the pain point of traditional KBS systems being unable to adjudicate complex and ambiguous dispute cases, necessitating manual intervention. By recalling historical problem cases and generating reference information as prompts for the model based on the problem-solving chains of these historical cases, knowledge is accumulated and reused. The model's output decision results are fairer and more in line with the platform's past precedents, avoiding the subjective arbitrariness of manual judgments. By summarizing the judgment process through a large model, new cases refer to this judgment process rather than relying entirely on existing cases, and by combining product and industry-related knowledge, the accuracy of judgments is improved.

[0031] See Figure 2 , Figure 2 A flowchart of an order processing method according to an embodiment of this specification is shown, which specifically includes the following steps.

[0032] Step 202: In response to a problem handling request for the target order, determine the target order data for the target order.

[0033] In this context, a problem-handling request can be understood as an instruction initiated by a user (buyer or seller) on the e-commerce platform client, requesting after-sales service or dispute resolution for a specific order. It serves as the trigger signal for the entire intelligent adjudication process. Problem-handling requests can be initiated by either the buyer or the seller, and may include order identifiers, dispute types, appeal descriptions, supporting documentation, and user identity. The target order can be understood as the specific order for which the "problem-handling request" has been initiated. It is the object of processing in this dispute resolution process. An order may have multiple different problem-handling requests initiated during its lifecycle (e.g., first requesting a refund only, which is rejected, then requesting a return and refund). Target order data can be understood as all the original project data and contextual information related to the "target order" that can be used to determine the attribution of responsibility for the dispute. This is the foundational material for subsequent rule matching, case retrieval, and model inference. Target order data can originate from system platforms such as e-commerce system platforms, and may include basic order information, transaction status, user information, dispute context, and supporting documentation.

[0034] In practical applications, when determining the order data of a target order, it is necessary to preprocess the incoming data, such as integrating, deduplicating, and standardizing it to ensure that the data format is consistent and the quality meets the standards. Input is standardized through atomic capabilities such as feature recall and information completion. Determining the target order data includes: determining the original order data of the target order; and performing data preprocessing on the original order data to obtain the target order data of the target order.

[0035] Raw order data can be understood as a collection of raw, scattered, and heterogeneous data directly read from various project systems (orders, logistics, after-sales, communication, etc.) without any processing. Data preprocessing of raw order data can be understood as performing a series of cleaning, transformation, integration, and enhancement operations to convert it into a high-quality, structured, and standardized input suitable for subsequent model inference. After data preprocessing, the target order data is obtained.

[0036] In practice, data preprocessing can include data cleaning (such as handling missing values, outliers, and deduplication), data integration and fusion, data transformation and standardization, and feature extraction and information enhancement. Cleaned data reduces noise interference, resolves the issue of inconsistent data standards between different project systems, and significantly improves the performance of downstream tasks.

[0037] In a specific embodiment of this specification, a buyer purchases a mobile phone, discovers scratches on the screen after signing for it, and considers it a product defect. The buyer then opens the e-commerce platform to file a complaint and initiate a problem-solving request. During the process, the buyer can select the dispute type and provide evidence through the after-sales function. The client packages these complaint operations into a problem-solving request, which can then be sent to the server for processing. The server responds to the problem-solving request for the target order, determining the target order data, which may include data from multiple dimensions such as order details, logistics, after-sales service, and communication.

[0038] Step 204: Based on the target order data, determine the historical problem cases that are related to the target order in the historical problem case library, and the case information of the historical problem cases.

[0039] The historical case library can be understood as a database storing past adjudicated dispute cases. It serves as the system's "case law library" or "knowledge base." Each case includes its historical order data (similar to the current target order data) and the final judgment. These can be manually submitted "selected cases" or "high-quality cases" accumulated after automatic system judgments. Recalling cases from the historical case library can be understood as the process of quickly selecting a small subset of candidate cases most similar and relevant to the current target order from the vast "historical case library" according to certain strategies. Historical case cases related to the target order can be understood as cases where the current "target order" and a certain "historical case" share similarities in key characteristics, making the judgment experience of that historical case potentially applicable to the current situation.

[0040] In practical applications, the criteria for judging historical problem cases can include structured feature matching, semantic similarity, or comprehensive similarity. A historical problem case is a specific historical case that is recalled and has a "relationship" with the target order. The case information for a historical problem case is the detailed data package of the recalled case. This information may include historical order data, judgment results, and the judgment processing chain for that case.

[0041] In a specific embodiment of this specification, a buyer initiates after-sales service for a quality issue, citing the phone's inability to power on, and uploads a video of the phone's black screen. The server retrieves past issue cases from its historical problem case database based on the identified target order data, such as Case A: "A certain brand of mobile phone failed to power on 8 hours after receipt, with video evidence; judgment: return and refund supported." The problem handling chain states: "A high-end flagship phone experienced a functional failure in a very short time, and the buyer provided timely evidence; the responsibility is highly likely to lie in product quality or the manufacturing process, and the platform should tend to protect consumer rights." Based on this, by recalling related historical cases, we ensure that target orders with similar disputes to those historical cases are handled in a similar manner, thus avoiding the problem of "different judgments for the same case," enhancing users' trust in the platform, achieving "same judgment for similar cases," and improving the consistency and fairness of judgments.

[0042] Furthermore, the construction of the historical problem case library includes: determining case collection information, selecting a first problem case set from the original case library based on the case collection information; receiving a second problem case set reported by operations personnel, and constructing the historical problem case library based on the first problem case set and the second problem case set.

[0043] The original case library can be understood as the raw data warehouse of all completed dispute cases on the platform. It is a massive, unfiltered "case pool." The original case library contains complete data on all past adjudicated order disputes on the platform (whether manually or automatically), but it has not undergone quality assessment or knowledge extraction. Case collection information can be understood as a set of preset rules or standards used to automatically filter high-quality cases from the "original case library." Based on the case collection information, a first set of problematic cases can be selected from the original case library. This first set of problematic cases can be understood as a batch of high-quality, representative historical dispute cases automatically selected from the "original case library" based on the "case collection information." The first set of problematic cases has a large number of cases and a wide coverage, representing the "high-quality cases" in the case library.

[0044] Correspondingly, the second set of problematic cases can be understood as special or typical dispute cases proactively reported or submitted by platform operators, arbitration experts, or senior customer service personnel. While the number of problematic cases in the second set is small, their quality is extremely high and instructive, typically containing complex judgment logic or innovative handling methods. These problematic cases are considered "selected cases," including problem-solving chains written by operators—that is, the judgment logic filled in during manual dispute handling. After determining the first and second sets of problematic cases, a historical problematic case database can be constructed.

[0045] In a specific embodiment of this specification, see [link to embodiment]. Figure 3 , Figure 3This diagram illustrates a case library construction method according to an embodiment of this specification, wherein the case library is constructed by separately constructing a first problem case set and a second problem case set. In the second problem case set, operators or experts can submit selected cases via documents such as Excel or the system platform. Submitted cases undergo review to ensure their quality and representativeness. Approved cases are formally submitted to the system, and the relevant code or processing logic of the selected cases can be reused, improving efficiency. Manual submission aims to collect high-quality, representative dispute cases as a source for the "second problem case set," supplementing the deficiencies of automated screening.

[0046] The first issue case set is divided into full synchronization and incremental synchronization. Full synchronization refers to the cases that were cold-started before the first issue case set went live, while incremental synchronization refers to the cases added after the first issue case set went live, as well as cases updated due to project modifications. Full synchronization includes a full table, D2 tasks, and database synchronization. The full table extracts all case data that meets the criteria from the original case database; the D2 task packages all the data into a D2 task for subsequent data processing; and database synchronization synchronizes the data from the D2 task to the target database. Incremental synchronization includes a partitioned table, table generation messages, and a scheduler. The partitioned table manages the original case database by partitioning it for incremental updates; the table generation message generates a notification message when a new case is generated; and the scheduler, upon receiving the message, triggers an incremental synchronization task to add the new case to the case database. Adding the first and second issue case sets to the case database creates the historical issue case database.

[0047] Based on this, a historical problem case library is constructed by combining automatic screening with manual submission. This not only ensures the quality and diversity of the historical problem case library, but also provides a data foundation for subsequent problem processing chain generation and large model reasoning, thereby further improving the accuracy and efficiency of decision-making results.

[0048] Furthermore, before determining historical problem cases related to the target order and their case information in the historical problem case library based on the target order data, the method further includes: determining general domain knowledge corresponding to the target order data, and inputting the general domain knowledge and the target order data into the problem decision model; if the problem decision model outputs an initial problem decision result, using the initial problem decision result as the problem decision result corresponding to the target order; if the problem decision model does not output an initial problem decision result, performing the process of determining historical problem cases related to the target order and their case information in the historical problem case library based on the target order data.

[0049] In this context, general domain knowledge can be understood as universally applicable rules or common sense that are widely applicable within a specific project domain (such as e-commerce) and do not depend on specific cases. General domain knowledge can be stored in a KBS knowledge database system. For example, general domain knowledge could be "Buyers can apply for a return without reason within 7 days of receiving the goods (except for certain categories)". By inputting general domain knowledge and target order data into the problem decision model, it is possible to perform conventional model judgments on problem disputes of the target order and obtain the judgment results.

[0050] In practical applications, after receiving a user's issue handling request, the model can first combine industry knowledge to make a judgment and obtain a decision result. The model will output an initial issue decision result, which can be understood as a judgment suggestion directly output by the issue decision model based solely on "target order data" and "general domain knowledge". The initial issue decision result does not rely on specific historical cases recalled from the case library, making the reasoning process lighter and faster. In this case, the initial issue decision result can be directly used as the issue decision result for the target order. However, if the issue decision model only has general domain knowledge, it may not be able to output an initial issue decision result. In this case, the case recall step needs to be performed.

[0051] In a specific embodiment of this specification, see [link to embodiment]. Figure 4A , Figure 4A This diagram illustrates a model decision-making process according to an embodiment of this specification, where the problem-solving request is a user-initiated order dispute request. Upon receiving the problem-solving request, it is first input into the problem decision-making model. The model then uses industry-standard knowledge to perform decision analysis. If the analysis is successful based on industry-standard knowledge, a decision result is directly output. If analysis is not possible, a suitable tool is selected for case recall. Based on the recalled cases, the problem-solving chain, i.e., the decision-making logic, is determined. Reference information is generated based on the tool results and dynamic feature analysis, and returned to the problem decision-making model. The problem decision-making model then re-determines the decision based on the context prompts and outputs the final decision result.

[0052] In a specific embodiment of this specification, the problem decision model will provide a problem decision result in practical applications. However, this problem decision result will carry a corresponding confidence level. Only when the confidence level reaches a preset threshold can the result be regarded as the initial problem decision result, that is, the problem decision model is considered to have output an initial problem decision result. Otherwise, it is considered that the model has not output an initial problem decision result.

[0053] Based on this, by combining rule-based judgment and case-driven judgment, case-driven judgment is only used when rule-based judgment cannot solve the problem. That is, the solution path is dynamically selected according to the complexity of the problem, making the entire intelligent dispute resolution system more mature, robust and efficient.

[0054] Furthermore, based on the target order data, historical problem cases related to the target order are identified in the historical problem case library, along with case information for those historical problem cases. This includes: performing conditional recall in the historical problem case library based on the target order data, and / or performing similarity recall in the historical problem case library based on the target order data; and determining historical problem cases related to the target order based on the recall results, along with case information for those historical problem cases.

[0055] Conditional recall can be understood as a recall method that uses structured, discrete, or interval-based features in the "target order data" to perform precise matching or range filtering in the "historical problem case library." Recall criteria can include product type, dispute cause, etc. Similarity recall can be understood as converting unstructured text information (such as dispute descriptions or chat logs) in the "target order data" into vectors, and then recalling the semantically most similar cases by calculating the distance between vectors (such as cosine similarity) in the vector index of the "historical problem case library."

[0056] In practical applications, you can choose to use conditional recall or similarity recall. You can also use both conditional recall and similarity recall simultaneously by performing conditional recall and similarity recall separately, and then merging (removing duplicates) the two recall sets.

[0057] In one specific embodiment of this specification, a conditional recall is performed in the historical problem case library based on the target order data, and the recall result is "Case A, B, and C". A similarity recall is performed in the historical problem case library based on the target order data, and the recall result is "Case A and D". Therefore, the final recall result is "Case A, B, C, and D".

[0058] Based on this, a hybrid strategy of conditional recall and similarity recall is adopted to ensure that subsequent large models can perform high-quality inference, thereby further improving the accuracy of the model's inference results.

[0059] Furthermore, conditional recall is performed in the historical problem case library based on the target order data, including: determining case recall conditions based on the target order data; and selecting historical problem cases in the historical problem case library based on the case recall conditions.

[0060] Case recall criteria can be understood as a set of structured query parameters or logical expressions used to filter candidate cases from the "historical problem case library". It serves as the filter or selection criterion for the "conditional recall" operation. From the processed "target order data", a set of structured fields that best represent the characteristics of the current dispute is identified, extracted, and combined to form a "condition set" that can be used for database queries as case recall criteria.

[0061] In practical applications, selecting historical and problem cases from the historical problem case database based on case recall conditions can be understood as submitting the "case recall conditions" as a query statement to the database or search engine of the "historical problem case database" to perform data retrieval operations and return all case records that meet the conditions.

[0062] In a specific embodiment of this specification, the case recall conditions are determined based on the target order data. The case recall conditions are "Product Category: Home Appliances - Vacuum Cleaner...". Based on the case recall conditions, historical problem cases related to "vacuum cleaner" are recalled from the historical problem case database.

[0063] Based on this, conditional recall is performed using case recall criteria to achieve efficient and accurate filtering, providing data support for subsequent model inference.

[0064] Furthermore, similarity retrieval based on the target order data in the historical problem case library includes: extracting features from the target order data to obtain the order feature vector corresponding to the target order data; determining the case feature vector in the historical problem case library, calculating feature similarity based on the case feature vector and the order feature vector, and selecting historical problem cases from the historical problem case library based on the feature similarity.

[0065] Feature extraction can be understood as identifying and extracting key textual information from the "target order data" for calculating semantic similarity, and using it as input to the embedding model. The order feature vector can be understood as a high-dimensional numerical array encoded from unstructured textual information (such as dispute descriptions, chat logs, and product titles) in the target order data using a specific embedding model. The case feature vector can be understood as the feature vector corresponding to each historical case in the historical problem case library.

[0066] In practical applications, when building the case library, the system pre-vectorizes the core textual information of each historical case (such as dispute descriptions and judgment summaries) and stores the results (vectors) along with the case ID in a vector database. Feature similarity can be understood as the mathematical distance between an "order feature vector" and a certain "case feature vector," used to measure the semantic similarity between two cases. Calculation methods can include cosine similarity or Euclidean distance. Based on feature similarity, historical problem cases can be selected from the historical problem case library.

[0067] In practice, after calculating the feature similarity, the historical problem cases in the historical problem case library can be sorted in descending order according to the feature similarity, and the Top-K cases with the highest scores can be selected as the recall results.

[0068] In a specific embodiment of this specification, the system extracts key text information from the target order data, concatenates the text into a query text, and uses a text embedding model to convert the query text into a feature vector, thereby obtaining the order feature vector. Based on the order feature vector and the case feature vector of each case in the historical problem case library, feature similarity is calculated. Based on each feature similarity, several historical problem cases with higher scores are selected from the historical problem case library.

[0069] Based on this, similarity-based recall overcomes the limitations of keyword matching, achieving semantic-level recall. The recalled cases are highly similar to the current case in essence, and their "problem-processing chain" has extremely high inspirational value for large-scale models. This provides high-quality "analogical reasoning" material for large-scale models.

[0070] Step 206: Determine the problem handling chain corresponding to the historical problem case based on the case information, and generate reference information corresponding to the target order based on the problem handling chain and the target order data.

[0071] The problem-solving chain can be understood as the structured decision-making logic and reasoning path extracted from the judgment process of "historical problem cases." The problem-solving chain is not simply a judgment result, but rather the complete thought process behind why that judgment was made. It can include information such as problem analysis, attribution judgment, rule application, responsibility determination, and the basis for the judgment. The reference information can be understood as a core component of the input prompt (Prompt) constructed to generate the final judgment recommendation, containing context and examples. The reference information typically includes a summary of key information from the recalled historical problem case, the problem-solving chain for that case, and the final judgment result of that case.

[0072] Furthermore, determining the problem handling chain corresponding to the historical problem case based on the case information includes: if the historical problem case is of the manually reported type, determining the problem handling chain corresponding to the historical problem case in the case information; if the historical problem case is of the timed collection type, inputting the case information into the case analysis model to obtain the problem handling chain corresponding to the historical problem case output by the case analysis model.

[0073] The "manually submitted" type can be understood as cases that were proactively submitted, reviewed, and added to the database by operations personnel, arbitration experts, or senior customer service staff. The manually submitted type refers to the selected cases from the second set of problem cases in the above embodiments. The "timely collected" type can be understood as cases that were selected and added to the database from the "original case library" through automated system tasks (such as daily timed tasks) based on preset rules (such as automatic judgment based on high confidence, no complaints). The timely collected type refers to the high-quality cases from the first set of problem cases in the above embodiments.

[0074] In practical applications, the methods for determining the problem-solving chain differ depending on the type of historical problem case. Since manually reported historical problem cases are proactively reported and recorded by operations personnel, these cases already contain manually summarized judgment approaches, allowing the problem-solving chain to be directly determined from the case information. However, historical problem cases collected on a scheduled basis are selected based on specific criteria; therefore, the case information for this type of historical problem case does not contain a problem-solving chain, requiring further analysis using a model.

[0075] In practice, a case analysis model can be used to analyze the problem-solving chain of historical problem cases collected periodically. The case analysis model can be understood as a large language model or model system specifically designed to automatically extract and generate the "problem-solving chain" from the raw data (such as order information, chat logs, vouchers, and judgment results) of historical cases collected periodically. Its input is case information, and its output is a structured natural language text describing the judgment logic of the case, i.e., the "problem-solving chain."

[0076] In a specific embodiment of this specification, the type of historical problem cases is determined to be the manual reporting type. During the reporting process, the operators submit the judgment reasoning or cause together to form case information. Therefore, the problem handling chain can be directly obtained from the case information of the case.

[0077] In another specific embodiment of this specification, the type of historical problem cases is determined to be the timed collection type. The case information of the case is input into the case analysis model, and the case analysis model is used to analyze it to obtain the problem handling chain of the case. For example, "A high-end flagship mobile phone developed a serious functional failure that could not be turned on within a very short time (within 2 hours) after being signed for, and the buyer has provided valid video evidence, ruling out the possibility of damage during logistics or improper use by the user. This situation is highly likely to point to a defect in product quality or the manufacturing process. In order to protect consumer rights and the platform's reputation, returns and refunds should be supported." Based on this, different problem processing chain determination methods are adopted for different types of historical problem cases. For "manually reported" cases, existing knowledge is directly read, resulting in zero computational overhead. For "timed collection" cases, although the model needs to be invoked, it can be processed in batches, and the model inference can be optimized. This optimizes system resources and processing efficiency.

[0078] Furthermore, generating reference information corresponding to the target order based on the problem processing chain and the target order data includes: parsing the problem processing chain and the target order data to obtain order prompt information; and using the order prompt information as reference information for the target order.

[0079] The parsing of the problem processing chain and target order data can be understood as the process of extracting, formatting, and reorganizing information from the unstructured "problem processing chain" text and the structured "target order data." The goal is to transform the messy information into a clear, concise, and machine-friendly format for efficient construction of the final Prompt. Order prompt information can be understood as an intermediate product generated after the "parsing" step, directly used to construct the Large Model (LLM) input prompt. Order prompt information is a structured representation or semi-finished product of the "reference information." It already possesses the core content required for the final Prompt, but may not yet be fully assembled into natural language paragraphs. Specifically, it can take the form of key-value pairs, JSON objects, or text blocks divided by role. In subsequent processes where order prompt information is used as reference information, it can be populated into a predefined Prompt template, or the order prompt information itself can be used directly as reference information.

[0080] In a specific embodiment of this specification, the problem handling chain is defined as follows: "For high-end brand vacuum cleaners, if a complaint of 'weak suction' arises in a short period of time, and the buyer can provide video evidence of insufficient performance without any signs of improper use, it should be considered a product quality issue, and the consumer is supported in returning the product and receiving a refund." The problem handling chain and target order data are parsed to obtain order notification information. This order notification information can be a JSON object containing all necessary contextual information. The system calls a "Prompt Generator" module, which reads the "order notification information" and renders it into a piece of natural language text. "Reference Case A: Situation: A high-end brand vacuum cleaner experienced a short-term decrease in suction power, and the buyer provided video evidence. Handling logic: Considered a product quality issue. Judgment: Return and refund supported."

[0081] Current order: Product: Vacuum cleaner. Problem: Suction power is extremely weak 3 days after receipt. Evidence: Video available. Request: Return and refund.

[0082] Instruction: Based on the processing logic of the reference case, analyze the current order situation and provide a suggestion and reason for whether to support a return and refund. This text is the final "Reference Information" and is appended to the complete Prompt of the large model.

[0083] Based on this, order notification information is obtained by parsing the problem processing chain and target order data. Standardized information extraction and formatting are performed to avoid formatting chaos, information omissions, or redundancy caused by manual assembly. This ensures the high quality and consistency of reference information. Subsequently, using the order notification information as reference information allows the model to grasp key information and reasoning logic more quickly and accurately during inference, reducing misunderstandings caused by information redundancy or unclear expression, thereby generating better decision suggestions.

[0084] Step 208: Input the reference information into the problem decision model to obtain the problem decision result corresponding to the target order output by the problem decision model.

[0085] The problem decision model, specifically the Large Language Model (LLM), is used to generate the final judgment recommendation. It can be a domain-specific, fine-tuned model. The problem decision model can understand complex contexts, such as the case background and reasoning logic provided in the reference information; perform analogical reasoning, such as applying the problem-solving chain of the reference case to the current new case; and generate natural language output, such as outputting a clearly structured and well-reasoned judgment recommendation, thereby increasing the user's trust in the judgment result.

[0086] In practical applications, inputting reference information into the problem decision model yields the problem decision result output by the model. This problem decision result is the final judgment suggestion generated by the model.

[0087] For specific implementation details, please refer to the model's decision logic. Figure 4B , Figure 4B A schematic diagram of a model decision logic provided according to an embodiment of this specification is shown. Figure 4B The steps in the process include: Define the problem scenario: First, it is necessary to clearly define and understand the specific circumstances of the problem or dispute you are facing.

[0088] Check if there is a mutually agreed-upon solution: If there is a mutually agreed-upon solution ("Yes"), proceed directly to the evaluation platform solution; if there is no mutually agreed-upon solution ("No"), information correction is required to ensure that all relevant information is accurate.

[0089] Information correction: Correcting inaccurate or incomplete information to ensure the accuracy of subsequent judgments.

[0090] Check for merchant shortcomings: If any shortcomings are found, directly evaluate the platform's solution; if no shortcomings are found, review historical case reference information.

[0091] View historical case reference information: If historical cases are not available, view the general scenario judgment ideas to derive a solution; if historical cases are available, derive a solution by referring to historical cases.

[0092] Review the general decision-making process for common scenarios: Develop solutions based on general processing experience and logic.

[0093] Platform evaluation solution: Based on the conclusions drawn from the previous steps, evaluate and determine the platform's processing solution.

[0094] Furthermore, the method also includes: generating target problem cases corresponding to the target order based on the target order data and the problem decision results; and storing the target problem cases in the historical problem case library.

[0095] In this context, a target problem case can be understood as a new, complete, and historical case that can be used for future reference, consisting of the current "target order" and its final "problem decision result." The case information for a target problem case may include raw data, problem description, decision-making process, and decision result. Storing the target problem case and its information in a historical problem case database allows it to serve as reference information to assist model reasoning during subsequent problem judgments.

[0096] In practical applications, target order data and problem decision results can be integrated, formatted, and standardized, and encapsulated into a complete case object conforming to the data structure of the "historical problem case library". This allows the "target problem case" to be persisted as a new record to the "historical problem case library" database.

[0097] In a specific embodiment of this specification, the system integrates the target order data and the problem decision results to generate a structured target problem case. The generated target problem case is written into the database of the historical problem case library via API, and its type is determined to be the timed collection type, that is, the case is a selected case.

[0098] Based on this, by rewriting completed cases as problem cases and adding them to the historical problem case library, a positive feedback loop is formed, thereby improving the overall system performance.

[0099] This specification provides an order processing method, comprising: responding to a problem handling request for a target order; determining target order data for the target order; identifying historical problem cases related to the target order and their case information from a historical problem case library based on the target order data; determining a problem handling chain corresponding to the historical problem cases based on the case information; generating reference information for the target order based on the problem handling chain and the target order data; inputting the reference information into a problem decision model to obtain a problem decision result for the target order output by the problem decision model. This method achieves the following: when processing a problem handling request for a target order, the method determines the target order data for the target order; it identifies historical problem cases related to the target order and their cases from a historical problem case library based on the target order data; it then determines the problem handling chain for the historical problem cases based on the case information, clarifying the processing approach for the historical problem cases, enabling the generation of reference information for the target order based on the problem handling chain and the target order data, thus providing a reference for the current problem handling through historical problem cases related to the target order. By using reference information as input to the problem decision model, the model can process the problems of the target order according to the reference information, thereby improving the efficiency and accuracy of problem processing.

[0100] The following is in conjunction with the appendix Figure 5 Taking the application of the order processing method provided in this specification on an online shopping platform as an example, the order processing method will be further explained. Figure 5 The present specification illustrates a flowchart of an order processing method according to an embodiment, which includes the following steps.

[0101] Step 502: In response to the issue handling request for the target order, determine the target order data for the target order.

[0102] In one instance, a user purchased a mobile phone on a shopping platform, only to find it wouldn't turn on after receiving it. The user uploaded a video showing the phone failing to respond to repeated presses of the power button. The user initiated a customer service request for a refund. The seller argued that the video was insufficient to support a refund request and requested the platform's intervention to resolve the after-sales dispute.

[0103] Step 504: Determine the general domain knowledge corresponding to the target order data, and input the general domain knowledge and the target order data into the problem decision model.

[0104] In one embodiment, the order processing system of the online shopping platform determines the corresponding "general domain knowledge" based on the key features in the target order data (product category = mobile phone, problem = unable to turn on, time = within 2 hours of signing for receipt, evidence = video). The "general domain knowledge" is merged with the "target order data" to construct a preliminary Prompt, which is then input into the "problem decision model".

[0105] Step 506: If the problem decision model does not output the initial problem decision result, perform the following steps: based on the target order data, determine the historical problem cases that are related to the target order in the historical problem case library, and the case information of the historical problem cases.

[0106] In one embodiment, after receiving input, the "problem decision model" analyzes the data and concludes that it cannot make a high-confidence deterministic decision based solely on "general domain knowledge" and basic data (e.g., the model outputs "uncertain, more reference cases are needed" or no output at all). Therefore, a recall process is triggered.

[0107] Step 508: Based on the target order data, identify historical problem cases related to the target order in the historical problem case library, and the case information of the historical problem cases.

[0108] In one embodiment, case recall conditions are determined based on target order data, and historical problem cases are selected from the historical problem case library based on these conditions. The system determines case recall conditions, such as product category = "mobile phone", dispute reason = "cannot be turned on", etc. Historical problem cases and their corresponding case information are then selected based on these conditions.

[0109] In one embodiment, feature extraction is performed on the target order data to obtain the order feature vector corresponding to the target order data. Case feature vectors in the historical problem case library are determined, and feature similarity is calculated based on the case feature vectors and the order feature vectors. Historical problem cases are then selected from the historical problem case library based on the feature similarity. The system extracts features from the target order data and encodes them as order feature vectors. The vector database calculates the cosine similarity between the order feature vector and all case feature vectors in the library, and selects historical problem cases and their corresponding case information based on the cosine similarity.

[0110] Step 510: Determine the problem handling chain corresponding to the historical problem cases based on the case information.

[0111] In one embodiment, when the historical problem cases are manually reported, the problem handling chain corresponding to the historical problem cases is determined in the case information, such as "the mobile phone cannot be turned on within a very short time after signing for receipt, and the buyer provides valid video, excluding logistics damage and improper use by the user, it is determined to be a product quality problem, and a return and refund are supported".

[0112] In one embodiment, when the historical problem cases are collected periodically, the case information is input into the case analysis model to obtain the problem handling chain corresponding to the historical problem cases output by the case analysis model, such as "the mobile phone cannot be turned on within a very short time after signing for receipt, and the buyer provides valid video, excluding logistics damage and improper use by the user, it is determined to be a product quality problem, and a return and refund are supported."

[0113] Step 512: Analyze the problem processing chain and target order data to obtain order prompt information, and use the order prompt information as reference information for the target order.

[0114] In one embodiment, the system extracts and structures information from the "problem handling chain" and "target order data" to obtain order notification information such as "sunmary: mobile phone cannot be turned on 1 hour after receipt + video; reasoning: determined to be a product quality issue..." etc. This order notification information is used as reference information.

[0115] Step 514: Input the reference information into the problem decision model to obtain the problem decision result corresponding to the target order output by the problem decision model.

[0116] In one embodiment, the system concatenates the "reference information" into a standard instruction Prompt to form a complete input, which is then sent to the "problem decision model." The model generates and returns a "Judgment: Support return and refund. Reason: The current order is for a high-value flagship mobile phone, which developed a serious functional failure of "unable to power on" within 2 hours of receipt, and the buyer has provided valid video evidence. Referring to the handling logic of historical cases A and C (high-end mobile phones), this situation should be considered a product quality issue. To protect consumer rights, the buyer's request for return and refund is supported."

[0117] One embodiment of this specification implements the process of determining the target order data when handling a problem-solving request for a target order. Based on the target order data, historical problem cases and their related cases are identified in a historical problem case library. Subsequently, the problem-solving chain of the historical problem cases is determined based on the case information, clarifying the processing approach for each historical problem case. This allows reference information for the target order to be generated based on the problem-solving chain and the target order data, achieving the effect of providing a reference for the current problem-solving process using historical problem cases related to the target order. The reference information is then input into a problem decision model, enabling the model to process the target order's problem according to the reference information, improving problem-solving efficiency and accuracy.

[0118] Corresponding to the above method embodiments, this specification also provides embodiments of an order processing apparatus. Figure 6 A schematic diagram of an order processing apparatus according to one embodiment of this specification is shown. Figure 6 As shown, the device includes: The determination module 602 is configured to determine the target order data of the target order in response to a problem handling request for the target order; The recall module 604 is configured to determine, based on the target order data, historical problem cases related to the target order in the historical problem case library, and case information of the historical problem cases; The generation module 606 is configured to determine the problem processing chain corresponding to the historical problem case based on the case information, and generate reference information corresponding to the target order based on the problem processing chain and the target order data; The output module 608 is configured to input the reference information into the problem decision model and obtain the problem decision result corresponding to the target order output by the problem decision model.

[0119] Optionally, the method further includes a decision module configured to determine the general domain knowledge corresponding to the target order data, and input the general domain knowledge and the target order data into the problem decision model; if the problem decision model outputs an initial problem decision result, the initial problem decision result is used as the problem decision result corresponding to the target order; if the problem decision model does not output an initial problem decision result, historical problem cases related to the target order are determined in the historical problem case library based on the target order data, and the case information of the historical problem cases is also determined.

[0120] Optionally, the recall module 604 is further configured to perform conditional recall in the historical problem case library based on the target order data, and / or perform similarity recall in the historical problem case library based on the target order data; and determine historical problem cases that are associated with the target order and case information of the historical problem cases based on the recall results.

[0121] Optionally, the recall module 604 is further configured to determine case recall conditions based on the target order data; and select historical problem cases from the historical problem case library based on the case recall conditions.

[0122] Optionally, the recall module 604 is further configured to extract features from the target order data to obtain the order feature vector corresponding to the target order data; determine the case feature vector in the historical problem case library; calculate the feature similarity based on the case feature vector and the order feature vector; and select a historical problem case from the historical problem case library according to the feature similarity.

[0123] Optionally, the generation module 606 is further configured to, when the historical problem case is of the manually reported type, determine the problem processing chain corresponding to the historical problem case in the case information; and when the historical problem case is of the timed collection type, input the case information into the case analysis model to obtain the problem processing chain corresponding to the historical problem case output by the case analysis model.

[0124] Optionally, the generation module 606 is further configured to parse the problem processing chain and the target order data to obtain order prompt information; and use the order prompt information as reference information for the target order.

[0125] Optionally, the method further includes a construction module configured to determine case collection information, select a first set of problem cases from the original case library based on the case collection information, receive a second set of problem cases reported by operations personnel, and construct the historical problem case library based on the first set of problem cases and the second set of problem cases.

[0126] Optionally, the method further includes a storage module configured to generate target problem cases corresponding to the target order based on the target order data and the problem decision results; and store the target problem cases in the historical problem case library.

[0127] This specification provides an order processing apparatus, comprising: a determining module configured to determine target order data of the target order in response to a problem processing request for a target order; a recall module configured to determine, based on the target order data, historical problem cases related to the target order in a historical problem case library, and case information of the historical problem cases; a generating module configured to determine a problem processing chain corresponding to the historical problem cases based on the case information, and generate reference information corresponding to the target order based on the problem processing chain and the target order data; and an output module configured to input the reference information into a problem decision model to obtain a problem decision result corresponding to the target order output by the problem decision model. This achieves the following: when processing a problem processing request for a target order, the determining of the target order data of the target order is implemented. Based on the target order data, historical problem cases related to the target order and their corresponding cases are determined in the historical problem case library. Subsequently, the problem processing chain of the historical problem cases is determined based on the case information, clarifying the processing approach of the historical problem cases. This allows reference information for the target order to be generated based on the problem processing chain and the target order data, achieving the effect of providing a reference for the current problem processing through historical problem cases related to the target order. By using reference information as input to the problem decision model, the model can process the problems of the target order according to the reference information, thereby improving the efficiency and accuracy of problem processing.

[0128] The above is an illustrative scheme of an order processing device according to this embodiment. It should be noted that the technical solution of this order processing device and the technical solution of the order processing method described above belong to the same concept. For details not described in detail in the technical solution of the order processing device, please refer to the description of the technical solution of the order processing method described above.

[0129] See Figure 7 , Figure 7 A flowchart of an order processing method according to an embodiment of this specification is shown, which specifically includes the following steps.

[0130] Step 702: In response to an after-sales issue handling request for the target order, determine the target order data for the target order.

[0131] Step 704: Based on the target order data, determine the historical after-sales problem cases that are related to the target order in the historical after-sales problem case library, and the case information of the historical after-sales problem cases.

[0132] Step 706: Determine the after-sales problem handling chain corresponding to the historical after-sales problem case based on the case information, and generate reference information corresponding to the target order based on the after-sales problem handling chain and the target order data.

[0133] Step 708: Input the reference information into the problem decision model to obtain the after-sales problem decision result corresponding to the target order output by the problem decision model.

[0134] In this context, after-sales issue handling requests can be understood as requests initiated by buyers after order completion, involving services such as returns, repairs, refunds only, and replacements. The historical after-sales issue case library can be understood as a knowledge base specifically storing resolved, high-quality after-sales dispute cases. The after-sales issue handling chain can be understood as the decision-making logic and reasoning path extracted from historical after-sales cases regarding how to determine responsibility, which after-sales policy to apply, and how to ultimately implement it. Reference information can be understood as input prompts constructed to generate final after-sales judgment suggestions for the large model; its core is "reusable after-sales decision examples." The after-sales issue decision result can be understood as the specific handling plan and rationale output by the problem decision model for the current after-sales request.

[0135] This specification provides an order processing method that, when processing after-sales issue handling requests for a target order, first identifies the target order data. Based on the target order data, it identifies historical after-sales issue cases and their related cases from a historical after-sales issue case library. Subsequently, it determines the issue handling chain of these historical after-sales issue cases based on the case information, clarifying the processing approach for each case. This allows reference information for the target order to be generated based on the issue handling chain and the target order data, providing a reference for handling the current after-sales issue using historical issue cases related to the target order. This reference information is then input into a problem decision model, enabling the model to process the after-sales issue of the target order according to the reference information, improving the efficiency and accuracy of after-sales issue handling.

[0136] Corresponding to the above method embodiments, this specification also provides embodiments of an order processing apparatus. Figure 8 A schematic diagram of an order processing apparatus according to one embodiment of this specification is shown. Figure 8As shown, the device includes: The determination module 802 is configured to determine the target order data of the target order in response to an after-sales problem handling request for the target order; The recall module 804 is configured to determine, based on the target order data, historical after-sales problem cases related to the target order in the historical after-sales problem case library, and case information of the historical after-sales problem cases; The generation module 806 is configured to determine the after-sales problem handling chain corresponding to the historical after-sales problem case based on the case information, and generate reference information corresponding to the target order based on the after-sales problem handling chain and the target order data; The output module 808 is configured to input the reference information into the problem decision model and obtain the after-sales problem decision result corresponding to the target order output by the problem decision model.

[0137] This specification provides an order processing device that, when processing after-sales issue handling requests for a target order, first identifies the target order data. Based on the target order data, it identifies historical after-sales issue cases and their related cases from a historical after-sales issue case library. Subsequently, it determines the issue handling chain of these historical after-sales issue cases based on the case information, clarifying the handling approach for each case. This allows reference information for the target order to be generated based on the issue handling chain and the target order data, providing a reference for handling the current after-sales issue using historical issue cases related to the target order. The reference information is then input into a problem decision model, enabling the model to handle the after-sales issue of the target order according to the reference information, improving the efficiency and accuracy of after-sales issue handling.

[0138] Figure 9 A structural block diagram of a computing device 900 according to one embodiment of this specification is shown. The components of the computing device 900 include, but are not limited to, a memory 910 and a processor 920. The processor 920 is connected to the memory 910 via a bus 930, and a database 950 is used to store data.

[0139] The computing device 900 also includes an access device 940, which enables the computing device 900 to communicate via one or more networks 960. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 940 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a Near Field Communication (NFC) interface.

[0140] In one embodiment of this specification, the above-described components of the computing device 900 and Figure 9 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 9 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0141] The computing device 900 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 900 can also be a mobile or stationary server.

[0142] The processor 920 is used to execute the following computer-executable instructions, which, when executed by the processor, implement the steps of the above-described order processing method.

[0143] The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device and the technical solution of the order processing method described above belong to the same concept. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the order processing method described above.

[0144] An embodiment of this specification also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the above-described order processing method.

[0145] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium and the technical solution of the order processing method described above belong to the same concept. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the order processing method described above.

[0146] An embodiment of this specification also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the steps of the order processing method described above.

[0147] The above is an illustrative scheme of a computer program product according to this embodiment. It should be noted that the technical solution of this computer program product and the technical solution of the order processing method described above belong to the same concept. For details not described in detail in the technical solution of the computer program product, please refer to the description of the technical solution of the order processing method described above.

[0148] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0149] The computer program / instructions include computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added or removed according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0150] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, because according to the embodiments in this specification, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.

[0151] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0152] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. An order processing method, comprising: In response to a problem handling request for a target order, the target order data for the target order is determined; Based on the target order data, identify historical problem cases related to the target order in the historical problem case database, and obtain case information for those historical problem cases; Based on the case information, determine the problem handling chain corresponding to the historical problem case, and generate reference information corresponding to the target order based on the problem handling chain and the target order data; The reference information is input into the problem decision model to obtain the problem decision result corresponding to the target order output by the problem decision model.

2. The method according to claim 1, before determining the historical problem cases related to the target order and the case information of the historical problem cases in the historical problem case library based on the target order data, the method further includes: Determine the general domain knowledge corresponding to the target order data, and input the general domain knowledge and the target order data into the problem decision model; When the problem decision model outputs an initial problem decision result, the initial problem decision result is used as the problem decision result corresponding to the target order; If the problem decision model does not output an initial problem decision result, the system performs the following steps: based on the target order data, it determines historical problem cases that are related to the target order in the historical problem case library, and obtains case information for those historical problem cases.

3. The method according to claim 1, wherein, based on the target order data, historical problem cases related to the target order are determined in a historical problem case database, and case information of the historical problem cases is obtained, including: Conditional recall is performed in the historical problem case library based on the target order data, and / or similarity recall is performed in the historical problem case library based on the target order data; Based on the recall results, identify historical problem cases that are associated with the target order, and the case information of the historical problem cases.

4. The method according to claim 1, wherein conditional recall is performed in the historical problem case database based on the target order data, comprising: Determine the case recall conditions based on the target order data; Historical problem cases are selected from the historical problem case library based on the aforementioned case recall criteria.

5. The method according to claim 1, wherein similarity retrieval is performed in a historical problem case database based on the target order data, comprising: Feature extraction is performed on the target order data to obtain the order feature vector corresponding to the target order data; Determine the case feature vectors in the historical problem case library, calculate the feature similarity based on the case feature vectors and the order feature vectors, and select historical problem cases from the historical problem case library according to the feature similarity.

6. The method according to claim 1, wherein determining the problem processing chain corresponding to the historical problem case based on the case information includes: If the historical problem case is of the manually submitted type, the problem processing chain corresponding to the historical problem case is determined in the case information; When the historical problem cases are collected periodically, the case information is input into the case analysis model to obtain the problem processing chain corresponding to the historical problem cases output by the case analysis model.

7. The method according to claim 1, wherein generating reference information corresponding to the target order based on the problem processing chain and the target order data includes: The problem processing chain and the target order data are parsed to obtain order notification information; The order notification information is used as reference information for the target order.

8. The method according to any one of claims 1-7, wherein the construction of the historical problem case library comprises: Determine the case collection information, and select the first set of problem cases from the original case library based on the case collection information; Receive the second set of problem cases reported by the operations staff, and construct the historical problem case library based on the first set of problem cases and the second set of problem cases.

9. The method according to any one of claims 1-7, further comprising: Generate target problem cases corresponding to the target orders based on target order data and problem decision results; The target problem case is stored in the historical problem case library.

10. An order processing method, comprising: In response to an after-sales issue handling request for a target order, determine the target order data for the target order; Based on the target order data, identify historical after-sales problem cases related to the target order in the historical after-sales problem case database, and obtain case information for the historical after-sales problem cases; Based on the case information, determine the after-sales problem handling chain corresponding to the historical after-sales problem case, and generate reference information corresponding to the target order based on the after-sales problem handling chain and the target order data; The reference information is input into the problem decision model to obtain the after-sales problem decision result corresponding to the target order output by the problem decision model.

11. An order processing apparatus, comprising: The determination module is configured to determine the target order data of the target order in response to a problem handling request for the target order; The recall module is configured to determine, based on the target order data, historical problem cases related to the target order in the historical problem case library, and case information of the historical problem cases; The generation module is configured to determine the problem processing chain corresponding to the historical problem case based on the case information, and generate reference information corresponding to the target order based on the problem processing chain and the target order data; The output module is configured to input the reference information into the problem decision model and obtain the problem decision result corresponding to the target order output by the problem decision model.

12. An order processing apparatus, comprising: The determination module is configured to determine the target order data of the target order in response to an after-sales problem handling request for the target order; The recall module is configured to determine, based on the target order data, historical after-sales problem cases related to the target order in the historical after-sales problem case library, and case information of the historical after-sales problem cases; The generation module is configured to determine the after-sales problem handling chain corresponding to the historical after-sales problem case based on the case information, and generate reference information corresponding to the target order based on the after-sales problem handling chain and the target order data; The output module is configured to input the reference information into the problem decision model and obtain the after-sales problem decision result corresponding to the target order output by the problem decision model.

13. An order processing system, the system comprising a client and a server, wherein, The client generates a problem handling request for the target order and sends the problem handling request to the server. The server, in response to the problem handling request, determines the target order data for the target order; Based on the target order data, identify historical problem cases related to the target order in the historical problem case database, and obtain case information for those historical problem cases; Based on the case information, determine the problem handling chain corresponding to the historical problem case, and generate reference information corresponding to the target order based on the problem handling chain and the target order data; The reference information is input into the problem decision model to obtain the problem decision result corresponding to the target order output by the problem decision model; The decision result of the problem is sent to the client.

14. A computing device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the order processing method according to any one of claims 1 to 10.

15. A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the order processing method according to any one of claims 1 to 10.

16. A computer program product comprising a computer program or instructions which, when executed by a processor, implement the steps of the order processing method according to any one of claims 1 to 10.

Images