Instant messaging-based conversation message processing method and device, electronic equipment and program product

Abstract

The present disclosure discloses a method and device for processing conversation messages based on instant messaging, an electronic device and a program product. The method comprises: obtaining a conversation message in an instant messaging conversation; performing intent recognition on the conversation message, identifying a target conversation message with a target intent of requesting a virtual resource configuration suggestion, and determining a target intent of a conversation party associated with the target conversation message; generating a task tree according to the target intent, the task tree defining a plurality of processing tasks and an arrangement logic of the plurality of processing tasks; according to the task tree, scheduling a target intelligent agent corresponding to the processing task to perform the plurality of processing tasks according to the arrangement logic, and generating a virtual resource configuration suggestion; and returning the virtual resource configuration suggestion to the instant messaging conversation. The method according to the embodiment of the present disclosure can significantly improve the flexibility and processing efficiency of conversation message processing.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, specifically to a method, apparatus, electronic device, and program product for processing conversational messages based on instant messaging. Background Technology

[0002] Currently, with the popularization of instant messaging technology, more and more enterprises and organizations are interacting with users through instant messaging platforms, and users can express their service needs for virtual resources through instant messaging sessions.

[0003] To address this, solutions have been proposed for processing text messages in conversations and providing responses to users based on certain rules or models. However, as conversation formats diversify and user needs become increasingly complex, known conversation message processing solutions have shown significant shortcomings in response efficiency and handling complex tasks.

[0004] Therefore, we hope to provide a technical solution that can process instant messaging messages more flexibly and efficiently to meet users' complex and diverse service needs.

[0005] The background description is provided for the purpose of understanding the relevant technologies in this field and is not intended as an admission of prior art. Summary of the Invention

[0006] The embodiments disclosed herein aim to provide a method, apparatus, electronic device, and program product for processing conversational messages based on instant messaging, which can significantly improve the flexibility and efficiency of conversational message processing.

[0007] In a first aspect, embodiments of this disclosure provide a session message processing method based on instant messaging, comprising:

[0008] Retrieve session messages from an instant messaging session;

[0009] The session messages are subjected to intent recognition to identify target session messages with the target intent of requesting virtual resource configuration suggestions, and to determine the target intent of the session party associated with the target session message.

[0010] A task tree is generated based on the target intent, and the task tree defines one or more processing tasks and the orchestration logic of the one or more processing tasks;

[0011] Based on the orchestration logic defined in the task tree, the target intelligent agent corresponding to the one or more processing tasks is scheduled to execute the one or more processing tasks, and virtual resource configuration suggestions are generated.

[0012] The virtual resource configuration suggestion is returned to the instant messaging session.

[0013] In some embodiments, the processing tasks are multiple and include resource processing tasks and content generation tasks;

[0014] The orchestration logic defines at least one of the output data of the resource processing task as the input data of the content generation task, wherein the output data includes state analysis results generated based on data related to virtual resources.

[0015] In some embodiments, the resource processing task includes a resource status analysis task, and the orchestration logic defines the output data of the resource status analysis task as the input data of the content generation task;

[0016] The step of scheduling target agents corresponding to the one or more processing tasks to execute the one or more processing tasks according to the orchestration logic defined in the task tree, and generating virtual resource configuration suggestions, includes:

[0017] The scheduling data analysis agent executes the resource status analysis task to obtain virtual resource data from at least one virtual resource data source and generate status analysis results based on the virtual resource data;

[0018] The content generation agent is scheduled to execute the content generation task to perform semantic rewriting on the state analysis results and generate virtual resource configuration suggestions.

[0019] In some embodiments, the resource processing task includes a resource status analysis task and a resource scheme recommendation task, and the orchestration logic defines the output data of the resource status analysis task and / or the output data of the resource scheme recommendation task as the input data of the content generation task;

[0020] The step of scheduling target agents corresponding to the one or more processing tasks to execute the one or more processing tasks according to the orchestration logic defined in the task tree, and generating virtual resource configuration suggestions, includes:

[0021] The scheduling data analysis agent executes the resource status analysis task to obtain virtual resource data from at least one virtual resource data source and generate status analysis results based on the virtual resource data;

[0022] The scheduling scheme recommendation agent executes the resource scheme recommendation task to obtain the identity tag of the session party, and selects at least one recommended scheme from multiple candidate virtual resource configuration schemes based on the identity tag to obtain the scheme recommendation result;

[0023] The content generation agent is scheduled to execute the content generation task to perform semantic rewriting on the state analysis results and / or the recommendation results, and to perform semantic merging on the state analysis results and the recommendation results to generate the virtual resource configuration suggestions.

[0024] In some embodiments, the resource processing task includes a resource holding analysis task and a resource scheme recommendation task, and the orchestration logic defines the output data of the resource holding analysis task as the input data of the content generation task;

[0025] The step of scheduling target agents corresponding to the one or more processing tasks to execute the one or more processing tasks according to the orchestration logic defined in the task tree, and generating virtual resource configuration suggestions, includes:

[0026] Based on the resource holding data of the session party, the resource preference tag of the session party is determined, wherein the resource holding data represents at least a portion of the virtual resources held by the session party;

[0027] The resource holding analysis agent is scheduled to execute the resource holding analysis task to generate a status analysis result of the virtual resources held by the session party based on the held resource data and the resource preference tags;

[0028] The content generation agent is scheduled to execute the content generation task to perform semantic rewriting on the state analysis results and generate the virtual resource configuration suggestions.

[0029] In some embodiments, the resource processing task includes a resource holding analysis task and a resource scheme recommendation task, and the orchestration logic defines the output data of the resource holding analysis task and / or the resource scheme recommendation task as the input data of the content generation task;

[0030] The step of scheduling target agents corresponding to the one or more processing tasks to execute the one or more processing tasks according to the orchestration logic defined in the task tree, and generating virtual resource configuration suggestions, includes:

[0031] Based on the resource holding data of the session party, the resource preference tag of the session party is determined, wherein the resource holding data represents at least a portion of the virtual resources held by the session party;

[0032] The scheduling resource holding analysis agent executes the resource holding analysis task, and generates a status analysis result of the virtual resources held by the session party based on the held resource data and the resource preference tags. The held resource analysis result includes adjustment suggestions for the held resources.

[0033] The scheduling scheme recommendation agent executes the resource scheme recommendation task and generates scheme recommendation results based on the resource holding analysis results and the resource preference tags;

[0034] The content generation agent is scheduled to execute the content generation task, perform semantic rewriting on the state analysis results and / or the scheme recommendation results, and perform semantic merging on the state analysis results and the scheme recommendation results to generate the virtual resource configuration suggestion.

[0035] In some embodiments, generating the state analysis result of the virtual resources held by the session party based on the held resource data and the resource preference tags includes:

[0036] Obtain virtual resource data corresponding to the virtual resources held by the session party from at least one virtual resource data source;

[0037] The resource status of the virtual resources held by the session party is determined based on the corresponding virtual resource data;

[0038] Based on the resource status and the resource preference tags, the resource risk characteristics of the session party are determined;

[0039] Based on the resource risk characteristics, a status analysis result of the held resources is generated.

[0040] In some embodiments, generating a scheme recommendation result based on the resource holding analysis result and the resource preference label includes:

[0041] From multiple candidate resource configuration schemes, at least one recommended scheme is selected that matches the risk characteristics with the resource preference label to obtain the scheme recommendation result, wherein the virtual resources involved in the recommended scheme are at least partially different from the virtual resources held by the session party.

[0042] In some embodiments, generating resource holding analysis results based on the held resource data and the resource preference tags includes:

[0043] Determine the resource status of at least a portion of the virtual resources held by the session party;

[0044] Based on the resource status and the resource preference tags, the resource risk characteristics of the session party are determined;

[0045] Based on the resource risk characteristics, determine the analysis results of the held resources, including the adjustment recommendations;

[0046] The step of generating a scheme recommendation result based on the resource holding analysis result and the resource preference label includes:

[0047] From multiple candidate resource configuration schemes, at least one recommended scheme is selected that matches the risk characteristics with the resource preference label and meets the adjustment suggestions, to obtain the scheme recommendation result, wherein the virtual resources involved in the recommended scheme are at least partially different from the virtual resources held by the session party.

[0048] In some embodiments, the instant messaging session is a group session;

[0049] The step of obtaining session messages in an instant messaging session includes: obtaining multiple session messages from the group session, wherein at least some of the multiple sessions come from different parties.

[0050] The step of performing intent recognition on the session messages, identifying target session messages with the target intent of requesting virtual resource configuration suggestions, and determining the target intent of the session party associated with the target session message, includes:

[0051] Identify multiple target session messages with the target intent of requesting virtual resource configuration suggestions;

[0052] The multiple target session messages are aggregated to obtain at least one intent category, and each intent category corresponds to a target intent;

[0053] The step of returning the virtual resource configuration suggestion to the instant messaging session includes: returning the virtual resource configuration suggestion generated for the at least one intent category to the group session, and optionally, providing targeted reminders to the session participants associated with the target session message.

[0054] In a second aspect, embodiments of this disclosure provide a session message processing apparatus based on instant messaging, comprising:

[0055] The session acquisition unit is configured to acquire session messages in an instant messaging session.

[0056] The intent recognition unit is configured to perform intent recognition on the session message, identify the target session message with the target intent of requesting virtual resource configuration suggestions, and determine the target intent of the session party associated with the target session message;

[0057] The task tree unit is configured to generate a task tree based on the target intent, wherein the task tree defines multiple processing tasks and the orchestration logic of the multiple processing tasks;

[0058] The scheduling unit is configured to schedule the target agent corresponding to the one or more processing tasks to execute the one or more processing tasks according to the orchestration logic defined in the task tree, and generate virtual resource configuration suggestions.

[0059] The return unit is configured to return the virtual resource configuration suggestion to the instant messaging session.

[0060] In a third aspect, an electronic device is provided, comprising: a processor and a memory storing a computer program, the processor being configured to implement the method as described in the first aspect when the computer program is executed.

[0061] In a fourth aspect, a program product is provided, comprising a computer program, wherein the computer program, when executed by a processor, implements the method as described in the first aspect.

[0062] The instant messaging-based session message processing method provided in this disclosure can acquire session messages in an instant messaging session; perform intent recognition on the session messages to identify target session messages with a target intent requesting virtual resource configuration suggestions, and determine the target intent of the session party associated with the target session message; generate a task tree based on the target intent, the task tree defining one or more processing tasks and the orchestration logic of the one or more processing tasks; schedule target agents corresponding to the one or more processing tasks to execute the one or more processing tasks according to the orchestration logic defined in the task tree, and generate virtual resource configuration suggestions; and return the virtual resource configuration suggestions to the instant messaging session. This scheme, by employing a task tree orchestration mechanism, can dynamically plan processing tasks and their execution order according to different target intents, enabling multiple agents to work collaboratively according to predetermined orchestration logic, thereby significantly improving the flexibility and efficiency of session message processing.

[0063] In a further embodiment of this disclosure, the resource processing task may include a resource status analysis task, and the orchestration logic defines that the output data of the resource status analysis task can be used as the input data of the content generation task. This further solution enables effective linkage between virtual resource status data and configuration suggestion generation. By scheduling a data analysis agent to obtain virtual resource data from external data sources and generate status analysis results, and then having the content generation agent perform semantic rewriting on the analysis results, it is possible to generate easy-to-understand configuration suggestions based on real-time resource status data, thereby improving the timeliness and understandability of the configuration suggestions.

[0064] In a further embodiment of this disclosure, the resource processing task may further include a resource holding analysis task, which can determine resource preference tags based on the resource holding data of the session party, and generate status analysis results based on the resource holding data and resource preference tags. This further solution can solve the technical problem of providing targeted services based on the personalized characteristics of the session party. By determining resource preference tags based on the resource holding data and combining the tags with the analysis task, configuration suggestions that match the resource preferences of the session party can be generated to achieve personalized services, thereby significantly improving the accuracy of configuration suggestions.

[0065] Other optional features and technical effects of the embodiments of this disclosure are described in part below, and in part will be apparent from reading this document. Attached Figure Description

[0066] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings. The elements shown are not limited to the scale shown in the drawings, and the same or similar reference numerals in the drawings denote the same or similar elements, wherein:

[0067] Figure 1 An exemplary flowchart of a session message processing method based on instant messaging according to an embodiment of the present disclosure is shown;

[0068] Figure 2 An exemplary flowchart of a session message processing method based on instant messaging according to an embodiment of the present disclosure is shown;

[0069] Figure 3 An exemplary flowchart of a session message processing method based on instant messaging according to an embodiment of the present disclosure is shown;

[0070] Figure 4 An exemplary flowchart of a session message processing method based on instant messaging according to an embodiment of the present disclosure is shown;

[0071] Figure 5 An exemplary flowchart of a session message processing method based on instant messaging according to an embodiment of the present disclosure is shown;

[0072] Figure 6 An exemplary flowchart of a session message processing method based on instant messaging according to an embodiment of the present disclosure is shown;

[0073] Figure 7 An exemplary flowchart of a session message processing method based on instant messaging according to an embodiment of the present disclosure is shown;

[0074] Figure 8 An exemplary flowchart of a session message processing method based on instant messaging according to an embodiment of the present disclosure is shown;

[0075] Figure 9An exemplary block diagram of an instant messaging-based session message processing apparatus according to embodiments of the present disclosure is shown; and

[0076] Figure 10 An exemplary structural diagram of an electronic device that can implement an instant messaging-based session message processing method according to embodiments of the present disclosure is shown. Detailed Implementation

[0077] To make the objectives, technical solutions, and advantages of this disclosure clearer, the disclosure will be further described in detail below with reference to specific embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this disclosure are used to explain this disclosure, but are not intended to limit this disclosure.

[0078] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0079] The user data, data acquisition, and / or use involved in the embodiments of this disclosure strictly comply with the laws, regulations, and industry standards of relevant countries and regions. The collection and acquisition of data involved in the embodiments of this disclosure are all done in advance by actively prompting or prominently displaying information to inform users and obtaining authorization, or by obtaining full authorization from all parties. The processing, manipulation, forwarding, and use of data involved in the embodiments of this disclosure are all carried out on the premise that the user or relevant party is fully informed and authorized. When implementing the embodiments of this disclosure, the types of data or information, scope of use, and usage scenarios that may be involved are informed to users or relevant parties and authorization is obtained through appropriate means. The specific methods of notification and authorization may vary according to actual circumstances, and this disclosure is not limited in this regard. The processing of personal information involved in the embodiments of this disclosure is carried out under the premise of having a legal basis (such as obtaining the consent of the personal information subject or being necessary for the performance of a contract), and is only processed within the prescribed or agreed scope. Sensitive personal information such as biometric information, medical and health information, financial account information, and precise location information involved in the embodiments of this disclosure are all processed under the premise of having a specific purpose and sufficient necessity, and with the separate authorization and consent of the user or relevant party. In some embodiments of this disclosure, if a user or related party refuses to process personal information other than the information necessary for the basic functions, it will not affect the use of the basic functions of the embodiments of this disclosure.

[0080] As mentioned earlier, in instant messaging applications, users may express diverse service needs through conversation messages. These needs may involve querying, analyzing, or providing configuration suggestions for virtual resources. Currently, with the increasing complexity of user needs, higher demands are being placed on conversation message processing capabilities.

[0081] In this regard, this disclosure recognizes that some known session message processing solutions have certain limitations when facing complex requirements, specifically in the following aspects:

[0082] First, some known solutions employ relatively simple response mechanisms at the task scheduling level. Specifically, some known solutions typically map intent recognition results directly to the corresponding processing modules, thus simply establishing an "intent-to-process" mapping. However, in real-world scenarios, a user's request may require the collaborative completion of multiple processing stages. For example, when a user wants virtual resource configuration suggestions, it may be necessary to first obtain and analyze resource status data from an external data source, then recommend solutions based on the analysis results and user characteristics, and finally translate the professional analysis conclusions into an easily understandable expression. This disclosure recognizes that known solutions neglect the data dependencies between these processing stages. When the input of subsequent stages depends on the output of preceding stages, the simple "intent-to-process" mapping mechanism is insufficient to effectively define and coordinate the execution order and data flow relationships of these multiple stages, potentially leading to broken processing flows or incomplete results.

[0083] Secondly, some known solutions fail to fully leverage the personalized characteristics of session participants to achieve personalized services. Specifically, some known solutions typically generate configuration recommendations based on general rules or models, without considering the virtual resources held by the session participants or their preferences. However, this disclosure recognizes that different session participants have significant differences in resource holdings and risk preferences. The configuration recommendations generated by known solutions fail to match the personalized characteristics of the session participants, resulting in insufficient targeting and practicality of the recommendations, making it difficult to meet the current users' differentiated service needs.

[0084] Furthermore, some known solutions handle the needs of multiple participants in a group session independently. However, this disclosure recognizes that multiple participants in a group session may express similar or related needs within a short period of time. In this case, if each need is identified and processed independently, it may result in redundant processing overhead, and it will be difficult to provide a unified and efficient response to the relevant participants.

[0085] To address this, this disclosure proposes a method for processing instant messaging-based conversational messages, along with corresponding devices, electronic equipment, and program products. The instant messaging-based conversational message processing method of this disclosure can be executed by a conversational message processing system / device, which includes, but is not limited to, the backend service of an instant messaging platform, conversational robots, third-party application plugins, enterprise financial service platforms, or enterprise middleware. The instant messaging-based conversational message processing method and device of this disclosure can be used in application scenarios that require understanding user requests, task orchestration, collaborative processing, and returning results within an instant messaging conversation, including but not limited to: enterprise customer operations and services, and intelligent investment advisory / financial consulting. The method and device of this disclosure will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0086] In some embodiments, reference Figure 1 The instant messaging-based session message processing method according to the embodiments of this disclosure may include the following steps 110, 120, 130, 140 and 150.

[0087] 110: Retrieve session messages from an instant messaging session.

[0088] In this embodiment of the disclosure, the instant messaging session includes a real-time communication session established based on an instant messaging platform. In some embodiments, the instant messaging platform may include a personal instant messaging platform or an enterprise instant messaging platform, and this embodiment of the disclosure does not impose specific limitations on the instant messaging platform.

[0089] In some embodiments, instant messaging sessions may include private chat sessions and group sessions. In some embodiments, a private chat session includes a one-to-one session between two parties. In some embodiments, a group session includes a many-to-many session involving multiple parties. In some embodiments, parties may be, for example, but not limited to, natural person users, bot accounts, or service accounts.

[0090] In some embodiments, session messages in an instant messaging session can be retrieved through the session archiving interface of the instant messaging session. By way of explanation and not limitation, enterprise instant messaging platforms may provide session archiving functionality, allowing authorized applications to retrieve session messages via callback interfaces. In one example, session message retrieval is implemented through an API (Application Programming Interface).

[0091] In some embodiments, the session archiving interface can provide real-time push or periodic retrieval of messages. In some embodiments, messages are pushed in real time, meaning that when a new message is generated in a session, the instant messaging platform can push the message to the processing apparatus / system implementing the session message processing method according to embodiments of this disclosure via a callback interface. In some embodiments, messages are retrieved periodically, meaning that the processing system can retrieve session messages from the instant messaging platform at preset time intervals.

[0092] In embodiments of this disclosure, conversation messages may include, but are not limited to, text messages and / or non-text messages. In some embodiments, text messages include message content presented in text form, and non-text messages include message content presented in non-text form, such as voice messages, image messages, file messages, and video messages. Accordingly, in some embodiments, the step of obtaining conversation messages in an instant messaging session may include: obtaining text messages and / or non-text messages in the instant messaging session.

[0093] In some embodiments, when the non-text message is a voice message, a speech recognition model can be used to convert the voice message into text data. In this embodiment, the speech recognition model may employ Automatic Speech Recognition (ASR) technology. In some embodiments, the speech recognition model may also load a domain-specific corpus to improve recognition accuracy. In some embodiments, when the non-text message is an image message, an image recognition model can be used to extract structured data from the image message. In this embodiment, the image recognition model may employ Optical Character Recognition (OCR) technology to extract text information from the image. In some embodiments, when the non-text message is a file message, a document parsing model can be used to extract structured data from the file message. By way of explanation and not limitation, the format of the file message may include, but is not limited to, xlsx, pdf, docx, etc. In some embodiments, the document parsing model may employ an appropriate parsing strategy according to the file format. In a specific example, for an xlsx file, cell data in the xlsx file can be read sequentially.

[0094] In some embodiments, when the instant messaging session is a group session, the step of obtaining session messages in the instant messaging session may include: obtaining multiple session messages of the group session. In this embodiment, at least some of the sessions (messages) in the multiple sessions (messages) come from different parties. For explanation, in a group session scenario, multiple parties may send messages simultaneously or sequentially in the group session, and the obtained session messages may come from different parties.

[0095] 120: Perform intent identification on session messages, identify target session messages with the target intent of requesting virtual resource configuration suggestions, and determine the target intent of the session party associated with the session message.

[0096] In this embodiment of the disclosure, virtual resources may include non-physical certificates of rights or assets. By way of explanation and not limitation, virtual resources may include, but are not limited to, various financial products, digital assets, virtual certificates of rights, etc., existing in data form.

[0097] In this disclosure, the virtual resource configuration recommendations may include suggestions on the configuration, adjustment, and optimization of virtual resources. In some embodiments, the virtual resource configuration recommendations may include investment advisory advice. By way of explanation and not limitation, the virtual resource configuration recommendations may include, but are not limited to, analysis and evaluation of currently held virtual resources, optimization suggestions for virtual resource configuration structures, recommendations for specific virtual resources, and interpretation and analysis of the virtual resource market status (related to the current virtual resources).

[0098] In this embodiment of the disclosure, the target intent of requesting virtual resource configuration advice may include the intent of a session party to request services such as virtual resource configuration advice. In some embodiments, a session message containing such intent may be identified as a target session message. As previously described, multiple session parties in a group session may express similar or related intents within a short period of time. In this embodiment, there may be multiple target session messages, which may be associated with at least two session parties. In this embodiment, the session parties associated with a session message may include session parties that are associated with the sending, interaction, or content of the session message. In some examples, the session parties associated with a session message may include the message sender that sent the session message, the message sender that referenced the session message, and the session parties mentioned by the session message. In some examples, mentioning a user account can be achieved through a message in the format "@user account".

[0099] In this embodiment of the disclosure, intent identification includes identifying the service request intent of the conversation party from the conversation messages. As an explanation and not a limitation, not all conversation messages carry explicit service intents. In instant messaging scenarios, conversation messages may include various types of content, such as: casual greetings (e.g., "Good morning"), casual conversation (e.g., weather discussions), service inquiries (e.g., "Could you check the recent market situation for me?"), and specific requests (e.g., "Recommend some configuration schemes suitable for me?"). In this embodiment of the disclosure, for example, in step 120 above, a target conversation message with a target intent can be identified / filtered from multiple conversation messages.

[0100] In some embodiments, intent identification may be achieved through rule matching and / or semantic matching.

[0101] In some embodiments, one or more sets of matching rules can be predefined, and session messages can be matched and judged according to the matching rules. In some embodiments, the matching rules may include keyword rules, regular expression rules, pattern matching rules, etc. In a specific example, the keyword rules include the following keywords: "recommendation", "suggestion", "configuration", and "analysis", thereby identifying messages with any of the above keywords as messages that may contain target intent.

[0102] In some embodiments, semantic analysis techniques can be used to understand the semantic content of conversational messages and match them with predefined intent categories. In this embodiment, semantic matching can handle messages with diverse expressions and flexible wording.

[0103] In some embodiments, intent recognition of session messages can be performed according to a preset semantic matching algorithm, which may specifically include the following steps A1, A2 and A3.

[0104] A1: Construct a demand vector based on the session messages.

[0105] In some embodiments, the semantic features of the session messages can be analyzed and a demand vector can be constructed. For explanation, a demand vector is a vectorized representation of a session message. In some embodiments, the demand vector can be constructed using, for example, but not limited to, bag-of-words models and word embedding models; this disclosure does not limit this approach.

[0106] A2: Calculate the similarity between the demand vector and multiple candidate intents in the preset intent library.

[0107] In some embodiments, the intent library may include a predefined set of intent categories. In this embodiment, the intent library may include multiple intent categories, each corresponding to a type of service request. By way of explanation and not limitation, intent categories may include, but are not limited to: requesting virtual resource configuration suggestions, querying virtual resource status, inquiring about virtual resource information, and requesting analysis of held resources.

[0108] In some embodiments, each intent category may correspond to one or more (candidate) intent vectors. In this embodiment, similarity can be calculated using methods such as cosine similarity, Euclidean distance, Manhattan distance, etc. In some embodiments, the TF-IDF cosine similarity algorithm can be used to calculate the similarity between the demand vector and the (candidate) intent vector.

[0109] A3: Based on the comparison results between similarity and preset matching threshold, determine the target intent that matches the conversation message.

[0110] In some embodiments, the intent category with the highest similarity exceeding a preset matching threshold can be identified as the target intent. In this embodiment, the preset matching threshold can be configured according to the actual application scenario. For example, a higher threshold can improve the accuracy of recognition, while a lower threshold can improve the recall rate. This disclosure does not limit the specific value of the matching threshold.

[0111] In optional embodiments, historical interaction data of the parties involved in the conversation can also be obtained, and the identification result of the target intent can be adjusted based on the historical interaction data. In some embodiments, historical interaction data may include the parties' past service request records, preference settings, behavioral patterns, etc. As an explanation, and not a limitation, the intent identification result of the current message can be adjusted based on historical interaction data. In a specific example, if the parties involved in the conversation have frequently consulted a certain type of virtual resource in the past, then when the intent identification result of the current message is unclear, it may be more likely to be identified as an intent related to that type of virtual resource.

[0112] In some embodiments, reference Figure 2 In the case of an instant messaging session being a group session, the steps of performing intent identification on the session message, identifying the target session message with the target intent of requesting virtual resource configuration suggestions, and determining the target intent of the session party associated with the session message may include the following steps 210 and 220.

[0113] 210: Identify multiple target session messages with the target intent of requesting virtual resource configuration suggestions.

[0114] In this embodiment of the disclosure, unlike private chat sessions, group sessions can involve multiple participants, and the session messages can originate from multiple different participants. In this embodiment, there can be multiple target session messages, and these multiple target session messages can be associated with at least two participants. As an explanation, and not a limitation, it is possible for multiple participants to express their intention to request virtual resource configuration suggestions simultaneously within the same time period.

[0115] In some embodiments, such as in step 210 above, intent recognition can be performed on multiple session messages in a group session to identify messages with a target intent to request virtual resource configuration suggestions. In this embodiment, a detailed description of intent recognition can be found in the specific description of intent recognition in the private chat scenario described above; the difference is that in a group session scenario, the recognition result may include multiple target session messages.

[0116] 220: Aggregate multiple target session messages to obtain at least one intent category.

[0117] In some embodiments, each intent category may correspond to a target intent. In this embodiment, multiple participants in a group session may express similar or related needs within a short period of time. By way of explanation and not limitation, if each target session message is processed independently, this repetition will result in significant processing overhead and make it difficult to provide a unified and efficient response to the relevant participants.

[0118] In some embodiments, such as 220 above, similar demand intents from multiple target session messages can be categorized into intent categories through aggregation processing. In this embodiment, each intent category can aggregate multiple target session messages with similar intents. As an explanation, and not a limitation, a corresponding virtual resource configuration suggestion can be generated for each intent category, rather than generating a suggestion separately for each target session message, thereby providing a unified response to relevant session parties quickly and efficiently.

[0119] In some embodiments, the aggregation process may include clustering multiple target session messages based on semantic similarity. Specifically, in this embodiment, each target session message can be represented as a semantic vector, and the semantic similarity between the target session messages can be calculated. Session messages with semantic similarity higher than a preset clustering threshold are then grouped into the same intent category. In some embodiments, the clustering algorithm may include, but is not limited to, K-means clustering, hierarchical clustering, and DBSCAN clustering; this disclosure does not impose any limitations on this.

[0120] In an alternative embodiment, the aggregation process may also consider a time factor. In one example, similar intent messages with shorter time intervals (e.g., within 5 minutes) may be preferentially grouped into the same intent category, while similar intent messages with longer time intervals may be grouped into different intent categories. This is to explain why the above embodiments ensure timely responses and prevent unexpected confusion caused by requests spanning long periods.

[0121] In some embodiments, the step of returning virtual resource configuration suggestions to the instant messaging session in a group chat scenario may differ from that in a private chat scenario, which will be described in detail below.

[0122] In this embodiment of the disclosure, for example, the target intent of the conversation party associated with the conversation message is determined through the intent recognition process in step 120, and the identified target intent can provide input for the subsequent generation of the task tree.

[0123] 130: Generate a task tree based on the target intent.

[0124] In this embodiment, the task tree includes a data structure for planning and organizing multiple processing tasks. In this embodiment, the task tree can define one or more processing tasks and their orchestration logic. As an explanation, and not a limitation, some known solutions employ a simple "intent-to-process" mapping mechanism, whereby after identifying a user's intent, the intent is directly mapped to the corresponding processing module or function, which then performs the corresponding processing and returns the result. However, this disclosure recognizes that such a simple mapping mechanism is only suitable for handling simple, single service requests, but has limitations when facing complex service requests involving multiple processing stages. Therefore, this disclosure adopts a different technical approach from known solutions, namely, using a task tree to plan multiple processing tasks and their orchestration logic required to complete the target intent.

[0125] In this embodiment, the task tree can define multiple processing tasks required to complete the target intent. As an explanation, and not a limitation, a complex service request may require multiple processing stages to work together. In one example, a session party requests virtual resource configuration suggestions, which requires: obtaining virtual resource status data, analyzing the session party's preference characteristics, recommending a configuration scheme based on the analysis results, and translating the professional analysis conclusions into easily understandable expressions. This disclosure uses a task tree to abstract these processing stages into multiple processing tasks, thereby clarifying the different responsibilities of each processing task (processing stage), and thus clarifying all the tasks required to complete the target intent.

[0126] In this embodiment of the disclosure, the task tree can also define orchestration logic for multiple processing tasks. In this embodiment, the orchestration logic includes the organization and coordination relationships between multiple processing tasks. In some embodiments, the orchestration logic includes, but is not limited to: the execution order of tasks, data dependencies between tasks, and the serial or parallel execution mode of tasks.

[0127] In some embodiments, the orchestration logic may include serial execution logic and / or parallel execution logic.

[0128] In some embodiments, serial execution logic can define multiple processing tasks to be executed sequentially in a predetermined order. In this embodiment, under serial execution logic, subsequent tasks are executed only after the preceding task is completed. As an explanation, and not a limitation, serial execution logic can be used when the execution of a subsequent task depends on the output of a preceding task. For example, a content generation task needs to be processed based on the output of a state analysis task; therefore, the content generation task should be executed after the state analysis task is completed.

[0129] In some embodiments, parallel execution logic can define multiple processing tasks to be executed simultaneously or substantially simultaneously. In this embodiment, under parallel execution logic, multiple tasks can be started concurrently and executed independently. By way of explanation and not limitation, parallel execution logic can be used to shorten the overall execution time when there are no data dependencies between multiple tasks. For example, tasks that retrieve data from multiple different data sources can be executed in parallel to reduce waiting time.

[0130] In some embodiments, the orchestration logic may also define data flow relationships. In this embodiment, the data flow relationship may include how the output data of a preceding task is passed to a subsequent task as input data. By way of explanation and not limitation, the orchestration logic may define one or more of the following data flow relationships: the output data of a preceding task is used as input data for a subsequent task; the output data of multiple preceding tasks is aggregated and used as input data for a subsequent task (aggregation mode); and the output data of a preceding task is distributed to multiple subsequent tasks as input data (distribution mode).

[0131] In some embodiments, task tree generation can be based on a predefined task tree template. In some embodiments, corresponding task tree templates can be predefined for different types of target intents. When a target intent is identified, the corresponding task tree template is selected according to the intent type, and the template is parameterized according to the specific situation to generate the final task tree.

[0132] In other embodiments, task tree generation can also be implemented based on a rule engine. In this embodiment, the rule engine can dynamically determine which processing tasks should be included in the task tree and how to orchestrate these tasks based on factors such as the characteristics of the target intent, the attributes of the session parties, and the current context. However, it is understood that other task tree generation methods can also be used, and this disclosure does not specifically limit the method of task tree generation.

[0133] In this embodiment of the disclosure, by introducing a task tree as a multi-task planning and orchestration mechanism, after identifying the target intent of the conversation party, it is not directly mapped to a single processing module. Instead, a task tree can be generated according to the target intent. Subsequently, according to the multiple processing tasks required to complete the target intent as defined by the task tree and the orchestration logic between these processing tasks, the agents corresponding to each processing task can be further scheduled to coordinate and execute in a predetermined order and manner, thereby achieving orderly connection of multi-stage processing.

[0134] 140: Based on the orchestration logic defined in the task tree, schedule the target agent corresponding to one or more processing tasks to execute one or more processing tasks and generate virtual resource configuration suggestions.

[0135] In some embodiments, virtual resource configuration suggestions can be generated by scheduling and processing the target agent corresponding to the task according to the orchestration logic defined in the task tree to perform the above-mentioned processing tasks.

[0136] In some embodiments, an agent includes a software module or service unit with specific processing capabilities. In some embodiments, each agent may undertake one or more types of processing tasks. In some embodiments, multiple agents may be pre-configured to form an agent pool.

[0137] In some embodiments, the multiple processing tasks may include resource processing tasks and content generation tasks.

[0138] In some embodiments, resource processing tasks may include tasks that process data related to virtual resources. In some embodiments, resource processing tasks may include, but are not limited to, resource status analysis tasks, resource holding analysis tasks, and resource scheme recommendation tasks. This disclosure does not limit the specific type of resource processing task. For explanation, resource status analysis tasks may include analyzing market status data of virtual resources; resource holding analysis tasks may include analyzing virtual resources held by a session party; and resource scheme recommendation tasks may include recommending suitable configuration schemes from candidate schemes based on the analysis results. These resource processing tasks will be described in detail below.

[0139] In some embodiments, a content generation task may include a task that semantically rewrites the output data of at least one resource processing task to generate virtual resource configuration suggestions. By way of explanation and not limitation, the output data of a resource processing task typically contains specialized analytical conclusions, technical indicators, or structured data results. The content generation task can transform this specialized output data into natural language expressions that are easily understood by the conversational parties, thereby facilitating their comprehension. The content generation task will be further described below.

[0140] In some embodiments, the orchestration logic may define the output data of at least one resource processing task as the input data of a content generation task. In this embodiment, the output data of the resource processing task may include state analysis results generated based on data related to virtual resources. In this embodiment, the data related to virtual resources may include environmental virtual resource data obtained from external virtual resource data sources and / or the resource holding data of the session party. In some embodiments, environmental virtual resource data includes, but is not limited to, resource indicator value data, resource circulation data, and resource indicator value change magnitude data; held resource data includes, but is not limited to, the types, quantities, and initial holding values ​​of the virtual resources held. In some embodiments, the state analysis results may include an analysis and evaluation of the current state of the virtual resources, such as, but not limited to, resource indicator value trend analysis, held resource structure analysis, and risk assessment.

[0141] In some embodiments, under the above orchestration logic, a target agent can be scheduled to perform resource processing tasks to generate state analysis results based on data related to virtual resources, and a content generation agent can be scheduled to perform content generation tasks to semantically rewrite the state analysis results and generate virtual resource configuration suggestions. As an explanation, and not a limitation, semantic rewriting can transform specialized state analysis results into expressions easily understood by the conversational parties. Through the data flow relationship established by the orchestration logic, the specialized analysis results generated by the resource processing tasks can be transformed into easily understandable configuration suggestions via the content generation tasks, thereby achieving an effective transformation between data-driven analysis processing and user-oriented content generation. In some embodiments, the agents may include, but are not limited to, data analysis agents, solution recommendation agents, content generation agents, resource holding analysis agents, customer tagging agents, demand clue agents, product comparison agents, and product interpretation agents, but the embodiments disclosed herein are not limited to these.

[0142] In other embodiments, step 140 above and the steps performed by the agent described below can also be implemented by a large model, such as a large language model with multimodal understanding capabilities, or by a comprehensive processing unit consisting of a large model, an intermediate layer, and sub-agents, which falls within the protection scope of the embodiments of this disclosure.

[0143] In some embodiments, such as step 140 above, target (sub) agents corresponding to each processing task can be determined from the agent pool based on the processing tasks defined in the task tree. In some embodiments, the step of determining the target agent can be implemented based on the matching relationship between task type and agent capabilities. For example, a resource status analysis task can be matched with a data analysis agent, a resource solution recommendation task can be matched with a solution recommendation agent, and a content generation task can be matched with a content generation agent.

[0144] In some embodiments, when or after the target agent is determined, the target agent can be scheduled to execute multiple processing tasks according to the orchestration logic defined in the task tree. In some embodiments, the steps of scheduling the target agent may include: sending a task execution request to the target agent, transmitting the input data required for the task, monitoring the task execution status, receiving the task execution results, and coordinating the execution order and data flow of multiple tasks according to the orchestration logic.

[0145] In some embodiments, a virtual resource configuration recommendation can be generated through the collaborative execution of multiple processing tasks. In some embodiments, the virtual resource configuration recommendation may include the combined output of the processing tasks. For explanation, the virtual resource configuration recommendation may integrate the processing results from multiple stages, such as data analysis, solution recommendation, and content generation.

[0146] In the embodiments of this disclosure, when dealing with different combinations of processing tasks, such as one or more of the aforementioned resource status analysis task, resource holding analysis task, resource scheme recommendation task, content generation task, etc., the session message processing method of the embodiments of this disclosure, such as step 140 above, may have different specific implementations. The following will describe several implementations of step 140 in conjunction with the embodiments.

[0147] In some embodiments, reference Figure 3 The step of scheduling and processing the target intelligent agent corresponding to the task according to the task tree, executing multiple processing tasks according to the orchestration logic, and generating virtual resource configuration suggestions may include the following steps 310 and 320.

[0148] In some embodiments, resource processing tasks may include resource status analysis tasks, and orchestration logic defines that the output data of the resource status analysis task can be used as input data for the content generation task. By way of explanation and not limitation, the resource status analysis task and the content generation task may adopt a sequential orchestration pattern, i.e., the resource status analysis task comes first, followed by the content generation task, and the two tasks are connected through data flow relationships.

[0149] 310: The scheduling data analysis agent performs resource status analysis tasks, obtains virtual resource data from at least one virtual resource data source, and generates status analysis results based on the virtual resource data.

[0150] In some embodiments, the data analysis agent includes an agent for data acquisition and analysis. In this embodiment, the data analysis agent may interface with one or more virtual resource data sources, acquire virtual resource data from the data sources, and analyze and process the acquired data.

[0151] In some embodiments, the virtual resource data source may include data services or data interfaces that provide virtual resource-related data. In this embodiment, the virtual resource data source may include, but is not limited to, a virtual resource management platform data interface, a third-party data platform interface, or an internal database. In some embodiments, the data analysis agent may obtain data from the virtual resource data source through API calls, database queries, or other methods.

[0152] In some embodiments, virtual resource data may include various types of data related to virtual resources. In some embodiments, virtual resource data may also be referred to as virtual resource status data. In some embodiments, virtual resource data may include, but is not limited to, resource indicator value data, resource circulation data, resource indicator value change range data, resource flow direction data, etc. It is conceivable that virtual resource data may include real-time data or historical data.

[0153] In some embodiments, the data analysis agent can acquire virtual resource data and then generate status analysis results based on the virtual resource data. In this embodiment, the status analysis results include an analysis and evaluation of the current status of the virtual resource. In some embodiments, the status analysis results may include trend analysis of virtual resource indicator values, virtual resource attention analysis, risk assessment, trend prediction, etc.

[0154] In some embodiments, the data analysis agent may periodically retrieve data from a virtual resource data source to maintain the timeliness of the data. In some embodiments, a timed fetching strategy may be pre-configured to automatically retrieve data at specific times (such as a fixed time each day). It is understood that the timed fetching strategy can be configured according to actual needs, such as setting the fetching frequency, fetching time window, etc., and this disclosure does not impose any limitations on it.

[0155] In some embodiments, the step of obtaining virtual resource data from at least one virtual resource data source and generating status analysis results based on the virtual resource data may include the following steps B1, B2, B3, and B4.

[0156] B1: Get the time information corresponding to the session message.

[0157] In some embodiments, time information may include the time the session message was sent or received.

[0158] B2: Match time information with multiple preset time intervals, and determine the target acquisition time period based on the matching results.

[0159] In some embodiments, the preset time interval may include multiple time periods. In a specific example, the preset time interval includes: a first time interval (before the start of the preset business day's business cycle), a second time interval (within the preset business day's business cycle), and a third time interval (after the preset business day's business cycle ends). It is understood that different time intervals may correspond to different target acquisition time periods. In some embodiments, when the session message sending time is before the start of the preset business day's business cycle, the target acquisition time period may include the interval from the end of the previous preset business day's business cycle to the current time to obtain the virtual resource data trend distribution of the previous data update cycle; when the session message sending time is within the preset business day's business cycle, the target acquisition time period may include the interval from the start time of the current cycle to the current time to obtain the real-time virtual resource data trend distribution within the current cycle.

[0160] B3: Obtain the virtual resource data corresponding to the target acquisition period from at least one virtual resource data source.

[0161] In some embodiments, virtual resource data for a given target time period can be obtained from a data source.

[0162] B4: Determine the target analysis mode based on the target acquisition time period, analyze and process the virtual resource data according to the target analysis mode, and generate status analysis results.

[0163] In some embodiments, different target acquisition time periods may correspond to different analysis modes. In some embodiments, a first analysis mode may be used for virtual resource data of a data update cycle. In this embodiment, the first analysis mode focuses, for example, on analyzing the impact of overnight major events on the virtual resource environment. In some embodiments, a second analysis mode may be used for the trend distribution of real-time virtual resource data within the current cycle. In this embodiment, the second analysis mode focuses, for example, on analyzing the changing trends of resource indicator values ​​and the flow of virtual resources within the current cycle.

[0164] In some embodiments, the first analysis mode may include a multi-dimensional in-depth analysis mode to perform in-depth analysis of data in multiple dimensions; the second analysis mode may include a single-dimensional conventional analysis mode to perform conventional analysis of data in the main dimensions.

[0165] In optional embodiments, the analysis and processing of virtual resource data may include threshold-based differentiated processing. In some embodiments, the magnitude of change in virtual resource data can be calculated and compared with a preset threshold, and the corresponding analysis and processing method can be selected based on the comparison result. In some embodiments, if the magnitude of change in virtual resource data exceeds the preset threshold (e.g., the magnitude of change in a virtual resource indicator value exceeds 2%), it can be determined that the virtual resource has experienced significant indicator value fluctuations, and a deep analysis mode can be used to generate more detailed status analysis results; if the magnitude of change does not exceed the preset threshold, it can be determined that the environmental fluctuations are small, and a conventional analysis mode can be used for analysis. As an explanation, and not a limitation, the above threshold triggering mechanism can make the analysis and processing more intelligent, automatically increasing the depth of analysis when significant changes occur in the market, and providing more valuable information to the parties involved.

[0166] 320: The content generation agent is scheduled to perform content generation tasks, perform semantic rewriting on the state analysis results, and generate virtual resource configuration suggestions.

[0167] In some embodiments, the content generation agent includes an agent for content generation and text processing. In this embodiment, the content generation agent may receive state analysis results as input and perform semantic rewriting processing on them to generate user-understandable (semantically rewritten) virtual resource configuration suggestions. In some embodiments, the virtual resource configuration suggestions include semantically rewritten state analysis results.

[0168] In some embodiments, semantic rewriting processing includes transforming specialized analytical conclusions into easily understandable expressions. As an explanation, and not a limitation, state analysis results may contain technical terms, technical indicators, numerical data, etc., which may be difficult for ordinary users to understand directly. Embodiments of this disclosure transform this technical content into natural language expressions through semantic rewriting processing, making virtual resource allocation recommendations easier to understand and accept. In one specific example, "short-term indicator values ​​fluctuate significantly" is rewritten as "short-term increases are significant, with a risk of correction," etc. In another specific example, "high concentration of held resources, with a single resource accounting for 65%" is rewritten as "Your held resources are relatively concentrated, with most resources allocated to a single product; it is recommended to appropriately diversify your allocation to reduce risk."

[0169] In some embodiments, semantic rewriting may include adjusting the expression style based on the characteristics of the conversation participants. In some embodiments, a target expression style may be determined based on the identity tags of the conversation participants, and the state analysis results may be semantically rewritten according to the target expression style. In a specific example, some technical terms may be retained for conversation participants with strong professional backgrounds; for ordinary users, a more colloquial expression style may be used.

[0170] In some embodiments, in private chat scenarios, the identity tags of the participants may include customer identity tags. In some embodiments, in group chat / group conversation scenarios, the identity tags of the participants may include customer group identity tags.

[0171] In an optional embodiment, the step of semantically rewriting the state analysis results and / or recommendation results may include the following steps C1 and C2. C1: Determine the target expression style based on the identity tag; C2: Perform semantic rewriting on the state analysis results and / or recommendation results according to the target expression style.

[0172] In some embodiments, the identity tag may include a risk preference tag. In this embodiment, the risk preference tag can be used to reflect a participant's risk tolerance and preference level. In some embodiments, a participant's risk preference can be inferred based on the resource structure distribution in the held resource data. In a specific example, if a participant holds a high proportion of high-risk resources (e.g., more than 60%) in their virtual resources, it can be inferred that the participant's risk preference is aggressive; if a participant holds a high proportion of medium-risk resources (e.g., 50% to 70%), it can be inferred that the participant's risk preference is moderate; and if a participant holds a high proportion of low-risk resources (e.g., more than 60%), it can be inferred that the participant's risk preference is conservative.

[0173] In some embodiments, identity tags may also include tags that reflect the speaker's knowledge background, level of expertise, and expression preferences. In some embodiments, identity tags may include expertise level tags (such as professional, average, novice) and expression preference tags (such as concise, detailed).

[0174] In some embodiments, the target expression style may include a semantic rewriting style determined based on identity tags. In some embodiments, the target expression style may include a professional style (retaining technical terms and using industry-standard expressions), a colloquial style (converting technical terms into colloquial expressions and using everyday metaphors), a concise style (point-by-point expression, omitting detailed explanations), and a detailed style (complete description, with accompanying explanations), etc. As an explanation, and not a limitation, by determining the target expression style based on identity tags and performing semantic rewriting according to that style, virtual resource configuration suggestions that are more in line with the reading habits and comprehension abilities of the conversation participants can be generated.

[0175] In this embodiment of the disclosure, through the above steps 310 and 320, the data analysis agent can acquire virtual resource data and generate state analysis results, and the content generation agent can perform semantic rewriting processing on the analysis results to generate virtual resource configuration suggestions, which is particularly suitable for scenarios that need to provide configuration suggestions based on real-time data status.

[0176] In some embodiments, the resource processing task may include a resource status analysis task and a resource scheme recommendation task. The orchestration logic defines the output data of the resource status analysis task and / or the output data of the resource scheme recommendation task as the input data of the content generation task. In this embodiment, a resource scheme recommendation task is added compared to the previous embodiments, forming a three-task serial or converged orchestration mode.

[0177] In some embodiments, reference Figure 4 The step of scheduling and processing the target intelligent agent corresponding to the task according to the task tree to execute multiple processing tasks according to the orchestration logic and generate virtual resource configuration suggestions may further include the following steps 410, 420 and 430.

[0178] 410: The scheduling data analysis agent performs resource status analysis tasks, obtains virtual resource data from at least one virtual resource data source, and generates status analysis results based on the virtual resource data.

[0179] In some embodiments, the specific description of step 410 can be found in the specific description of step 310 above, and will not be repeated here.

[0180] 420: The scheduling scheme recommendation agent performs the resource scheme recommendation task, obtains the identity tags of the session parties, and selects at least one recommended scheme from multiple candidate virtual resource configuration schemes based on the identity tags to obtain the scheme recommendation result.

[0181] In some embodiments, the solution recommendation agent includes an agent for recommending solutions. In this embodiment, the solution recommendation agent may select a suitable solution from multiple candidate solutions based on the characteristics of the session party. In some embodiments, the solution recommendation result may include at least one selected recommended solution and its related information.

[0182] In some embodiments, such as step 420 above, the identity tag may include a tagged description of the session party's characteristics. In this embodiment, the specific description of the identity tag can be found in step 320 and other specific descriptions above, and will not be repeated here.

[0183] In some embodiments, when or after the scheme recommendation agent obtains the identity tags of the participants, at least one recommended scheme can be selected from multiple candidate virtual resource configuration schemes based at least on the identity tags. In some embodiments, the selection of a recommended scheme can be based on the matching degree between the identity tags and the scheme characteristics. In a specific example, for a participant with a conservative risk appetite, a scheme with a lower risk level can be recommended; for a participant with an aggressive risk appetite, a scheme with higher expected growth potential but also higher risk can be recommended.

[0184] In some embodiments, there may be no data dependency between the resource status analysis task and the resource scheme recommendation task, meaning that steps 410 and 420 can be executed in parallel. This can be explained by the fact that parallel execution of steps 410 and 420 can shorten the overall execution time.

[0185] 430: The content generation agent is scheduled to perform content generation tasks, perform semantic rewriting on the state analysis results and / or recommendation results, and perform semantic merging on the state analysis results and recommendation results to generate virtual resource configuration suggestions.

[0186] In some embodiments, such as step 430 above, the content generation agent may receive state analysis results and solution recommendation results as inputs, and may perform semantic rewriting and semantic merging processing on these two inputs. In this embodiment, the virtual resource configuration suggestion may include the semantically rewritten and semantically merged state analysis results and recommendation results.

[0187] In some embodiments, the content generation agent can receive state analysis results and solution recommendation results as input, perform semantic rewriting processing on them, and generate user-understandable virtual resource configuration suggestions. In this embodiment, a detailed description of the semantic rewriting processing can be found in the relevant descriptions in step 320 and other sections above, and will not be repeated here.

[0188] In some embodiments, semantic merging processing includes integrating content from different sources to form a unified and coherent output. In some embodiments, the state analysis result may focus on the analysis of the current virtual resource market state, and may also be referred to as "first recommended content" or "virtual resource configuration direction recommendation"; the scheme recommendation result may focus on recommending specific configuration schemes, and may also be referred to as "second recommended content" or "virtual resource configuration scheme recommendation". In some embodiments, semantic merging processing can integrate these two parts into a complete virtual resource configuration recommendation, so that the recommendation content includes both the analysis of the current state and the recommendation of specific configuration schemes.

[0189] In some embodiments, the content generation agent may employ various merging strategies during semantic merging processing. In this embodiment, the merging strategies may include, but are not limited to, sequential splicing strategies (presenting state analysis first, followed by solution recommendations), interleaved presentation strategies (interleaving state analysis and solution recommendations to form an "analysis-recommendation" correspondence), and summary integration strategies (extracting summaries from state analysis and solution recommendations separately and then integrating them for presentation), etc.

[0190] In this embodiment, for example, the coordination of the three stages of state analysis, solution recommendation, and content generation is achieved through steps 410, 420, and 430 described above. As an explanation, and not a limitation, this embodiment is particularly suitable for scenarios requiring simultaneous state analysis and solution recommendation, providing more comprehensive virtual resource configuration suggestions to the parties involved in the session.

[0191] In some embodiments, the resource processing task may further include a resource holding analysis task, and the orchestration logic defines that the output data of the resource holding analysis task can be directly or indirectly used as the input data of the content generation task. In this embodiment, unlike the aforementioned embodiments that obtain data based on virtual resource data sources, analysis can be performed based on the session party's held resource data to achieve personalized services driven by held resource data.

[0192] In some embodiments, reference Figure 5 The step of scheduling and processing the target intelligent agent corresponding to the task according to the task tree to execute multiple processing tasks according to the orchestration logic and generate virtual resource configuration suggestions may further include the following steps 510, 520 and 530.

[0193] 510: Determine the resource preference tags of the session participants based on their resource holding data.

[0194] In some embodiments, the resource holding data includes data related to the virtual resources currently held by the session party. In some embodiments, the resource holding data may include, but is not limited to, information such as the type of virtual resources held by the session party, the quantity of virtual resources, the holding time, the cost, and the current indicator value. In this embodiment, the resource holding data can be used to characterize at least a portion of the virtual resources held by the session party.

[0195] In some embodiments, resource holding data can be obtained through various methods. In some embodiments, the virtual resource holding data of the session participants can be automatically synchronized by connecting to relevant systems. In some embodiments, the resource holding data of the session participants can be automatically obtained by connecting to an account management system, a virtual resource management system, or an asset management system. In other embodiments, the session participants can provide resource holding data by uploading screenshots of the held resources, inputting resource holding information, etc. In a specific example, the session participant uploads a screenshot of the held resources in an instant messaging session, and then the held resource data can be extracted through the aforementioned image recognition model.

[0196] In some embodiments, resource preference tags include tags generated based on held resource data that reflect the resource preference characteristics of the session participants. In some embodiments, resource preference tags may include, but are not limited to, risk preference tags, product preference tags, and held resource health tags. In this embodiment, a detailed description of the risk preference tag can be found in step 320 above and other sections, and will not be repeated here.

[0197] In some embodiments, product preference tags can be used to reflect a session participant's preference for a specific type of virtual resource. In some embodiments, product preference tags may include a preference for highly liquid virtual resources, a preference for resources with stable index values, and a preference for resources with fluctuating index values. In some embodiments, a session participant's product preference can be inferred based on the distribution of resource types in the resource holding data. In a specific example, if a session participant primarily holds a certain type of virtual resource (e.g., this type of virtual resource accounts for more than 50% of total assets), it can be inferred that the session participant has a preference for that type of virtual resource.

[0198] In some embodiments, the resource health label can be used to reflect the health of a session's current resource holding structure. In some embodiments, the resource health can be assessed based on factors such as the dispersion of held resources, risk concentration, and the degree of matching with the session's risk tolerance. In a specific example, if a session's held resources are overly concentrated in a single resource or a single type of resource, the resource health can be identified as "high concentration"; if the resource holding structure does not match the session's risk tolerance, the resource health can be identified as "risk mismatch".

[0199] In some embodiments, resource preference labels can be determined using a rule engine or a machine learning model. In some embodiments, a rule engine can generate corresponding labels based on predefined rules and the characteristics of the held resource data. In some embodiments, a machine learning model can be trained based on historical data to learn the mapping relationship between the characteristics of the held resource data and preference labels.

[0200] 520: The scheduling resource holding analysis agent executes the resource holding analysis task and generates the status analysis results of the virtual resources held by the session party based on the held resource data and resource preference tags.

[0201] In some embodiments, the resource holding analysis agent includes an agent for analyzing the resources held by a session party. In this embodiment, unlike the aforementioned data analysis agents that focus on analyzing overall market data, the resource holding analysis agent focuses, for example, on analyzing the resource holding situation of a specific session party.

[0202] In some embodiments, the resource holding analysis agent can generate state analysis results based on held resource data and resource preference tags. In this embodiment, the state analysis results may include, but are not limited to, performance evaluations of each held resource, rationality analysis of the held resource structure, matching degree analysis with resource preference tags, and potential risk warnings. By way of explanation and not limitation, the state analysis results are an analysis and evaluation of the virtual resources held by the session party.

[0203] In some embodiments, the resource holding analysis agent can analyze the held virtual resources from multiple dimensions when or after generating the state analysis results.

[0204] In some embodiments, the analysis dimensions may include one or more of the following dimensions: indicator performance dimension, risk dimension, structural dimension, and matching dimension. In some embodiments, the indicator performance dimension may include the indicator performance of each held resource, such as the absolute indicator change rate, the relative indicator change rate (compared to the market benchmark), and the indicator change during the holding period; the risk dimension may include the risk characteristics of each held resource, such as indicator volatility, maximum indicator decline, and risk level; the structural dimension may include the overall structure of the holdings, such as asset type distribution, industry distribution, and geographical distribution; the matching dimension may include the degree of matching between the held resource structure and the session participant's resource preference labels. By way of explanation and not limitation, in cases where the held resource structure and preference labels do not match, potential risks or optimization directions may be indicated in the state analysis results.

[0205] 530: The content generation agent is scheduled to perform content generation tasks, perform semantic rewriting on the state analysis results, and generate virtual resource configuration suggestions.

[0206] In some embodiments, the description of step 530 can be referred to the relevant description of step 320 above, and will not be repeated here. In this embodiment, the difference is that the state analysis result in step 530 comes from step 520 above, for example.

[0207] In this embodiment of the disclosure, for example, personalized analysis and suggestion generation based on the resource holding data of the session party are realized through the processing of steps 510, 520, and 530. As an explanation, the solution of this embodiment can provide session parties with analysis and suggestions tailored to their individual holdings, achieving a personalized service that is "personalized for each individual."

[0208] In some embodiments, the multiple processing tasks may further include a resource holding analysis task, a resource scheme recommendation task, and a content generation task. The orchestration logic may define the output data of the resource holding analysis task and / or the resource scheme recommendation task as the input data of the content generation task. In this embodiment, the resource scheme recommendation task is added compared to the embodiments of steps 510, 520, and 530 described above. In this embodiment, the resource scheme recommendation task is used in particular in conjunction with the resource holding analysis task.

[0209] In some embodiments, reference Figure 6 The step of scheduling and processing the target intelligent agent corresponding to the task according to the task tree to execute multiple processing tasks according to the orchestration logic and generate virtual resource configuration suggestions may further include the following steps 610, 620, 630 and 640.

[0210] 610: Determine the resource preference tags of the session participants based on their resource holding data.

[0211] In some embodiments, the specific description of step 610 above can be referred to the relevant description of step 510 above, and will not be repeated here.

[0212] 620: Schedule the resource holding analysis agent to perform the resource holding analysis task, and generate the status analysis results of the virtual resources held by the session party based on the held resource data and resource preference tags.

[0213] In some embodiments, such as step 620 above, the resource holding analysis results may include not only an analysis and evaluation of the state of the held resources, but also adjustment suggestions for the held resources. It is understood that in other embodiments, adjustment suggestions for non-held resources may also be optionally included.

[0214] In some embodiments, the adjustment recommendations include suggested directions for optimizing the resource holding structure based on state analysis. In this embodiment, the adjustment recommendations may include, but are not limited to, increasing the allocation ratio of a certain type of resource, decreasing the allocation ratio of a certain type of resource, and adjusting the risk exposure of the held resources.

[0215] In some embodiments, the resource holding analysis agent can generate adjustment suggestions based on multiple analysis dimensions. In this embodiment, a detailed description of the resource holding analysis agent and the analysis dimensions can be found in the relevant description in step 520 above, and will not be repeated here.

[0216] In some embodiments, adjustment recommendations may include optimization directions based on problems identified through state analysis. In a specific example, if analysis reveals that the risk level of held resources is higher than the risk preference of the session party, adjustment recommendations are generated to "reduce the proportion of high-risk virtual resources" or "increase the proportion of low-risk virtual resources." If analysis reveals that held resources are too concentrated, adjustment recommendations are generated to "increase the dispersion of virtual resource types" or "consider configuring different types of virtual resources." By way of explanation and not limitation, adjustment recommendations can provide input for subsequent solution recommendation tasks, enabling recommended solutions to correspond to the problems identified in the diagnosis, thereby achieving a complete solution from virtual resource holding diagnosis and virtual resource optimization direction to recommendation.

[0217] 630: The scheduling scheme recommendation agent executes the resource scheme recommendation task, and generates scheme recommendation results based on the resource analysis results and resource preference tags.

[0218] In some embodiments, such as step 630 above, the solution recommendation agent can generate solution recommendation results based on the resource holding analysis results and resource preference tags. In this embodiment, a detailed description of the solution recommendation agent can be found in the relevant description in step 420 above, and will not be repeated here. In this embodiment, unlike the recommendation based on identity tags in step 420 above, the solution recommendation agent in the above embodiment also considers adjustment suggestions from the resource holding analysis results when generating solution recommendation results.

[0219] In some embodiments, the step of generating a scheme recommendation result based on the resource holding analysis results and resource preference tags may include the following steps: screening at least one recommended scheme from multiple candidate resource configuration schemes whose risk characteristics match the resource preference tags to obtain a scheme recommendation result. In this embodiment, the virtual resources involved in the recommended scheme are at least partially different from the virtual resources held by the session party. For illustrative purposes, and not as a limitation, in this embodiment, the recommended scheme does not simply repeat the virtual resources already held by the session party, but rather recommends alternative or supplementary virtual resources that can optimize the structure of the held resources. In a specific example, if the session party currently holds a high proportion of high-risk virtual resources, the recommended scheme will include lower-risk virtual resources to reduce overall risk; if the session party currently holds resources that are too concentrated in a certain type of resource, the recommended scheme will include other types of virtual resources to increase the dispersion of the held virtual resources.

[0220] In other embodiments, the step of generating a scheme recommendation result based on the resource holding analysis results and resource preference tags may include: screening at least one recommended scheme from multiple candidate resource configuration schemes whose risk characteristics match the resource preference tags and meet the adjustment recommendations, to obtain the scheme recommendation result. In this embodiment, the virtual resources involved in the recommended scheme are at least partially different from the virtual resources held by the session party. As an explanation and not a limitation, in this embodiment, the above-described dual screening mechanism can ensure that the recommended scheme matches both the session party's risk preferences and meets the adjustment recommendations derived from the resource holding diagnosis.

[0221] In some embodiments, the step of selecting at least one recommended scheme from multiple candidate resource allocation schemes that matches the risk characteristics and resource preference labels and meets the adjustment recommendations may include the following steps D1 and D2. D1: Selecting schemes from the candidate resource allocation schemes whose risk characteristics match the resource preference labels to form a first selection result; D2: Further selecting schemes from the first selection result that meet the adjustment recommendations to form a final recommended scheme. For example, if the adjustment recommendation is "increase the proportion of low-risk virtual resource allocation," then schemes containing low-risk resources are prioritized in the first selection result; if the adjustment recommendation is "improve the dispersion of virtual resources," then schemes involving multiple virtual resource types are prioritized in the first selection result.

[0222] 640: The content generation agent is scheduled to perform content generation tasks, perform semantic rewriting on the state analysis results and / or solution recommendation results, and perform semantic merging on the state analysis results and solution recommendation results to generate virtual resource configuration suggestions.

[0223] In some embodiments, such as step 640 above, a detailed description of the content generation agent can be found in steps 430, 530, and other related details, and will not be repeated here. The difference in this embodiment is that the input received by the content generation agent in step 640 may include resource analysis results and solution recommendation results. The content generation agent can perform semantic rewriting and semantic merging processing on the above inputs (together). In some embodiments, the resource analysis results may include state analysis results and adjustment suggestions.

[0224] In some embodiments, the description of semantic rewriting and semantic merging processes can be found in the detailed descriptions of steps 430, 530, and other related content mentioned above, and will not be repeated here. The difference in this embodiment is that the input to the semantic rewriting and semantic merging processes is the content received by the content-generating agent based on the analysis results of its held resources and the solution recommendation results.

[0225] In this embodiment of the disclosure, the processing of the solution enables a complete "diagnosis" and virtual resource recommendation service based on the virtual resource data held by the session party. Specifically, the solution of this embodiment can not only analyze (diagnose) the current status of the session party's held resources, but also provide targeted optimization solutions based on the analysis results, thereby providing the session party with a one-stop personalized service.

[0226] In some embodiments, the step of generating a status analysis result of a session party's virtual resources based on the resource holding data and resource preference tags may include the following steps 710, 720, 730, and 740.

[0227] 710: Obtain virtual resource data corresponding to the virtual resource held by the session party from at least one virtual resource data source.

[0228] In some embodiments, the resource holding analysis agent can interface with a virtual resource data source to obtain environmental virtual resource data corresponding to the specific virtual resources held by the session party. In this embodiment, a detailed description of the virtual resource data source can be found in step 310 and other sections described above, and will not be repeated here.

[0229] In some embodiments, the resource holding data of a session party may include identification information of the virtual resources it holds. The resource holding analysis agent can then use this identification information to obtain the current environmental virtual resource data of these specific virtual resources from a virtual resource data source. In one specific embodiment, the resource holding analysis agent can use the resource code included in the session party's resource holding data to obtain the latest current value, price change, exchange volume, and net resource index value of that resource code in the current market from the virtual resource data source. By way of explanation and not limitation, by obtaining real-time environmental virtual resource data from an external data source, the resource holding analysis agent can combine the session party's held virtual resource data with the real-time environmental virtual resource data for analysis, thereby providing more accurate and timely status analysis results.

[0230] 720: Determine the resource status of the virtual resources held by the session party based on the corresponding virtual resource data.

[0231] In some embodiments, such as step 720 above, the resource status may include an assessment of the current performance of the virtual resource. In some embodiments, the resource status may include, but is not limited to, resource indicator value trend status, relative performance status, and risk status. In one example, the resource indicator value trend status includes one of increasing, decreasing, or fluctuating; the relative performance status includes one of being better than the environmental average, worse than the environmental average, or on par with the environmental average; and the risk status includes one of high risk, medium risk, or low risk.

[0232] In some embodiments, such as in step 720 above, the resource holding analysis agent can analyze the current state of each held resource based on the acquired virtual resource data.

[0233] In some embodiments, the trend status of resource indicator values ​​can be determined based on the changing trend of virtual resource indicator values. In one example, if a resource indicator value has been rising continuously over a period of time, its trend status can be marked as "rising," and vice versa; if a resource indicator value fluctuates within a certain range, it can be marked as "oscillating." For relative performance status, the rate of change of resource indicator values ​​can be compared with the virtual resource benchmark reference value. If the rate of change of resource indicator values ​​is higher than the virtual resource benchmark reference value, its relative performance status can be marked as "better than the market average," and vice versa; for risk status, it can be assessed based on risk indicators such as the fluctuation range of resource indicator values ​​and the maximum decline range of indicator values. If the annualized fluctuation range of resource indicator values ​​exceeds a preset high-risk threshold (e.g., 30%), its risk status can be marked as "high risk," if it is lower than a preset low-risk threshold (e.g., 10%), it can be marked as "low risk," and otherwise marked as "medium risk."

[0234] 730: Determine the resource risk characteristics of the session participants based on resource status and resource preference tags.

[0235] In some embodiments, resource risk characteristics include risk assessment results derived from analysis of the current state of held resources and the preferences of the parties involved. In some embodiments, resource risk characteristics may include resource allocation risk characteristics. By way of explanation and not limitation, resource risk characteristics may reflect, for example, but not limited to, the actual risk level of current holdings, the degree of matching between the actual risk level and the risk preferences of the parties involved, and the potential level of risk exposure.

[0236] In some embodiments, the resource holding analysis agent can calculate the overall risk level of the session's holdings based on the resource status of each held resource. In some embodiments, the overall risk level can be calculated using a weighted average method, where the weight is the proportion of each resource in the total held resources.

[0237] In some embodiments, the resource holding analysis agent can compare the calculated overall risk level with the virtual resource preference label of the session party to determine the virtual resource risk characteristics. In this embodiment, the virtual resource preference label may include a risk preference label. For explanation, when the overall risk level matches the risk preference label, the virtual resource risk characteristic can be identified as "risk matched," meaning the current held resource risk matches the session party's risk tolerance preference; when the overall risk level is higher than the risk level corresponding to the risk preference label, the virtual resource risk characteristic can be identified as "risk mismatch - high," meaning the current held resource risk is higher than the session party's risk tolerance preference; when the session party's virtual resource preference label is conservative, but the virtual resource status shows a high proportion of high-risk virtual resources, the virtual resource risk characteristic can be identified as "risk mismatch - high"; when the overall risk level is lower than the risk level corresponding to the risk preference label, the virtual resource risk characteristic can be identified as "risk mismatch - low," meaning the current held resource risk is lower than the session party's risk tolerance preference.

[0238] 740: Generate status analysis results of held resources based on resource risk characteristics.

[0239] In some embodiments, such as step 740 above, the state analysis results generated based on resource risk characteristics may include an assessment of the current risk status of held resources, an analysis of risk sources, and an explanation of the degree of matching between risk and preference. For explanation, the state analysis results can provide a basis for subsequent solution recommendations and content generation.

[0240] In some embodiments, the status analysis results may include one or more of the following: overall risk assessment, risk source analysis, matching degree analysis, and risk alerts. For explanation, overall risk assessment may include evaluating the overall risk level of the party's holdings and determining the current risk level of the holdings; risk source analysis may include analyzing the main factors leading to the current risk level; matching degree analysis may include the degree of matching between the risk of the currently held resources and the party's risk preferences; risk alerts may include risk alerts corresponding to the risk issues identified in the analysis.

[0241] In this embodiment of the disclosure, the solution described above enables the resource holding analysis agent to combine virtual resource data from the external environment and the preferences of the session participants to conduct in-depth analysis of the resources held by the session participants, thereby generating more accurate and targeted state analysis results.

[0242] In other embodiments, unlike the embodiments described above, the resource status can be determined independently of data obtained from an external virtual resource data source, based on the resource holding data itself. As an explanation, and not a limitation, the resource holding data provided by the session party (such as a screenshot of resource holding) may include information such as the current resource holding value and changes in indicator values, thereby allowing the resource status to be determined based on this information without additional access to an external data source.

[0243] Accordingly, in some embodiments, reference is made to Figure 8 The step of generating the status analysis results of the virtual resources held by the session party based on the resource holding data and resource preference tags may include the following steps 810, 820 and 830.

[0244] 810: Determine the resource status of at least a portion of the virtual resources held by the session party.

[0245] In some embodiments, the resource holding analysis agent extracts relevant information about each resource from the held resource data and analyzes and determines the resource status. In this embodiment, a detailed description of the virtual resource data source can be found in the preceding steps 710 and other sections, and will not be repeated here. The difference in this embodiment is that in step 810, the data source, independent of external virtual resource data sources, analyzes and determines the resource status of at least some virtual resources based on the held resource data itself.

[0246] In some embodiments, resource status may include, but is not limited to, indicator value increase / decrease status, resource holding weight status, and risk level status.

[0247] In some embodiments, the increase or decrease of the indicator value for each resource can be calculated based on the initial value and current value of resource holding in the resource holding data. In one example, the indicator value increase or decrease status is determined as "indicator value increase" when the current value of resource holding is higher than the initial value of resource holding, as "indicator value decrease" when the current value of resource holding is lower than the initial value of resource holding, and as "unchanged" when the two are close.

[0248] In some embodiments, the weight status of held resources can be determined based on the proportion of each resource in the total held resources. In one example, it can be determined whether there is a situation where the held resources are too concentrated: if the proportion of a single resource exceeds a preset threshold (30%), the weight status of the held resources is determined to be "high concentration".

[0249] 820: Determine the resource risk characteristics of the session participants based on resource status and resource preference tags.

[0250] In some embodiments, the virtual resource risk characteristics include a risk assessment result derived by comprehensively considering the current state of the held resources and the preference characteristics of the session participants. In this embodiment, a detailed description of the virtual resource risk characteristics can be found in the preceding steps 730 and other sections, and will not be repeated here.

[0251] In optional embodiments, the virtual resource risk characteristics may further include an assessment of risk concentration. In some embodiments, if a single resource accounts for an excessively high proportion in the holdings, the virtual resource risk characteristics may be further identified as "risk concentration" to prompt the session participants to pay attention to risk diversification.

[0252] 830: Based on the resource risk characteristics, determine the results of the resource holding analysis, including adjustment recommendations.

[0253] In some embodiments, such as step 830 above, the resource holding analysis results may include not only an analysis and evaluation of the current holdings but also adjustment recommendations. In this embodiment, the adjustment recommendations include optimization directions determined based on the risk characteristics of virtual resources.

[0254] In some embodiments, the resource holding analysis agent can generate adjustment suggestions based on resource risk characteristics. In some embodiments, adjustment suggestions can be configured for multiple resource risk characteristics respectively. In other embodiments, multiple resource risk characteristics can be aggregated to recall the most matching adjustment suggestion from a preset set of adjustment suggestions. As an explanation, and not a limitation, the resource holding analysis results can integrate state analysis and adjustment suggestions and provide input for subsequent solution recommendation tasks.

[0255] In one specific embodiment, when the resource risk characteristic is "risk mismatch - high", the adjustment suggestion is to "reduce the allocation ratio of high-risk virtual resources" or "increase the allocation ratio of low-risk virtual resources" to match the risk level of the held resources with the risk preferences of the parties. In another specific embodiment, when the virtual resource risk characteristic is "risk mismatch - low", the adjustment suggestion is to "appropriately increase the allocation ratio of virtual risky assets" or "consider allocating resources with higher potential for indicator value growth" to enhance the potential for indicator value growth in the holdings. In yet another specific embodiment, when the virtual resource risk characteristic includes "risk concentration", the adjustment suggestion is to "increase the diversification of virtual resources" or "reduce the allocation ratio of a single virtual resource" to diversify risk.

[0256] In some embodiments, the subsequent step of generating a scheme recommendation result based on the resource holding analysis results and resource preference tags may include: screening at least one recommended scheme from multiple candidate resource configuration schemes whose risk characteristics match the resource preference tags and which meet the adjustment recommendations, to obtain the scheme recommendation result. In this embodiment, the virtual resources involved in the recommended scheme are at least partially different from the virtual resources held by the session party. As an explanation, the scheme of the above embodiment implements a dual screening mechanism of diagnosis and recommendation: first, schemes whose risk characteristics match the preference tags are screened, and then schemes that meet the adjustment recommendations are screened from the matched schemes. This dual screening ensures that the recommended scheme is suitable for the session party's risk preferences and can specifically solve the problems found in the diagnosis.

[0257] In some embodiments of this disclosure, the solutions described in the above embodiments (e.g., steps 510 to 530, 610 to 640, 710 to 740, and 810 to 830) are implemented. Figures 5 to 8 (Corresponding embodiments) can realize personalized services based on resource data held by the session party. This service model can provide external data-driven solutions. Figure 3 and Figure 4 Corresponding embodiments) and internally held resource data-driven solutions ( Figures 5 to 8 (Corresponding embodiments) thereby enabling the provision of more accurate and personalized virtual resource configuration suggestions to the parties involved in the conversation.

[0258] In other embodiments of this disclosure, the aforementioned state analysis based on virtual resource data sources ( Figure 3 and Figure 4 Corresponding implementation examples) and position analysis based on resource holding data ( Figures 5 to 8 (Corresponding embodiments) can be used in combination. In one example, the overall virtual resource status data of the market and the resource holding data of the session party can be obtained simultaneously. After comprehensive analysis, a virtual resource allocation suggestion that reflects both the overall status of virtual resources and the individual holdings of the session party can be generated, which falls within the protection scope of this disclosure.

[0259] 150: Return virtual resource configuration suggestions to the instant messaging session.

[0260] In some embodiments, the virtual resource configuration suggestions generated in the preceding steps may be returned to the session participants. In some embodiments, the method and channel for returning the virtual resource configuration suggestions may be adapted to the type of instant messaging session.

[0261] In some embodiments, in a private chat scenario, virtual resource configuration suggestions can be sent directly to the private chat session, and the participants can view the received suggestions in the chat interface. By way of explanation and not limitation, the returned virtual resource configuration suggestions can be sent in the form of a text message, or in the form of a rich text message, card message, etc., to provide a better reading experience; this disclosure does not impose any limitations on this.

[0262] In some embodiments, the return of virtual resource configuration suggestions in a group session scenario can take several forms. In some embodiments, the virtual resource configuration suggestions can be sent to the group session as a regular message, which can be seen by all group members. In other embodiments, the virtual resource configuration suggestions can be published through a group announcement interface to present them as an announcement, thereby facilitating viewing by all group members and ensuring the persistent visibility of the message.

[0263] In some embodiments, in a group chat scenario, the step of returning a virtual resource configuration suggestion to an instant messaging session may include: returning a virtual resource configuration suggestion generated for at least one intent category to the group chat.

[0264] In an optional embodiment, targeted notifications may also be sent to the session parties associated with the target session message. By way of explanation, and not limitation, targeted notifications ensure that the relevant session parties are aware of the suggestions related to their request.

[0265] In some embodiments, targeted alerts may include steps such as tagging relevant session participants with @ in the message, sending separate alert messages to relevant session participants, and marking relevant session participants in group announcements. In some embodiments, when sending virtual resource configuration suggestions, the session participant who made the relevant request may be tagged with @ at the beginning of the message to receive a special alert; alternatively, a private message may be sent to the relevant session participant after sending a group message to remind them to check the group message.

[0266] In some embodiments, the returned virtual resource configuration recommendations may include various types of content. In some embodiments, the virtual resource configuration recommendations may include, but are not limited to, a first recommendation regarding the direction of resource configuration and a second recommendation regarding the resource configuration scheme.

[0267] In some embodiments, the first recommendation may include configuration direction suggestions based on virtual resource status analysis, i.e., macro-level configuration direction guidance. In some embodiments, the first recommendation may include suggestions such as "pay attention to a certain type of virtual resource in the near future" or "it is recommended to appropriately control the configuration ratio of a certain type of virtual resource."

[0268] In some embodiments, the second recommendation may include specific configuration scheme recommendations derived from the scheme recommendations, that is, specific scheme recommendations focusing on the micro level. In some embodiments, the second recommendation may include "consider configuring virtual resource 1" or "recommend a certain virtual resource configuration scheme", etc.

[0269] In some embodiments, when a virtual resource configuration suggestion includes both a first suggestion and a second suggestion, the content generation agent can integrate the two parts into a coherent suggestion text through semantic merging processing. In this embodiment, the integrated suggestion text can be presented as a structure combining directional guidance with specific solutions, thereby enabling the parties involved in the conversation to understand both the overall configuration direction and obtain specific solutions for reference.

[0270] In some embodiments, the returned virtual resource configuration recommendations may also include risk warnings. In some embodiments, the risk warnings may include: reminding the session party of the inherent risks of virtual resource configuration, reminding the session party to make prudent decisions based on its own circumstances, and reminding the session party of the potential impact of changes in virtual resource indicator values, etc.

[0271] In some embodiments of this disclosure, the solutions described above can achieve a complete processing loop from obtaining session messages, identifying intent, generating a task tree, scheduling execution, to returning.

[0272] In optional embodiments, in addition to responding to proactive requests from session participants, proactive push services can also be implemented. In some embodiments, scheduled tasks or triggering conditions can be pre-configured to proactively push virtual resource configuration suggestions to session participants when specific conditions are met.

[0273] In some embodiments, scheduled push notifications may include automatically executing push notifications according to a preset time schedule. In some embodiments, a market analysis summary may be automatically generated and proactively pushed to subscribers of the service at a specific time on each preset business day (e.g., before the start of a business cycle or after the end of a business cycle).

[0274] In some embodiments, triggering push notifications may include automatically triggering push notifications when specific conditions are detected. In some embodiments, virtual resource abnormal fluctuation trigger conditions may be pre-configured, thereby automatically triggering analysis and proactively pushing notifications to relevant session parties when a large fluctuation in virtual resource indicator values ​​is detected (such as a change in virtual resource indicator values ​​exceeding a preset threshold, for example, 2%). As an explanation, and not a limitation, compared to the traditional passive response mode, proactive push services have stronger service initiative and timeliness, providing indicator value information in advance before the session party has even noticed the change in virtual resource indicator values.

[0275] In some embodiments, the proactive push service can be combined with the virtual resource data held by the session party. In some embodiments, when a significant fluctuation is detected in a virtual resource held by the session party, analysis and suggestions regarding that virtual resource can be proactively pushed to the session party. In some embodiments, personalized proactive push may include the following processes: monitoring the virtual resource data held by the session party; automatically triggering analysis processing when the virtual resource held by the session party is detected to meet preset conditions (such as a single-day price change exceeding a threshold, a cumulative price change exceeding a threshold, or triggering a preset indicator value change trigger condition); generating targeted analysis and suggestions based on the session party's resource data and the triggering event; and proactively pushing the analysis and suggestions to the session party.

[0276] In some embodiments, the proactive push service may support subscription and preference settings for the parties involved in the conversation. In some embodiments, the parties involved in the conversation may choose to subscribe to or unsubscribe from the proactive push service, specify the timing of the push, and select the content preferences for the push.

[0277] In an optional embodiment, the instant messaging-based session message processing method may include: deploying agents corresponding to multiple processing tasks in a localized environment; and / or performing data compliance processing on session messages and / or virtual resource configuration suggestions.

[0278] In some embodiments, localized deployment includes deploying the agent in a local data center or private cloud environment. By way of explanation and not limitation, localized deployment prevents session data and analytics data from leaving the local environment, avoiding the risk of data interception or leakage during transmission; it also better protects sensitive data such as personal and asset information of the parties involved in the session and meets compliance requirements.

[0279] In some embodiments, data compliance processing may include processing data to comply with regulatory and policy requirements. In some embodiments, data compliance processing may include, but is not limited to, de-identification of sensitive information, audit logging, encrypted data storage, and access control. As an explanation and not a limitation, the methods of this disclosure embodiment, based on the above optional embodiments, can be extended and enhanced in multiple aspects to meet the needs of different application scenarios.

[0280] The instant messaging-based session message processing method according to the above embodiments of this disclosure can have at least some of the following technical effects:

[0281] The instant messaging-based conversation message processing method of this disclosure can identify target conversation messages with the target intent of requesting virtual resource configuration suggestions by performing intent recognition on conversation messages. This allows for accurate identification of target conversation messages with service intent from conversation messages containing greetings, casual conversation, etc., providing accurate input for subsequent processing. Furthermore, in group conversation scenarios, it aggregates multiple target conversation messages, classifying similar intents into intent categories, avoiding redundant processing and improving processing efficiency in group conversation scenarios.

[0282] In this embodiment of the disclosure, a task tree is also generated according to the target intent. The task tree defines multiple processing tasks and orchestration logic. Compared with the simple "intent to processing" mapping mechanism, this scheme realizes the planned processing of complex service requests: the task tree mechanism can define the execution order of multiple tasks, data dependencies, serial or parallel execution methods, etc., so that multiple agents can work together according to the predetermined orchestration logic, avoiding the problems of broken processing flow or incomplete results.

[0283] In this embodiment of the disclosure, multiple processing tasks are also scheduled to be performed by the target intelligent agent according to the task tree. This enables the coordinated execution of specialized intelligent agents such as data analysis intelligent agents, solution recommendation intelligent agents, and content generation intelligent agents to achieve a complete chain of data acquisition, analysis and processing, solution recommendation, and content generation, thereby generating high-quality virtual resource configuration suggestions.

[0284] In a further embodiment of this disclosure, a data analysis agent acquires real-time virtual resource data from an external data source and generates state analysis results based on the real-time data. Simultaneously, a content generation agent performs semantic rewriting on the analysis results, transforming the professional analysis conclusions into easily understandable expressions. This embodiment's solution can provide configuration suggestions based on real-time data, ensuring the timeliness of the suggestions; at the same time, semantic rewriting improves the comprehensibility of the suggestions, enabling the parties involved in the conversation to easily understand the professional analysis conclusions.

[0285] In a further embodiment of this disclosure, resource preference tags are determined based on the resource data held by the session participants, and these tags are applied to subsequent analysis and recommendation tasks. Compared to services based on general rules, this embodiment enables personalized services: services based on resource data fully consider the personalized characteristics of the session participants, generating configuration suggestions that match their resource preferences, significantly improving the accuracy of configuration suggestions and achieving a "personalized" service.

[0286] In a further embodiment of this disclosure, by providing adjustment suggestions in addition to the analysis of current holdings, the resource holding analysis results can simultaneously consider risk preference matching and the suitability of adjustment suggestions when recommending a solution. This embodiment provides a complete chain from resource holding diagnosis and optimization direction to dual screening, ensuring that the recommended solution is suitable for the client's risk preferences and can specifically address the problems identified in the diagnosis.

[0287] In this embodiment of the disclosure, a conversational message processing apparatus 900 based on instant messaging is also provided. In some embodiments, reference is made to... Figure 9 The instant messaging-based session message processing device 900 may include a session acquisition unit 910, an intent recognition unit 920, a task tree unit 930, a scheduling unit 940, and a return unit 950.

[0288] In some embodiments, the session acquisition unit 910 is configured to acquire session messages in an instant messaging session. In some embodiments, the session acquisition unit 910 is configured to perform the processing of step 110 in the foregoing method embodiments.

[0289] In some embodiments, the intent recognition unit 920 is configured to perform intent recognition on session messages, identify target session messages with a target intent requesting virtual resource configuration suggestions, and determine the target intent of the session party associated with the session message. In some embodiments, the intent recognition unit 920 is configured to perform the processing of step 120 in the foregoing method embodiments.

[0290] In some embodiments, the task tree unit 930 is configured to generate a task tree according to a target intent. In this embodiment, the task tree defines multiple processing tasks and the orchestration logic for these processing tasks. In some embodiments, the task tree unit 930 is configured to perform the processing of step 130 in the foregoing method embodiments.

[0291] In some embodiments, the scheduling unit 940 is configured to schedule a target agent corresponding to one or more processing tasks to execute one or more processing tasks according to the orchestration logic defined in the task tree, thereby generating a virtual resource configuration suggestion. In some embodiments, the scheduling unit 940 is configured to perform the processing of step 140 in the foregoing method embodiments.

[0292] In some embodiments, the return unit 950 is configured to return a virtual resource configuration suggestion to the instant messaging session. In some embodiments, the return unit 950 is configured to perform the processing of step 150 in the foregoing method embodiments.

[0293] In the embodiments of this disclosure, each unit in the instant messaging-based session message processing apparatus 900 can be implemented in software, hardware, or a combination of both. In some embodiments, in the case of software implementation, each unit may include a software module running on a processor. In some embodiments, in the case of hardware implementation, each unit may include dedicated hardware circuitry or chips. In some embodiments, in the case of a combined software and hardware implementation, some units may be implemented in software, and some units may be implemented in hardware. Further features and execution details of each unit of the instant messaging-based session message processing apparatus 900 in the embodiments of this disclosure can be found in the description of the foregoing method embodiments.

[0294] The instant messaging-based conversational message processing apparatus and its components, modules, units, and features described in the embodiments of this disclosure can be incorporated into the instant messaging-based conversational message processing method of the embodiments of this disclosure in a non-contradictory manner, and will not be elaborated here. Conversely, the instant messaging-based conversational message processing method and its steps, sub-steps, and features described in the embodiments of this disclosure can also be incorporated into the instant messaging-based conversational message processing apparatus of the embodiments of this disclosure in a non-contradictory manner.

[0295] In this disclosure, an electronic device is also provided, including: a processor and a memory storing a computer program, the processor being configured to perform the method of any of the embodiments of this disclosure when running the computer program.

[0296] In some embodiments, reference Figure 10 The diagram illustrates a session messaging method based on instant messaging that can be implemented according to embodiments of the present disclosure, or an electronic device 1000 that implements embodiments of the present disclosure. In some embodiments, it may include more or fewer electronic devices than illustrated. In some embodiments, it may be implemented using a single or multiple electronic devices. In some embodiments, it may be implemented using cloud-based or distributed electronic devices.

[0297] In some embodiments, reference Figure 10The electronic device 1000 includes a processor 1001, which can perform various appropriate operations and processes based on programs and / or data stored in read-only memory (ROM) 1002 or loaded from storage portion 1008 into random access memory (RAM) 1003. The processor 1001 may be a multi-core processor or may contain multiple processors. In some embodiments, the processor 1001 may include a general-purpose main processor and one or more special coprocessors, such as a graphics processing unit (GPU), a neural network processor (NPU), a digital signal processor (DSP), etc. Various programs and data required for the operation of the electronic device 1000 are also stored in RAM 1003. The processor 1001, ROM 1002, and RAM 1003 are interconnected via bus 1004. An input / output (I / O) interface 1005 is also connected to bus 1004.

[0298] The processor and memory described above are used together to execute a program stored in the memory. When the program is executed by a computer, it can implement the steps or functions of the methods described in the above embodiments.

[0299] The following components are connected to I / O interface 1005: an input section 1006 including a keyboard, mouse, touchscreen, etc.; an output section 1007 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 1008 including a hard disk, etc.; and a communication section 1009 including a network interface card such as a LAN card, modem, etc. The communication section 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to I / O interface 1005 as needed. A removable medium 1011, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 1010 as needed so that computer programs read from it can be installed into storage section 1008 as needed. Figure 10 The diagram only shows a portion of the components and does not imply that the computer system 1000 only includes... Figure 10 The components shown.

[0300] The systems, devices, modules, or units described in the above embodiments can be implemented by a computer or its associated components. The computer may be, for example, a mobile terminal, smartphone, personal computer, laptop computer, in-vehicle human-machine interface device, personal digital assistant, media player, navigation device, game console, tablet computer, wearable device, smart TV, Internet of Things system, smart home, industrial computer, server, or a combination thereof.

[0301] Although not shown, in embodiments of this disclosure, a program product is provided, the program product comprising a computer program configured to be run to implement the methods of any embodiment of this disclosure.

[0302] Although not shown, in embodiments of this disclosure, a storage medium is provided storing a computer program configured to be executed to implement the methods of any of the embodiments of this disclosure.

[0303] Storage media in embodiments of this disclosure include articles that are permanent and non-permanent, removable and non-removable, capable of storing information by any method or technology. Examples of storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0304] The methods, programs, systems, apparatuses, etc., of the embodiments of this disclosure can be executed or implemented in a single or multiple networked computers, or practiced in a distributed computing environment. In the embodiments of this specification, in these distributed computing environments, tasks can be performed by remote processing devices connected via a communication network.

[0305] Those skilled in the art will understand that the embodiments described in this specification can be provided as methods, systems, or computer program products. Therefore, those skilled in the art will realize that the functional modules / units or controllers and related method steps described in the above embodiments can be implemented in software, hardware, or a combination of both.

[0306] Unless explicitly stated otherwise, the actions or steps of the methods or procedures described in the embodiments of this disclosure do not necessarily have to be performed in a specific order and can still achieve the desired results. In some implementations, multitasking and parallel processing are also possible or may be advantageous.

[0307] This document describes several embodiments of the present disclosure; however, for the sake of brevity, the descriptions of the embodiments are not exhaustive, and identical or similar features or portions between the embodiments may be omitted. In this document, "one embodiment," "some embodiments," "example," "specific example," or "some examples" refers to at least one embodiment or example applicable to the present disclosure, but not all embodiments. The above terms do not necessarily refer to the same embodiment or example. Without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of the different embodiments or examples.

[0308] The exemplary systems and methods of this disclosure have been specifically shown and described with reference to the foregoing embodiments, and are merely examples of the best mode for implementing the systems and methods. Those skilled in the art will understand that various changes can be made to the embodiments of the systems and methods described herein without departing from the spirit and scope of this disclosure as defined in the appended claims when implementing the systems and / or methods.

Claims

1. A method for processing conversational messages based on instant messaging, characterized in that, include: Retrieve session messages from an instant messaging session; The session messages are subjected to intent recognition to identify target session messages with the target intent of requesting virtual resource configuration suggestions, and to determine the target intent of the session party associated with the target session message. A task tree is generated based on the target intent, and the task tree defines one or more processing tasks and the orchestration logic of the one or more processing tasks; Based on the orchestration logic defined in the task tree, the target intelligent agent corresponding to the one or more processing tasks is scheduled to execute the one or more processing tasks, and virtual resource configuration suggestions are generated. The virtual resource configuration suggestion is returned to the instant messaging session.

2. The session message processing method according to claim 1, characterized in that, The processing tasks are multiple and include resource processing tasks and content generation tasks; The orchestration logic defines at least one of the output data of the resource processing task as the input data of the content generation task, wherein the output data includes state analysis results generated based on data related to virtual resources.

3. The session message processing method according to claim 2, characterized in that, The resource processing task includes a resource status analysis task, and the orchestration logic defines the output data of the resource status analysis task as the input data of the content generation task; The step of scheduling target agents corresponding to the one or more processing tasks to execute the one or more processing tasks according to the orchestration logic defined in the task tree, and generating virtual resource configuration suggestions, includes: The scheduling data analysis agent executes the resource status analysis task, obtains virtual resource data from at least one virtual resource data source, and generates status analysis results based on the virtual resource data; The content generation agent is scheduled to execute the content generation task, perform semantic rewriting on the state analysis results, and generate virtual resource configuration suggestions.

4. The session message processing method according to claim 2, characterized in that, The resource processing task includes a resource status analysis task and a resource solution recommendation task. The orchestration logic defines the output data of the resource status analysis task and / or the output data of the resource solution recommendation task as the input data of the content generation task. The step of scheduling target agents corresponding to the one or more processing tasks to execute the one or more processing tasks according to the orchestration logic defined in the task tree, and generating virtual resource configuration suggestions, includes: The scheduling data analysis agent executes the resource status analysis task, obtains virtual resource data from at least one virtual resource data source, and generates status analysis results based on the virtual resource data; The scheduling scheme recommendation agent executes the resource scheme recommendation task, obtains the identity tag of the session party, and selects at least one recommended scheme from multiple candidate virtual resource configuration schemes based on the identity tag to obtain the scheme recommendation result; The content generation agent is scheduled to execute the content generation task, perform semantic rewriting on the state analysis results and / or the recommendation results, and perform semantic merging on the state analysis results and the recommendation results to generate the virtual resource configuration suggestion.

5. The session message processing method according to claim 2, characterized in that, The resource processing task includes a resource holding analysis task, and the orchestration logic defines the output data of the resource holding analysis task as the input data of the content generation task; The step of scheduling target agents corresponding to the one or more processing tasks to execute the one or more processing tasks according to the orchestration logic defined in the task tree, and generating virtual resource configuration suggestions, includes: Based on the resource holding data of the session party, the resource preference tag of the session party is determined, wherein the resource holding data represents at least a portion of the virtual resources held by the session party; The scheduling resource holding analysis agent executes the resource holding analysis task and generates a status analysis result of the virtual resources held by the session party based on the held resource data and the resource preference tags. The content generation agent is scheduled to execute the content generation task, perform semantic rewriting on the state analysis results, and generate the virtual resource configuration suggestions.

6. The session message processing method according to claim 2, characterized in that, The resource processing task includes a resource holding analysis task and a resource solution recommendation task. The orchestration logic defines the output data of the resource holding analysis task and / or the resource solution recommendation task as the input data of the content generation task. The step of scheduling target agents corresponding to the one or more processing tasks to execute the one or more processing tasks according to the orchestration logic defined in the task tree, and generating virtual resource configuration suggestions, includes: Based on the resource holding data of the session party, the resource preference tag of the session party is determined, wherein the resource holding data represents at least a portion of the virtual resources held by the session party; The scheduling resource holding analysis agent executes the resource holding analysis task, and generates a status analysis result of the virtual resources held by the session party based on the held resource data and the resource preference tags. The held resource analysis result includes adjustment suggestions for the held resources. The scheduling scheme recommendation agent executes the resource scheme recommendation task and generates scheme recommendation results based on the resource holding analysis results and the resource preference tags; The content generation agent is scheduled to execute the content generation task, perform semantic rewriting on the state analysis results and / or the scheme recommendation results, and perform semantic merging on the state analysis results and the scheme recommendation results to generate the virtual resource configuration suggestion.

7. The session message processing method according to claim 5 or 6, characterized in that, The step of generating the status analysis result of the virtual resources held by the session party based on the held resource data and the resource preference tags includes: Obtain virtual resource data corresponding to the virtual resources held by the session party from at least one virtual resource data source; The resource status of the virtual resources held by the session party is determined based on the corresponding virtual resource data; Based on the resource status and the resource preference tags, the resource risk characteristics of the session party are determined; Based on the resource risk characteristics, a status analysis result of the held resources is generated.

8. The session message processing method according to claim 6, characterized in that, The step of generating a scheme recommendation result based on the resource holding analysis results and the resource preference tags includes: From multiple candidate resource configuration schemes, at least one recommended scheme is selected that matches the risk characteristics with the resource preference label to obtain the scheme recommendation result, wherein the virtual resources involved in the recommended scheme are at least partially different from the virtual resources held by the session party.

9. The session message processing method according to claim 6, characterized in that, The step of generating resource holding analysis results based on the held resource data and the resource preference tags includes: Determine the resource status of at least a portion of the virtual resources held by the session party; Based on the resource status and the resource preference tags, the resource risk characteristics of the session party are determined; Based on the resource risk characteristics, determine the analysis results of the held resources, including the adjustment recommendations; The step of generating a scheme recommendation result based on the resource holding analysis result and the resource preference label includes: From multiple candidate resource configuration schemes, at least one recommended scheme is selected that matches the risk characteristics with the resource preference label and conforms to the adjustment suggestions, to obtain the scheme recommendation result, wherein the virtual resources involved in the recommended scheme are at least partially different from the virtual resources held by the session party.

10. The session message processing method according to any one of claims 1 to 6, characterized in that, The instant messaging session is a group session; The step of obtaining session messages in an instant messaging session includes: obtaining multiple session messages from the group session, wherein at least some of the multiple sessions come from different parties. The step of performing intent recognition on the session messages, identifying target session messages with the target intent of requesting virtual resource configuration suggestions, and determining the target intent of the session party associated with the target session message, includes: Identify multiple target session messages with the target intent of requesting virtual resource configuration suggestions; The multiple target session messages are aggregated to obtain at least one intent category, and each intent category corresponds to a target intent.

11. Among them, Returning the virtual resource configuration suggestion to the instant messaging session includes: returning the virtual resource configuration suggestion generated for the at least one intent category to the group session.

12. A conversational message processing device based on instant messaging, characterized in that, include: The session acquisition unit is configured to acquire session messages in an instant messaging session. The intent recognition unit is configured to perform intent recognition on the session message, identify the target session message with the target intent of requesting virtual resource configuration suggestions, and determine the target intent of the session party associated with the target session message; The task tree unit is configured to generate a task tree based on the target intent, wherein the task tree defines multiple processing tasks and the orchestration logic of the multiple processing tasks; The scheduling unit is configured to schedule the target agent corresponding to the one or more processing tasks to execute the one or more processing tasks according to the orchestration logic defined in the task tree, and generate virtual resource configuration suggestions. The return unit is configured to return the virtual resource configuration suggestion to the instant messaging session.

13. An electronic device, characterized in that, include: A processor and a memory storing a computer program, the processor being configured to implement the method as described in any one of claims 1 to 10 when the computer program is executed.

14. A program product comprising a computer program, wherein, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 10.

Images