Information processing method and device, electronic equipment and storage medium

Abstract

This disclosure relates to an information processing method, apparatus, electronic device, and storage medium. The method includes: responding to object association issue information of a target account for a target interactive object in an intelligent conversation page, obtaining intelligent conversation information and a case library from the intelligent conversation page; determining the target object exception type corresponding to the object association issue information based on the intelligent conversation information and the case library; determining structured exception result description information based on the target object exception type, object exception logs, object processing sequence information of the target interactive object, and object identification information of the target interactive object; recalling exception solutions based on the exception result description information to obtain a target exception solution; and invoking an intelligent response model to generate exception resolution response information for the object association issue information based on the exception result description information and the target exception solution. This makes the exception resolution response information more accurate and improves the operability of exception handling.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, and in particular to an information processing method, apparatus, electronic device and storage medium. Background Technology

[0002] With the proliferation of online platforms, online platform-based interactions are becoming increasingly common, such as product interactions on e-commerce platforms. In related technologies, for example, during the order fulfillment process on e-commerce platforms, merchants may encounter anomalies such as failed shipments or delayed packages, typically requiring them to seek solutions from the platform's customer service. However, due to varying capabilities of human customer service representatives, resolving these issues often involves long processing times, significant manpower investment, and inaccurate problem identification. This results in inaccurate responses to anomaly inquiries and poor operability in handling these issues. Summary of the Invention

[0003] This disclosure provides an information processing method, apparatus, electronic device, and storage medium to at least address issues in related technologies such as how to improve the accuracy and operability of responses to unusual inquiries. The technical solution of this disclosure is as follows: According to a first aspect of the present disclosure, an information processing method is provided, comprising: In response to the object association issue information of the target account for the target interaction object in the smart conversation page, the smart conversation information and case library in the smart conversation page are obtained. The case library is used to store the correspondence between object issues and object exception types. The target interaction object refers to the object generated by the target account through business interaction. Based on the intelligent session information and the case library, determine the target object anomaly type corresponding to the object-related problem information; Based on the target object's exception type, object exception log, object processing timing information of the target interactive object, and object identification information of the target interactive object, structured exception result description information is determined. Based on the description information of the abnormal results, an abnormal solution is recalled to obtain a target abnormal solution; the target abnormal solution refers to a solution overview used to resolve the object-related problem information. The intelligent response model is invoked to generate an anomaly resolution response information for the object-related problem information based on the anomaly result description information and the target anomaly solution; the anomaly resolution response information includes an overview of the solution and specific processing information of the solution corresponding to the solution overview.

[0004] In one possible implementation, determining the target object exception type corresponding to the object-associated problem information based on the intelligent session information and the case library includes: Based on the intelligent session information, the case library is queried to obtain the object anomaly type that matches the intelligent session information; Based on a large language model, fine-grained anomaly analysis is performed on the intelligent session information and the object anomaly type matched by the intelligent session information to obtain the target object anomaly type; the target object anomaly type is used to indicate the object processing flow node of the anomaly under the object anomaly type matched by the intelligent session information.

[0005] In one possible implementation, the case library is used to store the correspondence between the problem vector corresponding to the object problem and the object exception type vector corresponding to the object exception type; the step of querying the case library based on the smart session information to obtain the object exception type matched by the smart session information includes: The intelligent conversation information is transformed into a feature vector to obtain a conversation vector; Based on the session vector, the case library is queried to obtain the object anomaly type that matches the intelligent session information.

[0006] In one possible implementation, determining structured exception result description information based on the target object exception type, object exception log, object processing timing information of the target interactive object, and object identification information of the target interactive object includes: Extract historical interaction exception information corresponding to the object identifier information of the target interactive object from the object exception log; The abnormal result description information is determined based on the target object's abnormality type, the historical interaction abnormality information, and the object processing timing information of the target interaction object.

[0007] In one possible implementation, the target interaction object is a target order; determining the exception result description information based on the exception type of the target object, the historical interaction exception information, and the object processing sequence information of the target interaction object includes: In the case where the exception type of the target object is order delivery node failure, the exception result description information is determined based on the historical interaction exception information; Alternatively, if the exception type of the target object is an object processing flow node exception other than the shipping node failure, the exception result description information is determined based on the historical interaction exception information and the object processing sequence information of the target interaction object.

[0008] In one possible implementation, determining the exception result description information based on the target object exception type, the historical interaction exception information, and the object processing timing information of the target interaction object includes: The large language model is invoked to perform anomaly result analysis on the anomaly type of the target object, the historical interaction anomaly information, and the object processing timing information of the target interaction object, and generate the anomaly result description information.

[0009] In one possible implementation, the invocation of the intelligent response model generates exception resolution response information for the object-related problem information based on the exception result description information and the target exception solution, including: Based on the description of the abnormal result and the target abnormal solution, construct response prompt words; The intelligent reply model is invoked, and the reply prompt words are input into the intelligent reply model to generate reply information, thereby obtaining the abnormal resolution reply information for the object-related problem information.

[0010] In one possible implementation, the method further includes: Obtain reply restriction information, preset question-and-answer data pairs, and reply reasoning methods; The step of constructing response prompt words based on the abnormal result description information and the target abnormal solution includes: The reply prompt words are constructed based on the description of the abnormal result, the target abnormal solution, the reply restriction information, the preset question-and-answer data pair set, and the reply reasoning method.

[0011] In one possible implementation, the method further includes: The error resolution response information is evaluated to obtain the evaluation results; The anomaly resolution response information is adjusted based on the evaluation results to obtain the adjusted anomaly resolution response information; Based on the object-related question information and the adjusted exception resolution response information, a new question-and-answer data pair is constructed; The new question-and-answer data pairs are stored in the preset question-and-answer data pair set.

[0012] In one possible implementation, the method further includes: The error resolution response information is evaluated to obtain the evaluation results; The strategy for constructing the response prompts will be adjusted based on the evaluation results.

[0013] In one possible implementation, the step of recalling anomaly solutions based on the anomaly result description information to obtain a target anomaly solution includes: Based on the data structure of the anomaly knowledge base, the description information of the anomaly results is rewritten to obtain the query statement for the anomaly solution; The query statement is used to search for an anomaly solution in the anomaly knowledge base to obtain the target anomaly solution.

[0014] In one possible implementation, the method further includes: If there is an anomaly in the object processing timing information of the target interactive object on the object details page, intelligent session guidance will be performed. In response to the target account entering the smart conversation page based on the smart conversation guidance, the smart conversation page showing the target account conducting a smart conversation with the smart agent is displayed to the target account, and at least one object-related question in a preset question format is pushed to the smart conversation page based on the object identification information; When the target account selects a target object association question from the at least one object association question, the agent is invoked to generate the object association question information carrying the object identification information based on the preset question format, and the target account displays the object association question information for the target interactive object on the smart conversation page.

[0015] In one possible implementation, when the object-associated problem information indicates a response to the cause of the problem, the anomaly resolution response information also includes the anomaly result description information and the anomaly cause mining information generated by the intelligent response model based on the anomaly result description information.

[0016] According to a second aspect of the present disclosure, an information processing apparatus is provided, comprising: The information acquisition module is configured to respond to the target account's object association question information for the target interactive object in the smart conversation page, acquire the smart conversation information and case library in the smart conversation page, the case library being used to store the correspondence between object questions and object exception types; the target interactive object refers to the object generated by the target account through business interaction. The object anomaly type determination module is configured to determine the target object anomaly type corresponding to the object-related problem information based on the intelligent session information and the case library. The abnormal result description module is configured to determine structured abnormal result description information based on the abnormal type of the target object, the object abnormal log, the object processing timing information of the target interactive object, and the object identification information of the target interactive object. The exception solution recall module is configured to recall exception solutions based on the exception result description information to obtain a target exception solution; the target exception solution refers to a solution overview for resolving the object-related problem information. The exception resolution response information generation module is configured to execute and invoke the intelligent response model to generate exception resolution response information for the object-related problem information based on the exception result description information and the target exception solution; the exception resolution response information includes an overview of the solution and specific processing information of the solution corresponding to the solution overview.

[0017] In one possible implementation, the object exception type determination module includes: The object exception type determination unit is configured to perform a query in the case library based on the smart session information to obtain the object exception type that matches the smart session information. The target object anomaly type acquisition unit is configured to perform fine-grained anomaly analysis based on a large language model on the intelligent session information and the object anomaly type matched by the intelligent session information to obtain the target object anomaly type; the target object anomaly type is used to indicate the object processing flow node of the anomaly under the object anomaly type matched by the intelligent session information.

[0018] In one possible implementation, the case library is used to store the correspondence between the problem vector corresponding to the object problem and the object exception type vector corresponding to the object exception type; the object exception type determination unit includes: The session vector acquisition subunit is configured to perform feature vector transformation on the intelligent session information to obtain a session vector. The object exception type determination subunit is configured to perform a query in the case library based on the session vector to obtain the object exception type that matches the intelligent session information.

[0019] In one possible implementation, the abnormal result description module includes: The historical interaction anomaly information extraction unit is configured to extract historical interaction anomaly information corresponding to the object identifier information of the target interaction object from the object anomaly log. The abnormal result description unit is configured to determine the abnormal result description information based on the abnormal type of the target object, the historical interaction abnormal information, and the object processing timing information of the target interaction object.

[0020] In one possible implementation, the target interaction object is a target order; the abnormal result description unit includes: The first abnormal result description subunit is configured to determine the abnormal result description information based on the historical interaction abnormal information when the abnormal type of the target object is order delivery node failure. Alternatively, the second abnormal result description subunit is configured to determine the abnormal result description information based on the historical interaction abnormal information and the object processing sequence information of the target interaction object when the abnormality type of the target object is an object processing flow node abnormality other than the shipping node failure.

[0021] In one possible implementation, the abnormal result description unit includes: The third abnormal result description subunit is configured to execute the large language model to perform abnormal result analysis on the abnormal type of the target object, the historical interaction abnormal information, and the object processing timing information of the target interaction object, and generate the abnormal result description information.

[0022] In one possible implementation, the exception resolution response information generation module includes: The response prompt word construction unit is configured to construct response prompt words based on the abnormal result description information and the target abnormal solution; The exception resolution response information generation unit is configured to execute the call to the intelligent response model, input the response prompt words into the intelligent response model to generate response information, and obtain the exception resolution response information for the object-related problem information.

[0023] In one possible implementation, the device further includes: The prompt word association information acquisition module is configured to acquire reply restriction information, a preset set of question-and-answer data pairs, and reply reasoning methods; The reply prompt word construction unit is further configured to construct the reply prompt words based on the abnormal result description information, the target abnormal solution, the reply restriction information, the preset question-and-answer data pair set, and the reply reasoning method.

[0024] In one possible implementation, the device further includes: The response information evaluation module is configured to evaluate the exception resolution response information and obtain the evaluation results. The exception resolution response information adjustment module is configured to adjust the exception resolution response information based on the evaluation results to obtain the adjusted exception resolution response information; The question-and-answer data pair construction module is configured to construct new question-and-answer data pairs based on the object-associated question information and the adjusted exception resolution response information; The question-and-answer data pair set update module is configured to store the new question-and-answer data pair into the preset question-and-answer data pair set.

[0025] In one possible implementation, the device further includes: The response information evaluation module is configured to evaluate the exception resolution response information and obtain the evaluation results. The prompt word construction strategy adjustment module is configured to execute a strategy for adjusting the construction of the response prompt words based on the evaluation results.

[0026] In one possible implementation, the exception solution recall module includes: The query statement acquisition unit is configured to execute a data structure based on the exception knowledge base, rewrite the exception result description information, and obtain a query statement for the exception solution; The exception solution query unit is configured to perform an exception solution query in the exception knowledge base according to the query statement to obtain the target exception solution.

[0027] In one possible implementation, the device further includes: The intelligent conversation guidance module is configured to perform intelligent conversation guidance when there is an anomaly in the object processing timing information of the target interactive object in the object details page of the target interactive object. The object association question push module is configured to respond to the target account entering the smart conversation page based on the smart conversation guidance, display the smart conversation page where the target account and the smart agent are having a smart conversation to the target account, and push at least one object association question in a preset question format to the smart conversation page based on the object identification information; The object association question information processing module is configured to, when the target account selects a target object association question from the at least one object association question, call the intelligent agent to generate the object association question information carrying the object identification information based on the preset question format, and display the object association question information of the target account for the target interactive object on the intelligent conversation page.

[0028] In one possible implementation, when the object-associated problem information indicates a response to the cause of the problem, the anomaly resolution response information also includes the anomaly result description information and the anomaly cause mining information generated by the intelligent response model based on the anomaly result description information.

[0029] According to a third aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the method as described in any one of the first aspects above.

[0030] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided such that, when instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform any of the methods described in the first aspect of the present disclosure.

[0031] According to a fifth aspect of the present disclosure, a computer program product is provided, including computer instructions that, when executed by a processor, cause a computer to perform the method described in any one of the first aspects of the present disclosure.

[0032] The technical solutions provided by the embodiments of this disclosure have at least the following beneficial effects: By responding to the object-related issue information of the target account for the target interaction object in the intelligent conversation page, the system obtains the intelligent conversation information and case library in the intelligent conversation page. The case library is used to store the correspondence between object issues and object exception types. The target interaction object refers to the object generated by the target account through business interaction. Based on the intelligent conversation information and the case library, the target object exception type corresponding to the object-related issue information is determined. Based on the target object exception type, object exception logs, object processing sequence information of the target interaction object, and object identification information of the target interaction object, structured exception result description information is determined. This approach combines historical object exception logs with object processing sequence information including the current state, making the exception result description information more accurate. Furthermore, setting the exception result description information to a structured format further improves the accuracy of the exception result description information in describing the exception issue, thus providing a foundation for the accuracy of subsequent exception resolution responses.

[0033] Furthermore, based on the abnormal result description information, abnormal solution recall is performed to obtain a target abnormal solution. The target abnormal solution refers to a solution overview for resolving the object-related problem information, making the recall efficiency of abnormal solutions higher. An intelligent response model is then invoked to generate abnormal solution response information for the object-related problem information based on the abnormal result description information and the target abnormal solution. The abnormal solution response information includes the solution overview and the specific processing information corresponding to the solution overview. This results in more accurate abnormal responses and stronger operability in abnormal handling. This ensures that the abnormal solution response information not only includes a solution overview, allowing the target account to quickly locate the abnormal problem, but also includes the specific processing information corresponding to the solution overview, improving the operability of abnormal problem handling.

[0034] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0035] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure, and are not intended to unduly limit this disclosure.

[0036] Figure 1 This is a schematic diagram illustrating an application environment according to an exemplary embodiment.

[0037] Figure 2 This is a flowchart illustrating an information processing method according to an exemplary embodiment.

[0038] Figure 3a This is a schematic diagram illustrating a smart conversation page according to an exemplary embodiment.

[0039] Figure 3b This is a schematic diagram illustrating intelligent session guidance information and an intelligent session page according to an exemplary embodiment.

[0040] Figure 4 This is a flowchart illustrating, according to an exemplary embodiment, an evaluation of exception resolution response information and an update of a preset set of question-and-answer data pairs based on the evaluation results.

[0041] Figure 5 This is a schematic diagram illustrating an exception resolution response message according to an exemplary embodiment.

[0042] Figure 6 This is a schematic diagram of the flow architecture of an information processing method according to an exemplary embodiment.

[0043] Figure 7 This is a block diagram of an information processing apparatus according to an exemplary embodiment.

[0044] Figure 8 This is a block diagram illustrating an electronic device for information processing according to an exemplary embodiment.

[0045] Figure 9 This is a block diagram illustrating an electronic device for information processing based on an exemplary embodiment. Detailed Implementation

[0046] To enable those skilled in the art to better understand the technical solutions of this disclosure, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings.

[0047] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0048] 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 display, data used for analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties. Furthermore, in the specific embodiments of this disclosure, when user-related data is involved, user permission or consent is required when the following embodiments of this disclosure are applied to specific products or technologies, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0049] Artificial intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or computers-controlled machines to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. AI software technology mainly includes computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0050] In recent years, with the research and progress of artificial intelligence technology, artificial intelligence technology has been widely used in many fields. The solutions provided in this application involve technologies such as natural language processing, which are specifically illustrated through the following embodiments.

[0051] Please see Figure 1 , Figure 1 This is a schematic diagram illustrating an application environment according to an exemplary embodiment, such as... Figure 1 As shown, the application environment may include server 01 and terminal 02.

[0052] In an optional embodiment, terminal 02 can be used for services such as information processing. Specifically, terminal 02 can be, but is not limited to, electronic devices such as smartphones, desktop computers, tablets, laptops, smart speakers, digital assistants, augmented reality (AR) / virtual reality (VR) devices, and smart wearable devices. Optionally, the operating system running on the electronic device can be, but is not limited to, Android, iOS, Linux, and Windows.

[0053] In an optional embodiment, server 01 can provide background services for terminal 02. Specifically, server 01 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0054] In addition, it should be noted that, Figure 1 The example shown is merely one application environment of the information processing method provided in this disclosure. For instance, information processing can be performed by server 01, while terminal 02 provides front-end interactive support such as intelligent conversation pages; or information processing can be performed collaboratively by server 01 and terminal 02.

[0055] In the embodiments described in this specification, the server 01 and the terminal 02 can be directly or indirectly connected through wired or wireless communication, and this application does not impose any restrictions on this.

[0056] Figure 2 This is a flowchart illustrating an information processing method according to an exemplary embodiment. For example... Figure 2 As shown, the information processing method may include the following steps.

[0057] In step S201, in response to the object association problem information of the target account for the target interactive object in the smart conversation page, the smart conversation information and case library in the smart conversation page are obtained. The case library is used to store the correspondence between object problems and object exception types.

[0058] In the embodiments of this specification, the intelligent conversation page can refer to a page used for conversations between an account and an intelligent agent. The intelligent agent can refer to a processing assistant based on AI technology, such as one capable of conversations, generating object-related question information, etc., but this disclosure does not limit this. For example, the intelligent conversation page can be as follows: Figure 3a As shown, or as Figure 3b As shown in the middle and right figures, this disclosure does not limit the scope of the information.

[0059] In the embodiments of this specification, the aforementioned target interaction object can refer to an object (or interaction object) generated by the target account through business interaction. Business interaction can refer to the interaction of business within the target platform (e.g., an online platform). For example, business can include, but is not limited to, online sales of goods (products), online purchase of goods, return services, approval services, etc. Accordingly, the object generated through business interaction can include, but is not limited to, product orders, approval events, etc., which are not limited in this disclosure. Among them, products can include physical products or virtual products. Physical products can include products obtained online based on virtual resource interaction, such as clothes purchased from an e-commerce platform, or meals ordered. Virtual products can include services, e-books, etc., which are not limited in this disclosure.

[0060] For example, the target platform may include an e-commerce platform, which can refer to an online platform for conducting transactions, such as a platform that provides transaction services over the internet. Alternatively, the target platform may be an enterprise's approval management platform. This disclosure does not limit this.

[0061] In the embodiments described in this specification, the target account can refer to an account on a target platform. For example, if the target platform is an e-commerce platform, the target account can be a target merchant (any merchant) on that e-commerce platform. When the target merchant sells a certain product, a corresponding product order (which can be simply referred to as an order, i.e., the target interaction object) will be generated.

[0062] In the embodiments of this specification, object-related problem information may refer to information used to describe problems related to the target interactive object, that is, information used for consulting on related problems of the target interactive object.

[0063] In one example, object-related question information can be conversation information (or conversation messages) sent by the target account in the smart conversation page, addressing a question asked by the target interactive object. In other words, object-related question information can be manually entered and sent by the target account in the smart conversation page. For example... Figure 3a As shown in 31.

[0064] In another example, the object-related question information can be generated by calling the aforementioned intelligent agent based on a preset question format. This makes the question description clearer and facilitates subsequent parsing and processing. This disclosure does not limit the preset question format. For example, taking the target interaction object as the target order, the preset question format can be as follows.

[0065] Order XXX ( / / order process node) failed: Order XXX ( / / order number) failed. Please help me locate the cause and provide a solution.

[0066] In the embodiments of this specification, the smart conversation information in the smart conversation page can refer to the information in the smart conversation messages sent by the target account on the smart conversation page. For example, it can be conversation information related to the target interaction object. For example, the smart conversation page can be triggered based on a target order, and the smart conversation information in the smart conversation page is related to the target order. Alternatively, the smart conversation page also retains historical conversation information, and the conversation information can be understood based on a large language model to identify smart conversation information related to the target order, such as based on the order number, the semantics of the conversation messages, and the relationship between the conversation messages. This disclosure does not limit this.

[0067] In the embodiments of this specification, the case library can be used to store the correspondence between object problems and object exception types. For example, this correspondence can be preset and can be updated periodically; this disclosure does not limit this. In one example, the object problem can be object problem information in a preset question format, or it can be object problem description information manually entered by the user, or it can be extracted from a large amount of historical interaction object question consultation data; this disclosure does not limit this.

[0068] In a specific embodiment, the object exception type can be a type that is pre-classified or summarized for exceptions of the interactive object. For example, exception types can be divided into one or more levels based on the exception situations of the interactive object. Here, multiple levels of exception types can mean that the exception type can have multiple levels, such as a coarse-grained (or coarse-grained level) exception type, and each coarse-grained exception type is further subdivided into further subdivided exception types (or fine-grained levels). Taking an interactive object as an example of an order, the coarse-grained exception type can include order fulfillment exceptions, order after-sales exceptions, etc.; the further subdivision of exception types under each coarse-grained exception type can refer to the process node exception type of the processing flow node corresponding to each coarse-grained exception type, which is not limited in this disclosure.

[0069] In one example, the processing flow nodes corresponding to order fulfillment may include the process from order creation to order delivery, such as including but not limited to: order creation, order payment, order confirmation, inventory preparation, shipment, logistics and distribution (which may include logistics information such as pickup and transit), and order delivery, etc., which are not limited in this disclosure. Correspondingly, the process node exception types under the order fulfillment exception type may include the process node exception types corresponding to the above-mentioned multiple processing flow nodes.

[0070] In one possible implementation, the above method may further include: If there is an anomaly in the object processing timing information of the target interactive object on the object details page (e.g., the order details page of the target order), intelligent session guidance will be performed. For example, the object processing timing information can refer to the processing timing information generated by the execution of the aforementioned processing flow nodes corresponding to the interactive object, arranged in the order of execution. Anomalies in the object processing timing information can be obtained by the intelligent agent based on the parsing of the object processing timing information, or they can be identified by the platform based on a preset detection strategy.

[0071] In one example, intelligent conversation guidance can be displayed in a way that provides intelligent conversation guidance information, such as a pop-up window. Figure 3b As shown in 35 of the left-hand diagram, the smart session guidance information can be used to guide users into the smart session page. For example, the smart session guidance information can be as follows: Figure 3b As shown in Figure 34 on the left. Optionally, the object exception type, exception cause, interactive controls, etc., corresponding to the existence of an exception in the object processing timing information may also be displayed, but this disclosure does not limit this.

[0072] Next, in response to the target account entering the smart conversation page based on smart conversation guidance (e.g., clicking), the smart conversation page showing the target account engaging in a smart conversation with the agent can be displayed to the target account. Based on the object identification information, at least one object-related question in a preset question format can be pushed to the smart conversation page, for example... Figure 3a As shown in 32, or as Figure 3b As shown in 33.

[0073] Reference Figure 3a and Figure 3b As an example, the object details page of the target interactive object can be displayed, for example... Figure 3b The left-hand image shows the process before the intelligent session guidance information is displayed. The object details page displays the object processing sequence information of the target interactive object. If there is an anomaly in the object processing sequence information, the intelligent session guidance information can be displayed; this is in response to the target account's triggering of the intelligent session guidance information, such as clicking... Figure 3b 34 in the diagram can display the smart conversation page where the target account engages in a smart conversation with the smart agent, for example... Figure 3b As shown in the intermediate diagram, or as... Figure 3a As shown in the diagram on the right.

[0074] Furthermore, when the target account selects a target object association question from the at least one object association question, the intelligent agent is invoked to generate object association question information carrying object identification information based on a preset question format, and the target account displays the object association question information for the target interactive object on the intelligent conversation page, for example... Figure 3b As shown in 36. For example, if the target account clicks... Figure 3b 33 in the middle can be displayed Figure 3b As shown in the image on the right, the smart conversation page displays object association questions sent by the target account for the target interaction object, such as... Figure 3b As shown in 36.

[0075] In cases where there are anomalies in the object processing sequence information of the target interactive object on the object details page, intelligent conversation guidance can be provided, and at least one object-related question in a preset question format can be pushed. This makes the object-related question information for the target interactive object more accurate, thereby providing a foundation for the accuracy of subsequent determination of the target object's anomaly type and the accuracy of generating anomaly resolution response information.

[0076] In step S203, the target object anomaly type corresponding to the object-related problem information is determined based on the intelligent session information and the case library.

[0077] In one possible implementation, the intelligent session information can be used to query a case library to obtain the object anomaly type that matches the intelligent session information. For example, the similarity between the intelligent session information and object questions in the case library can be calculated, and the object anomaly type corresponding to the object question with the highest similarity can be determined as the target object anomaly type corresponding to the object association question information.

[0078] Optionally, the intelligent session information can be preprocessed before querying, such as filtering out duplicate or irrelevant information and extracting question keywords, so as to query the case library based on the preprocessed intelligent session information and obtain the matching object exception type.

[0079] In another possible implementation, the object exception types in the case library are the coarser-grained exception types mentioned above. Taking orders as an example, object exception types can include order fulfillment exceptions, order after-sales exceptions, etc. Based on this, the above-mentioned determination of the target object exception type corresponding to the object-related problem information according to the intelligent session information and the case library can include: Based on the intelligent session information, a query is performed in the case library to obtain the object anomaly type that matches the intelligent session information. For details, please refer to the relevant content above, which will not be repeated here. The matched object anomaly type can be one of the coarser-grained anomaly types mentioned above, such as order fulfillment anomaly.

[0080] Furthermore, based on a large language model, fine-grained anomaly analysis can be performed on the intelligent conversation information and the object anomaly types matched by the intelligent conversation information to obtain the target object anomaly type. This target object anomaly type can be used to indicate the abnormal object processing flow nodes under the aforementioned object anomaly type matched by the intelligent conversation information. In other words, it is possible to further determine which specific object processing flow node is abnormal under the object anomaly type. For example, in the case of order fulfillment anomaly, the anomaly of the object processing flow nodes included in order fulfillment can be located, such as shipping failure.

[0081] By querying the case library using intelligent conversation information, the abnormal object type matched by the intelligent conversation information is obtained. Based on the large language model, fine-grained anomaly analysis is performed on the intelligent conversation information and the abnormal object type matched by the intelligent conversation information to obtain the target object anomaly type. This target object anomaly type can be used to indicate the abnormal object processing flow node under the above-mentioned abnormal object type matched by the intelligent conversation information, making the target object anomaly type more accurate.

[0082] In one optional implementation, the case library can be used to store the correspondence between the question vector corresponding to an object problem and the object exception type vector corresponding to the object exception type. The step of querying the case library based on the intelligent session information to obtain the object exception type matching the intelligent session information includes: transforming the intelligent session information into a feature vector to obtain a session vector; and querying the case library based on the session vector to obtain the object exception type matching the intelligent session information. By setting the correspondence between the question vector corresponding to an object problem and the object exception type vector corresponding to the object exception type stored in the case library, the query efficiency for object exception types can be improved.

[0083] In step S205, based on the above-mentioned target object exception type, object exception log, object processing timing information of the above-mentioned target interaction object, and object identification information of the above-mentioned target interaction object, structured exception result description information is determined.

[0084] In the embodiments of this specification, the object exception log may refer to a log used to store exceptions that occur in interactive objects, such as a log recording exceptions that occur in orders.

[0085] In the embodiments of this specification, the object processing timing information of the target interactive object may refer to the processing timing information generated by the execution of the above-mentioned processing flow nodes corresponding to the target interactive object, and the processing information arranged in the execution order.

[0086] In the embodiments of this specification, the object identification information of the target interactive object can refer to information used to identify the target interactive object on the target platform. For example, the target interactive object is a target order, and correspondingly, the object identification information of the target interactive object can be an order number.

[0087] In the embodiments of this specification, structured abnormal result description information can refer to abnormal result description information expressed in a structured format. This disclosure does not limit the structure; the main purpose is to unify the format of abnormal result description information, making it more comprehensive and accurate. Abnormal result description information can refer to information describing the abnormality of an object.

[0088] In one possible implementation, historical interaction exception information corresponding to the object identifier information of the target interaction object can be extracted from the object exception log based on the object identifier information of the target interaction object. This allows for the combination of historical interaction exception information and the object processing sequence information of the target interaction object to mine and supplement the exception type of the target object (e.g., by calling a large language model for mining and supplementation), resulting in exception result description information. For example, structured exception result description information can be as follows: For the XX consolidated shipping order, the order was shipped using XXXXXXXX (logistics name) and the error message "Waymark number and courier company do not match" appeared. The error code is: XXXX. Here, "XX consolidated shipping order" can refer to the order category, and can also provide the logistics name, the reason for the error, and a description of the error result.

[0089] In another possible implementation, the structured exception result description information determined based on the target object exception type, object exception log, object processing timing information of the target interactive object, and object identification information of the target interactive object may include: Extract historical interaction exception information corresponding to the object identifier information of the target interaction object from the object exception log. For example, the object identifier information of the target interaction object can be used to perform a matching query in the object exception log to obtain the historical interaction exception information corresponding to the object identifier information. Furthermore, the description information of the abnormal result can be determined based on the abnormal type of the target object, historical interaction abnormal information, and the object processing sequence information of the target interaction object. By combining the abnormal type of the target object, historical interaction abnormal information, and the object processing sequence information of the target interaction object for abnormal result analysis, considering both historical interaction abnormal information and complete object processing sequence information, including the current processing status, the description information of the abnormal result can be more accurate.

[0090] In a specific embodiment, determining the abnormal result description information based on the target object's abnormality type, historical interaction abnormality information, and the object processing sequence information of the target interactive object can include: calling a large language model to perform abnormal result analysis on the target object's abnormality type, historical interaction abnormality information, and the object processing sequence information of the target interactive object, and generating the abnormal result description information. For example, the target object's abnormality type, historical interaction abnormality information, and the object processing sequence information of the target interactive object can be input into a large language model for abnormality understanding and abnormal result analysis to generate the abnormal result description information. By combining the target object's abnormality type, historical interaction abnormality information, and the object processing sequence information of the target interactive object for abnormal result analysis, both historical interaction abnormality information and complete object processing sequence information are considered. The object processing sequence information includes the current processing state, making the abnormal result description information more accurate. Furthermore, using a large language model to generate the abnormal result description information results in higher efficiency in obtaining the abnormal result description information, and the accuracy of the abnormal result description information can be further improved based on the intelligent understanding and mining capabilities of the large language model.

[0091] In one specific embodiment, the target interaction object can be a target order. Based on this, the above-mentioned determination of the abnormal result description information according to the target object's abnormality type, historical interaction abnormality information, and object processing sequence information of the target interaction object can include: When the target object's exception type is "order delivery node failure" (delivery failure), the exception result description information is determined based on historical interaction exception information. Since delivery failure scenarios typically involve successful order payment and haven't yet involved logistics tracking, they generally don't affect the object's processing sequence information. Therefore, it's preferable to determine the exception result description information based on historical interaction exception information. For example, a large language model can be invoked to perform exception analysis based on historical interaction exception information to obtain the exception result description information. Optionally, a large language model can be invoked to perform exception analysis based on both the target object's exception type and historical interaction exception information to obtain the exception result description information.

[0092] Alternatively, if the target object's exception type is an object processing node exception other than the shipping node failure, the exception result description information can be determined based on the historical interaction exception information and the object processing sequence information of the target interaction object. For example, a large language model can be invoked to perform exception analysis based on the historical interaction exception information and the object processing sequence information of the target interaction object to obtain the exception result description information. Optionally, a large language model can be invoked to perform exception analysis based on the target object's exception type, historical interaction exception information, and the object processing sequence information of the target interaction object to obtain the exception result description information.

[0093] By differentiating the processing of target orders based on different target object exception types, the goal is to obtain necessary exception information for exceptions at different processing nodes, avoid obtaining unnecessary exception information, thereby avoiding resource waste and improving the processing performance of exception result description.

[0094] In one specific embodiment, the large language model can be a pre-trained large language model (LLM), using its understanding capabilities for intent understanding (i.e., determining the anomaly type of the target object to understand the consultation intent) and anomaly result analysis and description. Alternatively, the large language model can be obtained by fine-tuning based on sample data, and this disclosure does not limit this. For example, the sample data may include sample consultation questions, with corresponding labels that can be object anomaly type labels; or the sample data may include sample object anomaly types, historical interaction anomaly information of sample interaction objects, and object processing sequence information of sample interaction objects, with corresponding labels that can be anomaly result description labels. This disclosure does not limit the specific method of fine-tuning.

[0095] In step S207, based on the above-mentioned abnormal result description information, an abnormal solution is recalled to obtain a target abnormal solution; the target abnormal solution refers to the solution overview used to solve the object association problem information.

[0096] In the embodiments of this specification, an exception knowledge base can be set up to store the exception solutions corresponding to multiple object exceptions. Based on this, the exception result description information can be matched with multiple object exceptions in the exception knowledge base to obtain the matching object exception, and the exception solution corresponding to the matching object exception can be used as the target exception solution for recall.

[0097] In a specific embodiment, the anomaly solutions stored in the anomaly knowledge base can refer to an overview of solutions for resolving corresponding object anomalies. Correspondingly, the target anomaly solution refers to an overview of solutions for resolving related object information. It is understood that the solution overview can be a summary description of the solution, such as "change to another logistics company," providing key information but omitting detailed operational steps, precautions, and other specific information. This approach saves storage resources in the anomaly knowledge base, allowing for the storage of richer anomaly solutions for object anomalies. It also provides more precise solution directions for subsequent intelligent response model input, effectively guiding and constraining the generation of anomaly resolution response information by the intelligent response model, thus improving the accuracy of the anomaly resolution response information.

[0098] In one possible implementation, the above-mentioned recall of anomaly solutions based on anomaly result description information to obtain the target anomaly solution may include: Based on the data structure of the anomaly knowledge base, the description information of anomaly results is rewritten to obtain query statements for anomaly solutions, making the query statements adaptable to the data structure of the anomaly knowledge base. For example, the data structure of the anomaly knowledge base refers to the data structure storing data in the anomaly knowledge base; this disclosure does not limit this.

[0099] Furthermore, an anomaly solution can be queried in the anomaly knowledge base based on the query statement to obtain the target anomaly solution. For example, the anomaly solution query in the anomaly knowledge base can be based on the object anomaly problem represented by the anomaly result description information carried in the query statement, thereby obtaining the target anomaly solution.

[0100] By rewriting the description information of the abnormal results through the data structure of the abnormal knowledge base, a query statement for the abnormal solution is obtained, making the query statement more accurate and improving the efficiency of obtaining the target abnormal solution.

[0101] In step S209, the intelligent response model is invoked to generate an anomaly resolution response information for the above-mentioned object-related problem information based on the above-mentioned anomaly result description information and the above-mentioned target anomaly solution; the anomaly resolution response information includes a solution overview and the specific processing information of the solution corresponding to the solution overview.

[0102] In the embodiments of this specification, the intelligent response model can be a pre-trained large language model (LLM); or it can be obtained by fine-tuning a pre-trained large language model based on a large amount of sample question-and-answer data or a large amount of sample data (e.g., including sample anomaly result description information and sample anomaly solutions, and the corresponding sample data labels can be sample response information). For example, based on the large amount of sample data and sample data labels here, a pre-trained large language model can be trained in a supervised manner to obtain an intelligent response model. This disclosure does not limit this.

[0103] In the embodiments of this specification, the specific processing information of the solution corresponding to the solution overview may refer to the specific operational information under the solution overview, such as the operation method and operation steps. In other words, the specific processing information of the solution is used to specifically complete the solution overview.

[0104] For example, the exception resolution response information can be as follows: Figure 5 As shown, the anomaly resolution response information provided by the agent may include at least one solution (e.g. Figure 5The solutions shown are an overview of "Solution 1: XXXXXXXXXXXXXXXXX" and "Solution 2: XXXXXXXXXXXXXXXXXXX" and the specific processing information for each solution (e.g., the information below Solution 1 and the information below Solution 2).

[0105] In one possible implementation, the invocation of the intelligent response model, based on the aforementioned abnormal result description information and the aforementioned target abnormal solution, to generate abnormal solution response information for the aforementioned object-related problem information, may include: The response prompt words are constructed based on the description of the abnormal result and the target solution to the abnormality. For example, the description of the abnormal result and the target solution to the abnormality can be concatenated to obtain the response prompt words, and this disclosure does not limit this approach.

[0106] Furthermore, an intelligent response model can be invoked, and response prompts can be input into the intelligent response model to generate response information, thereby obtaining abnormal resolution response information for the problem information associated with the object.

[0107] By constructing response prompts based on the description of abnormal results and the target abnormal solution, and inputting the response prompts into the intelligent response model to generate response information, the abnormal solution response information for the object-related problem information is obtained. This allows the response prompts to take into account both the description of the abnormal problem and the summary target abnormal solution for recall. It can effectively guide and constrain the generation of abnormal solution response information, thereby generating more accurate and specific abnormal handling methods.

[0108] In one possible implementation, the above method may further include: obtaining response restriction information, a preset set of question-and-answer data pairs, and a response reasoning method.

[0109] Accordingly, constructing response prompts based on the description of the abnormal results and the target solution to the abnormality may include: constructing response prompts based on the description of the abnormal results, the target solution to the abnormality, response restriction information, a preset set of question-and-answer data pairs, and the response reasoning method. For example, the description of the abnormal results, the target solution to the abnormality, the response restriction information, the preset set of question-and-answer data pairs, and the response reasoning method may be concatenated to obtain response prompts; this disclosure does not limit this approach.

[0110] In one specific implementation, the response restriction information can refer to information used to restrict the intelligent response model from generating responses. For example, it can restrict the model from generating responses based on the description information of the abnormal results and the aforementioned target abnormal solution, so as to avoid excessive divergence and resulting in false information responses.

[0111] In one specific implementation, the preset question-and-answer data pair set can refer to a pre-defined set of question-and-answer data pairs. For example, it may be pre-defined by an expert, and this disclosure does not limit this. Optionally, the formats of the questions and answers in the pre-defined question-and-answer data pairs in the preset question-and-answer data pair set can correspond to the object-related question information and the exception resolution response information. For example, the format of the questions in the pre-defined question-and-answer data pairs is consistent with the format of the object-related question information, and the format of the answers in the pre-defined question-and-answer data pairs is consistent with the format of the exception resolution response information.

[0112] In one specific implementation, the response reasoning method can be used to instruct the intelligent response model to perform reasoning response steps or processes.

[0113] By setting the dependency of the prompt words, including reply restriction information, preset question-and-answer data pairs, and reply reasoning methods, the reply prompt words are made more accurate, and the intelligent reply model is prevented from over-diverging, thus ensuring the accuracy of the reply information for resolving anomalies.

[0114] Reference Figure 4 In one possible implementation, the method may further include: In step S401, the exception resolution response information is evaluated to obtain the evaluation result.

[0115] In one specific embodiment, the evaluation result may include whether it is accurate or inaccurate, as well as the specific inaccurate content in the anomaly resolution response information. This can be viewed as an evaluation and annotation of the anomaly resolution response information. For example, AI evaluation or expert evaluation can be used, but this disclosure does not limit it.

[0116] In step S403, the anomaly resolution response information is adjusted based on the evaluation results to obtain the adjusted anomaly resolution response information.

[0117] In one specific embodiment, the anomaly resolution response information can be adjusted based on the evaluation results to obtain adjusted anomaly resolution response information. For example, if the evaluation result is accurate, the anomaly resolution response information may not be adjusted; if the evaluation result is inaccurate, the anomaly resolution response information may be adjusted to more accurate anomaly resolution response information, such as deleting inaccurate information or supplementing missing information, etc. This disclosure does not limit this.

[0118] In step S405, a new question-and-answer data pair is constructed based on the object-related question information and the adjusted exception resolution response information.

[0119] In one specific embodiment, the object-associated problem information and the adjusted exception resolution response information can be used as a new question-and-answer data pair, but this disclosure does not limit this.

[0120] In step S407, the new question-and-answer data pairs are stored in a preset question-and-answer data pair set.

[0121] In one specific embodiment, new question-and-answer data pairs can be stored in a preset question-and-answer data pair set to achieve timely updates to the preset question-and-answer data pair set.

[0122] The anomaly resolution response information is evaluated to obtain an evaluation result; the anomaly resolution response information is adjusted based on the evaluation result to obtain adjusted anomaly resolution response information; a new question-and-answer data pair is constructed based on the object-related question information and the adjusted anomaly resolution response information; the new question-and-answer data pair is stored in the preset question-and-answer data pair set, so that the question-and-answer data pair set can be updated in a timely manner, and the question-and-answer data pair set can cover more comprehensive and accurate question-and-answer data pairs, thereby improving the accuracy of response prompts and making the anomaly resolution response information more precise.

[0123] In one possible implementation, the method may further include: evaluating the exception resolution response information to obtain evaluation results; and adjusting the construction strategy of the response prompt words based on the evaluation results. For example, adjusting the construction strategy of the response prompt words based on the evaluation results can be set according to business requirements, and this disclosure does not limit this. By evaluating the exception resolution response information to obtain evaluation results, and adjusting the construction strategy of the response prompt words based on the evaluation results, the construction of the response prompt words becomes more accurate and effective, improving the accuracy and operability of the exception resolution response information.

[0124] By responding to the object-related issue information of the target account for the target interaction object in the intelligent conversation page, the system obtains the intelligent conversation information and case library in the intelligent conversation page. The case library is used to store the correspondence between object issues and object exception types. The target interaction object refers to the object generated by the target account through business interaction. Based on the intelligent conversation information and the case library, the target object exception type corresponding to the object-related issue information is determined. Based on the target object exception type, object exception logs, object processing sequence information of the target interaction object, and object identification information of the target interaction object, structured exception result description information is determined. This approach combines historical object exception logs with object processing sequence information including the current state, making the exception result description information more accurate. Furthermore, setting the exception result description information to a structured format further improves the accuracy of the exception result description information in describing the exception issue, thus providing a foundation for the accuracy of subsequent exception resolution responses.

[0125] Furthermore, based on the abnormal result description information, abnormal solution recall is performed to obtain a target abnormal solution. The target abnormal solution refers to a solution overview for resolving the object-related problem information, making the recall efficiency of abnormal solutions higher. An intelligent response model is then invoked to generate abnormal solution response information for the object-related problem information based on the abnormal result description information and the target abnormal solution. The abnormal solution response information includes the solution overview and the specific processing information corresponding to the solution overview. This results in more accurate abnormal responses and stronger operability in abnormal handling. This ensures that the abnormal solution response information not only includes a solution overview, allowing the target account to quickly locate the abnormal problem, but also includes the specific processing information corresponding to the solution overview, improving the operability of abnormal problem handling.

[0126] In an optional implementation, when the object-associated problem information indicates a response to the cause of the problem, the anomaly resolution response information may further include the anomaly result description information and anomaly cause mining information generated by the intelligent response model based on the anomaly result description information, for example... Figure 5 The corresponding content for "Cause of Problem" shown.

[0127] When the object-associated problem information indicates a response to the cause of the problem, the exception resolution response information also includes the exception result description information and the exception cause mining information generated by the intelligent response model based on the exception result description information. This allows the exception resolution response information to provide a more accurate exception cause, making the exception resolution response information more understandable and improving the user experience of exception handling interaction.

[0128] Reference Figure 6 Taking the target order as the target interaction object, if a merchant encounters a fulfillment exception, they can seek assistance. At this point, the merchant can either manually inquire or invoke a smart fulfillment exception component to display at least one related object issue, such as... Figure 6 The standard script shown.

[0129] Furthermore, intent understanding can be performed, for example through short-term conversational memory (i.e. Figure 6 The system uses short-term memory (i.e., intelligent conversation information) and a case library to identify the consultation intent (i.e., the type of abnormality of the target object) of merchants in abnormal performance scenarios, and can ensure that the corresponding performance abnormality assistant (Agent) is routed to perform abnormality diagnosis.

[0130] Next, problem diagnosis can be performed, for example, by analyzing the exception type of the target object, the object exception log (e.g., Figure 6 The exception log shown), and the object processing timing information of the target interaction object (e.g., the exception log shown). Figure 6 The processing sequence shown (such as logistics trajectory and other time-series information) and the object identification information of the target interactive object (such as order number) are used to determine structured abnormal result description information. For example, the structured abnormal result description information can be as follows: For the XX consolidated shipping order, the order was shipped using XXXXXXXX (logistics name) and the error message "Waymark number and courier company do not match" appeared. The error code is: XXXX.

[0131] This allows for solution retrieval; for example, based on the data structure of the anomaly knowledge base, the description information of the anomaly result can be rewritten to obtain a query statement for the anomaly solution (e.g., Figure 6 (The query rewrite shown); thus, an exception solution can be retrieved from the exception knowledge base based on the query statement to obtain the target exception solution. For example, the retrieved target exception solution can be as follows: 1. This is usually caused by the merchant choosing the wrong courier. We suggest: 1. Changing the courier; 2. Getting a new courier number. 2. The XX consolidation order platform requires that only electronic waybills be used and shipments be made through mainstream courier services.

[0132] Furthermore, responses can be generated based on AI, for example, based on descriptions of abnormal results (e.g.) Figure 6 The problem diagnosis information shown), and the target anomaly solution (e.g. Figure 6 The solution knowledge shown), and the response limitation information (e.g. Figure 6 The response restrictions shown), and the preset question-and-answer data pair set (e.g. Figure 6 The FewShot shown) and the response reasoning method (e.g. Figure 6 As shown in the diagram (Cot), a response suggestion word is constructed. The intelligent response model is then invoked, and the response suggestion word is input into the model to generate response information, resulting in an exception resolution response information for the object-related problem information. For example, the object-related problem information and the exception resolution response information can be as follows: Figure 5 As shown, this disclosure does not limit this.

[0133] Optionally, the prompts can also include the agent's role and skills. For example, role: You are an intelligent customer service assistant for merchants, representing the XX brand image, and solving merchants' daily operational problems with a professional, friendly, and accurate attitude. Skills: Generate corresponding standard procedures or operational suggestions based on the prompts.

[0134] As an example, response reasoning methods may include: 1. Build context-rich dynamic prompts that embed solution knowledge and problem diagnosis information.

[0135] 2. To generate an answer, you can do so in the following way: Provide conclusions and recommended solutions directly.

[0136] For operation instructions, please follow the step numbers.

[0137] Personalized descriptions based on associated data (such as order number, tracking number, time, etc.).

[0138] 3. If the information is insufficient (e.g., inquiring about an order but not providing the order number, inquiring about a package but not providing the tracking number), the following actions can be taken: clearly inform the user of the missing information and guide them to supplement the necessary information (e.g., order number, tracking number, etc.).

[0139] 4. Always maintain a professional, polite, and clear tone.

[0140] 5. If multiple rounds of dialogue have been conducted but the merchant's issue remains unresolved, the merchant can be guided to transfer to a human operator for processing.

[0141] As an example, response restrictions could include the following: 1. Answers must be based strictly on solution knowledge and problem diagnosis information; fabricated information is prohibited.

[0142] 2. For sensitive content, the company's standard terminology must be used, and the scope must not be exceeded.

[0143] 3. Style: Answers must be accurate, concise, and easy to understand. Professional and definitive responses are essential.

[0144] 4. Adhere to the brand's tone and avoid negative or offensive language.

[0145] 5. Try to avoid using emojis in your replies, and do not include special characters in your replies.

[0146] 6. Please check the replies in the history of the conversation before replying. Make full use of the user's context, naturally connect the preceding and following sentences, and negotiate with the user. Avoid mechanically repeating the same words and avoid repetitive or similar replies.

[0147] 7. Questions that are prohibited from being answered, such as sensitive company data, internal product logic, and personal privacy information.

[0148] 9. Prohibited words and phrases, such as "Do not answer with codes".

[0149] Reference Figure 6 It can also evaluate responses to anomalies, such as conducting AI evaluations (sentiment analysis, accuracy assessment) and manual annotations based on questions and answers from real online merchants, and continuously optimizing prompts based on the annotation results to ensure response accuracy.

[0150] Optionally, before going live, a test set evaluation can be conducted, and AI evaluation can be performed based on the test set. The results of the AI ​​evaluation can then be manually annotated to ensure the accuracy of the responses.

[0151] By combining historical object exception logs with object processing sequence information including the current state, the description of exception results becomes more accurate. Furthermore, setting the exception result description information to a structured format further enhances the accuracy of the description of the exception problem, thus providing a foundation for accurate subsequent exception resolution responses. Based on the exception result description information, exception solution retrieval is performed to obtain a target exception solution; the target exception solution refers to a solution overview for resolving the object-related problem information, making exception solution retrieval more efficient. Finally, an intelligent response model is invoked to generate exception resolution response information for the object-related problem information based on the exception result description information and the target exception solution; the exception resolution response information includes the solution overview and the specific processing information corresponding to the solution overview. This makes exception responses more accurate and exception handling more operable.

[0152] Figure 7 This is a block diagram of an information processing apparatus according to an exemplary embodiment. (Refer to...) Figure 7 The information processing device may include: The information acquisition module 701 is configured to respond to the target account's object association question information for the target interactive object in the smart conversation page, acquire the smart conversation information and case library in the smart conversation page, the case library being used to store the correspondence between object questions and object exception types; the target interactive object refers to the object generated by the target account through business interaction. The object anomaly type determination module 703 is configured to determine the target object anomaly type corresponding to the object-related problem information based on the intelligent session information and the case library; The abnormal result description module 705 is configured to determine structured abnormal result description information based on the target object's abnormal type, object abnormal log, object processing timing information of the target interactive object, and object identification information of the target interactive object. An anomaly solution recall module 707 is configured to perform anomaly solution recall based on the anomaly result description information to obtain a target anomaly solution; the target anomaly solution refers to a solution overview for resolving the object-related problem information. The exception resolution response information generation module 709 is configured to execute and call the intelligent response model to generate exception resolution response information for the object-related problem information based on the exception result description information and the target exception solution; the exception resolution response information includes an overview of the solution and specific processing information of the solution corresponding to the solution overview.

[0153] By responding to the object-related issue information of the target account for the target interaction object in the intelligent conversation page, the system obtains the intelligent conversation information and case library in the intelligent conversation page. The case library is used to store the correspondence between object issues and object exception types. The target interaction object refers to the object generated by the target account through business interaction. Based on the intelligent conversation information and the case library, the target object exception type corresponding to the object-related issue information is determined. Based on the target object exception type, object exception logs, object processing sequence information of the target interaction object, and object identification information of the target interaction object, structured exception result description information is determined. This approach combines historical object exception logs with object processing sequence information including the current state, making the exception result description information more accurate. Furthermore, setting the exception result description information to a structured format further improves the accuracy of the exception result description information in describing the exception issue, thus providing a foundation for the accuracy of subsequent exception resolution responses.

[0154] Furthermore, based on the abnormal result description information, abnormal solution recall is performed to obtain a target abnormal solution. The target abnormal solution refers to a solution overview for resolving the object-related problem information, making the recall efficiency of abnormal solutions higher. An intelligent response model is then invoked to generate abnormal solution response information for the object-related problem information based on the abnormal result description information and the target abnormal solution. The abnormal solution response information includes the solution overview and the specific processing information corresponding to the solution overview. This results in more accurate abnormal responses and stronger operability in abnormal handling. This ensures that the abnormal solution response information not only includes a solution overview, allowing the target account to quickly locate the abnormal problem, but also includes the specific processing information corresponding to the solution overview, improving the operability of abnormal problem handling.

[0155] In one possible implementation, the object exception type determination module 703 may include: The object exception type determination unit is configured to perform a query in the case library based on the smart session information to obtain the object exception type that matches the smart session information. The target object anomaly type acquisition unit is configured to perform fine-grained anomaly analysis based on a large language model on the intelligent session information and the object anomaly type matched by the intelligent session information to obtain the target object anomaly type; the target object anomaly type is used to indicate the object processing flow node of the anomaly under the object anomaly type matched by the intelligent session information.

[0156] In one possible implementation, the case library is used to store the correspondence between the problem vector corresponding to the object problem and the object exception type vector corresponding to the object exception type; the object exception type determination unit may include: The session vector acquisition subunit is configured to perform feature vector transformation on the intelligent session information to obtain a session vector. The object exception type determination subunit is configured to perform a query in the case library based on the session vector to obtain the object exception type that matches the intelligent session information.

[0157] In one possible implementation, the abnormal result description module 705 may include: The historical interaction anomaly information extraction unit is configured to extract historical interaction anomaly information corresponding to the object identifier information of the target interaction object from the object anomaly log. The abnormal result description unit is configured to determine the abnormal result description information based on the abnormal type of the target object, the historical interaction abnormal information, and the object processing timing information of the target interaction object.

[0158] In one possible implementation, the target interaction object is a target order; the abnormal result description unit may include: The first abnormal result description subunit is configured to determine the abnormal result description information based on the historical interaction abnormal information when the abnormal type of the target object is order delivery node failure. Alternatively, the second abnormal result description subunit is configured to determine the abnormal result description information based on the historical interaction abnormal information and the object processing sequence information of the target interaction object when the abnormality type of the target object is an object processing flow node abnormality other than the shipping node failure.

[0159] In one possible implementation, the abnormal result description unit may include: The third abnormal result description subunit is configured to execute the large language model to perform abnormal result analysis on the abnormal type of the target object, the historical interaction abnormal information, and the object processing timing information of the target interaction object, and generate the abnormal result description information.

[0160] In one possible implementation, the exception resolution response information generation module 709 may include: The response prompt word construction unit is configured to construct response prompt words based on the abnormal result description information and the target abnormal solution; The exception resolution response information generation unit is configured to execute the call to the intelligent response model, input the response prompt words into the intelligent response model to generate response information, and obtain the exception resolution response information for the object-related problem information.

[0161] In one possible implementation, the device may further include: The prompt word association information acquisition module is configured to acquire reply restriction information, a preset set of question-and-answer data pairs, and reply reasoning methods; The reply prompt word construction unit is further configured to construct the reply prompt words based on the abnormal result description information, the target abnormal solution, the reply restriction information, the preset question-and-answer data pair set, and the reply reasoning method.

[0162] In one possible implementation, the device may further include: The response information evaluation module is configured to evaluate the exception resolution response information and obtain the evaluation results. The exception resolution response information adjustment module is configured to adjust the exception resolution response information based on the evaluation results to obtain the adjusted exception resolution response information; The question-and-answer data pair construction module is configured to construct new question-and-answer data pairs based on the object-associated question information and the adjusted exception resolution response information; The question-and-answer data pair set update module is configured to store the new question-and-answer data pair into the preset question-and-answer data pair set.

[0163] In one possible implementation, the device may further include: The response information evaluation module is configured to evaluate the exception resolution response information and obtain the evaluation results. The prompt word construction strategy adjustment module is configured to execute a strategy for adjusting the construction of the response prompt words based on the evaluation results.

[0164] In one possible implementation, the exception solution recall module 707 may include: The query statement acquisition unit is configured to execute a data structure based on the exception knowledge base, rewrite the exception result description information, and obtain a query statement for the exception solution; The exception solution query unit is configured to perform an exception solution query in the exception knowledge base according to the query statement to obtain the target exception solution.

[0165] In one possible implementation, the device may further include: The intelligent conversation guidance module is configured to perform intelligent conversation guidance when there is an anomaly in the object processing timing information of the target interactive object in the object details page of the target interactive object. The object association question push module is configured to respond to the target account entering the smart conversation page based on the smart conversation guidance, display the smart conversation page where the target account and the smart agent are having a smart conversation to the target account, and push at least one object association question in a preset question format to the smart conversation page based on the object identification information; The object association question information processing module is configured to, when the target account selects a target object association question from the at least one object association question, call the intelligent agent to generate the object association question information carrying the object identification information based on the preset question format, and display the object association question information of the target account for the target interactive object on the intelligent conversation page.

[0166] In one possible implementation, when the object-associated problem information indicates a response to the cause of the problem, the anomaly resolution response information also includes the anomaly result description information and the anomaly cause mining information generated by the intelligent response model based on the anomaly result description information.

[0167] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0168] Figure 8 This is a block diagram illustrating an electronic device for information processing according to an exemplary embodiment. The electronic device may be a terminal, and its internal structure diagram may be as follows: Figure 8As shown, the electronic device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage medium. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements an information processing method. The display screen can be a liquid crystal display (LCD) or an e-ink display. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the device's casing, or an external keyboard, touchpad, or mouse.

[0169] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present disclosure and does not constitute a limitation on the electronic device to which the present disclosure is applied. A specific electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0170] Figure 9 This is a block diagram of an electronic device for information processing based on an exemplary embodiment. The electronic device may be a server, and its internal structure diagram may be as follows: Figure 9 As shown, this electronic device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements an information processing method.

[0171] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present disclosure and does not constitute a limitation on the electronic device to which the present disclosure is applied. A specific electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0172] In an exemplary embodiment, an electronic device is also provided, including: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the information processing method as described in the embodiments of this disclosure.

[0173] In an exemplary embodiment, a computer-readable storage medium is also provided, which, when executed by a processor of an electronic device, enables the electronic device to perform the information processing method of the present disclosure. The computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, or optical data storage device, etc.

[0174] In an exemplary embodiment, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to perform the information processing method of the embodiments of this disclosure.

[0175] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0176] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0177] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. An information processing method, characterized in that, include: In response to the object association question information of the target account for the target interactive object in the smart conversation page, obtain the smart conversation information and case library in the smart conversation page. The case library is used to store the correspondence between object questions and object exception types. The target interaction object refers to the object generated by the target account during business interactions. Based on the intelligent session information and the case library, determine the target object anomaly type corresponding to the object-related problem information; Based on the target object's exception type, object exception log, object processing timing information of the target interactive object, and object identification information of the target interactive object, structured exception result description information is determined. Based on the description information of the abnormal results, an abnormal solution is recalled to obtain the target abnormal solution; The target anomaly solution refers to an overview of the solution used to resolve the object association problem information; The intelligent response model is invoked to generate an anomaly resolution response information for the object-related problem information based on the anomaly result description information and the target anomaly solution; the anomaly resolution response information includes an overview of the solution and specific processing information of the solution corresponding to the solution overview.

2. The method according to claim 1, characterized in that, The step of determining the target object anomaly type corresponding to the object-associated problem information based on the intelligent session information and the case library includes: Based on the intelligent session information, the case library is queried to obtain the object anomaly type that matches the intelligent session information; Based on a large language model, fine-grained anomaly analysis is performed on the intelligent session information and the object anomaly type matched by the intelligent session information to obtain the target object anomaly type; the target object anomaly type is used to indicate the object processing flow node of the anomaly under the object anomaly type matched by the intelligent session information.

3. The method according to claim 2, characterized in that, The case library is used to store the correspondence between the problem vector corresponding to the object problem and the object exception type vector corresponding to the object exception type; the step of querying the case library based on the intelligent session information to obtain the object exception type matched by the intelligent session information includes: The intelligent conversation information is transformed into a feature vector to obtain a conversation vector; Based on the session vector, the case library is queried to obtain the object anomaly type that matches the intelligent session information.

4. The method according to claim 1, characterized in that, The step of determining structured exception result description information based on the target object's exception type, object exception log, object processing timing information of the target interactive object, and object identification information of the target interactive object includes: Extract historical interaction exception information corresponding to the object identifier information of the target interactive object from the object exception log; The abnormal result description information is determined based on the target object's abnormality type, the historical interaction abnormality information, and the object processing timing information of the target interaction object.

5. The method according to claim 4, characterized in that, The target interaction object is the target order; the step of determining the abnormal result description information based on the abnormal type of the target object, the historical interaction abnormal information, and the object processing sequence information of the target interaction object includes: In the case where the exception type of the target object is order delivery node failure, the exception result description information is determined based on the historical interaction exception information; Alternatively, if the exception type of the target object is an object processing flow node exception other than the shipping node failure, the exception result description information is determined based on the historical interaction exception information and the object processing sequence information of the target interaction object.

6. The method according to claim 4, characterized in that, The step of determining the anomaly result description information based on the target object anomaly type, the historical interaction anomaly information, and the object processing timing information of the target interaction object includes: The large language model is invoked to perform anomaly result analysis on the anomaly type of the target object, the historical interaction anomaly information, and the object processing timing information of the target interaction object, and generate the anomaly result description information.

7. The method according to claim 1, characterized in that, The intelligent response model is invoked to generate exception resolution response information for the object-related problem information based on the exception result description information and the target exception solution, including: Based on the description of the abnormal result and the target abnormal solution, construct response prompt words; The intelligent reply model is invoked, and the reply prompt words are input into the intelligent reply model to generate reply information, thereby obtaining the abnormal resolution reply information for the object-related problem information.

8. The method according to claim 7, characterized in that, The method further includes: Obtain reply restriction information, preset question-and-answer data pairs, and reply reasoning methods; The step of constructing response prompt words based on the abnormal result description information and the target abnormal solution includes: The reply prompt words are constructed based on the description of the abnormal result, the target abnormal solution, the reply restriction information, the preset question-and-answer data pair set, and the reply reasoning method.

9. The method according to claim 8, characterized in that, The method further includes: The error resolution response information is evaluated to obtain the evaluation results; The anomaly resolution response information is adjusted based on the evaluation results to obtain the adjusted anomaly resolution response information; Based on the object-related question information and the adjusted exception resolution response information, a new question-and-answer data pair is constructed; The new question-and-answer data pairs are stored in the preset question-and-answer data pair set.

10. The method according to claim 7, characterized in that, The method further includes: The error resolution response information is evaluated to obtain the evaluation results; The strategy for constructing the response prompts will be adjusted based on the evaluation results.

11. The method according to claim 1, characterized in that, The step of recalling anomaly solutions based on the anomaly result description information to obtain the target anomaly solution includes: Based on the data structure of the anomaly knowledge base, the description information of the anomaly results is rewritten to obtain the query statement for the anomaly solution; The query statement is used to search for an anomaly solution in the anomaly knowledge base to obtain the target anomaly solution.

12. The method according to claim 1, characterized in that, The method further includes: If there is an anomaly in the object processing timing information of the target interactive object on the object details page, intelligent session guidance will be performed. In response to the target account entering the smart conversation page based on the smart conversation guidance, the smart conversation page where the target account and the smart agent conduct a smart conversation is displayed to the target account, and at least one object-related question in a preset question format is pushed to the smart conversation page based on the object identification information; When the target account selects a target object association question from the at least one object association question, the agent is invoked to generate the object association question information carrying the object identification information based on the preset question format, and the target account displays the object association question information for the target interactive object on the smart conversation page.

13. The method according to claim 1, characterized in that, When the object-related problem information indicates that a response should be given to the cause of the problem, the anomaly resolution response information also includes the anomaly result description information and the anomaly cause mining information generated by the intelligent response model based on the anomaly result description information.

14. An information processing device, characterized in that, include: The information acquisition module is configured to respond to the target account's object association question information for the target interactive object in the smart conversation page, and to acquire the smart conversation information and case library in the smart conversation page. The case library is used to store the correspondence between object questions and object exception types. The target interaction object refers to the object generated by the target account during business interactions. The object anomaly type determination module is configured to determine the target object anomaly type corresponding to the object-related problem information based on the intelligent session information and the case library. The abnormal result description module is configured to determine structured abnormal result description information based on the target object's abnormal type, object abnormal log, object processing timing information of the target interactive object, and object identification information of the target interactive object. The abnormal solution recall module is configured to recall abnormal solutions based on the abnormal result description information to obtain the target abnormal solution; The target anomaly solution refers to an overview of the solution used to resolve the object association problem information; The exception resolution response information generation module is configured to execute and invoke the intelligent response model to generate exception resolution response information for the object-related problem information based on the exception result description information and the target exception solution; the exception resolution response information includes an overview of the solution and specific processing information of the solution corresponding to the solution overview.

15. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the information processing method as described in any one of claims 1 to 13.

16. A computer-readable storage medium, characterized in that, When the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device is able to perform the information processing method as described in any one of claims 1 to 13.

17. A computer program product, characterized in that, It includes computer instructions, which, when executed by a processor, cause the computer to perform the information processing method as described in any one of claims 1 to 13.

Images