Method and device for searching order information based on private domain large model, and program product

Abstract

The embodiment of the invention relates to the field of large models, and provides a method and a device for searching order information based on a private domain large model and a program product, applied to an order system based on the private domain large model, the method comprises the following steps: receiving a first input of a target user to a query icon, and displaying a session interface realized based on the private domain large model; receiving a second input, indicating order query information, of the target user on the session interface, and determining a first order query intention of the target user according to target key features of the order query information; determining a target query object according to the first order query intention; and according to the first order query intention and the target query object, obtaining first order information matched with the first order query intention by traversing an order system, and displaying the first order information on a session interface. According to the scheme, the condition screening mode is converted into semantic query order related information through a session mode, the order query efficiency and accuracy are improved, and the corresponding order operation is executed according to the queried order information.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of big model technology, and in particular, to a method, device and program product for searching order information based on a private domain big model. Background Art

[0002] In the field of e-commerce customer service, in the order system, it is often necessary to search based on order number, product name, merchant name and other information in order to quickly find the required order information. Merchants need to make a lot of inquiries in their daily work.

[0003] In the process of research and practice of the prior art, the inventors of the embodiments of the present application found that in the current order system, a large number of query conditions are set on the page, which requires manual selection, input and other repetitive operations on the page, which is time-consuming and labor-intensive, and the overall search efficiency is low. Summary of the invention

[0004] The embodiment of the present application provides a method, device and program product for searching order information based on a private domain big model. By implanting a private domain and multimodal intelligent customer service big model into the order system and interacting with users in a conversational manner, the intelligent customer service big model can respond to users' semantic inquiries and combine the general capabilities in the business system to help merchants find the required order information accurately, quickly and conveniently, and perform corresponding order operations according to the order status in the order information, thereby improving the work efficiency of merchants and saving the operating costs of the customer service system. It is especially suitable for application scenarios where batch orders are processed.

[0005] In a first aspect, an embodiment of the present application provides a method for searching order information based on a private domain big model, the method being applied to an e-commerce customer service system based on a multimodal big model, the method being applied to an order system based on a private domain big model, the order system currently displaying a first interface, the first interface including a query icon; the method comprising: Receiving a first input of the query icon by a target user; In response to the first input, displaying a conversation interface, wherein the conversation interface is implemented based on a private domain big model; receiving a second input from the target user in the conversation interface, where the second input indicates order query information, and the order query information is data in at least one mode; In response to the second input, extracting a target key feature of the order query information, and determining a first order query intention of the target user according to the target key feature, wherein the first order query intention may include at least one query condition for searching order information; Determining a target query object of the target user according to the first order query intention; Determining a target store associated with the target user; According to the first order query intent and the target query object, traverse the orders associated with the target store in the order system to obtain first order information matching the first order query intent; The first order information is displayed on the session interface.

[0006] In some implementations, the abnormal order is a logistics abnormal order, and the first order query is intended to query a second type of abnormal order associated with a second product of a target store within a second period of time; the method further includes: Traversing the logistics abnormality orders associated with the second product in the order system to obtain fourth order information; Verify the authenticity of the logistics status of each logistics abnormal order in the fourth order information, and filter the normal logistics orders; in the conversation interface, visualize the fourth order information according to the order status.

[0007] In a second aspect, an embodiment of the present application provides an order information processing device, which has the function of implementing the method for searching order information based on a private domain large model provided in the first aspect above. The function can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions, and the modules can be software and / or hardware.

[0008] In some implementations, the order information processing device includes an input / output module, a display module, and a processing module, wherein the processing module is used to control the input / output operation of the input / output module and the display operation of the display module: An input-output module, used for receiving a first input of the query icon by a target user; a processing module, configured to display a conversation interface through the display module in response to the first input received by the input / output module, wherein the conversation interface is implemented based on the private domain big model; the input / output module is further configured to receive a second input from the target user in the conversation interface, wherein the second input indicates order query information, and the order query information is data of at least one modality; The processing module is also used to extract the target key features of the order query information in response to the second input received by the input-output module, determine the first order query intention of the target user according to the target key features, and the first order query intention may include at least one query condition for searching order information; determine the target query object of the target user according to the first order query intention; determine the target store associated with the target user; and traverse the orders associated with the target store in the order system according to the first order query intention and the target query object to obtain the first order information matching the first order query intention; The display module is also used to display the first order information obtained by the processing module on the conversation interface.

[0009] In the third aspect, an embodiment of the present application provides an order information processing device, which includes: at least one processor and a memory; wherein the memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the above-mentioned first aspect, any one of the implementation methods of the first aspect, or any steps of the method for implementing the search for order information based on the private domain big model provided by the second aspect.

[0010] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a function of implementing the method for searching for order information based on a private domain large model provided in the first aspect above. The function can be implemented by hardware, or the corresponding software can be implemented by hardware. The hardware or software includes one or more modules corresponding to the above functions, and the modules can be software and / or hardware. Specifically, the computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for loading by a processor to execute the first aspect of the embodiment of the present application, any one of the implementation methods of the first aspect, or any step of the method for searching for order information based on a private domain large model provided in the second aspect.

[0011] Compared with the prior art, in the solution provided by the embodiment of the present application, since the intelligent customer service big model is based on multimodality, it can understand any modal data input by the user. Therefore, based on the preliminary identification of the first order query intention based on the target key features in the first input, on the one hand, the target user realizes the intelligent customer service big model to quickly locate the first order query intention of the target user through semantic inquiry, thereby significantly improving the query efficiency; on the other hand, since the order system of the e-commerce transaction platform is traversed directly based on the first order query intention, and the order information in the order system is all real transaction data recorded based on the consensus mechanism of the blockchain, therefore, the first order information output by the intelligent customer service big model is the real order information matching the first order query intention, thereby reducing or avoiding the invalid orders output by the intelligent customer service big model due to being attacked or the robustness to be improved. Usually, the invalid order is fabricated by the intelligent customer service big model itself, and the fabricated invalid order may be real, and it is impossible to distinguish the authenticity without careful observation, which will confuse the real order data and bring wrong reference data to the real-time query of the target user. In this way, the authenticity of the candidate answers presented to the target user on the conversation interface is guaranteed, and query efficiency and user experience are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1a A schematic diagram of a scenario of a communication system for implementing a method for searching order information based on a private domain big model provided in an embodiment of the present application; Figure 1b A schematic diagram of an interface for implementing order information search based on a conversational approach provided by this application; Figure 2a A flowchart of a method for searching order information based on a private domain big model provided in this application; Figure 2b A schematic diagram of query data enhancement performed by the query generation model provided in this application; Figure 2c A schematic diagram of a framework for adversarial training of a query generation model provided in this application; Figure 3 A schematic diagram of an interface for responding to after-sales inquiries about goods provided in an embodiment of the present application; Figure 4 A schematic diagram of an interface for responding to inquiries about the sales status of already distributed products provided by this application; Figure 5 A schematic diagram of an interface for responding to an inquiry about printing undelivered orders provided in an embodiment of the present application; Figure 6a A flowchart of a method for searching order information based on a private domain big model provided in this application; Figure 6b A schematic diagram of an interface for processing inquiries and printing of undelivered orders provided in an embodiment of the present application; Figure 7a A flow chart of a method for executing intelligent distribution of hot commodities based on order system search provided in an embodiment of the present application; Figure 7b A schematic diagram of an interface for performing intelligent distribution based on hot commodities in an order system provided in an embodiment of the present application; Figure 8 A schematic diagram of an interface for responding to financial data output provided in an embodiment of the present application; Figure 9a A schematic diagram of a process for executing intelligent distribution based on financial data searched by an order system according to an embodiment of the present application; Figure 9b A schematic diagram of an interface for executing intelligent distribution based on financial data searched by an order system provided in an embodiment of the present application; Fig.10 A schematic diagram of the structure of an order information processing device provided in an embodiment of the present application; Fig.11 A schematic diagram of a physical device structure for implementing a method for searching order information based on a private domain big model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0013] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0014] The terms "first", "second", etc. in the specification and claims of the embodiments of the present application and the above-mentioned drawings are used to distinguish similar objects (for example, the first input and the second input in the embodiments of the present application respectively represent different messages of the user input conversation interface), and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In some of the processes described in the embodiments of the present application and the drawings, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in the embodiments of the present application or may be executed in parallel. The serial numbers of the operations, such as 201, 202, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules that are clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. The division of modules that appear in the embodiments of the present application is only a logical division. There may be other division methods when implemented in actual applications. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between each other shown or discussed can be through some interfaces, and the indirect coupling or communication connection between modules can be electrical or other similar forms, which are not limited in the embodiments of the present application. In addition, the modules or submodules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed in multiple circuit modules, and some or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment of the present application.

[0015] The embodiments of the present application provide a method, apparatus and program product for searching order information based on a private domain big model, which can be used in servers or terminal devices. Specifically, it can be used in e-commerce trading platforms, social platforms, financial platforms, government platforms, retrieval systems and other platforms that need to introduce AI intelligent customer service to conduct efficient conversations with users. For example, it can be applied to a customer service system implanted in a business server. The intelligent customer service big model can conduct efficient conversations with users in real time offline or online, thereby accurately understanding the user's first order query intention, not limited to judging according to the rules preset by the e-commerce trading platform, timely avoiding or reducing misjudgments, and significantly improving the efficiency and quality of order queries, saving labor costs.

[0016] In some implementations, the present invention is applied to Figure 1aThe communication system shown in FIG. 1 may include at least one service server 1, a customer service system 2, and at least one user terminal 3 ( Figure 1a Only one user terminal is indicated in the figure). Among them, the customer service system 2 includes a query generation model (optional, not indicated in the figure), an adversarial generation model (optional, not indicated in the figure), and at least one intelligent customer service large model 22. The user terminal 3 can be a server or a terminal. When it is a terminal, it has a client installed. The client installed on each user terminal can be a client with the same or different version numbers, or a client with the same or different operating systems. The embodiment of the present application does not limit the version of the client and the type of operating system corresponding to the client. The following is a detailed description.

[0017] Business server 1: can be used to provide business services such as search, transaction, conversation, upload, download, etc. to users of each user terminal. For example, the server can provide users with commodity transactions (such as purchasing things on the e-commerce transaction platform), connect to customer service for real-time conversation, etc. The business server can be a platform for the e-commerce transaction platform to query order-related information. Business server 1 includes an order system.

[0018] Customer service system 2: used to respond to user order search requests. The query generation model can be used to perform data enhancement on the order query information received by the intelligent customer service big model 22 from the user and then feed it back to the intelligent customer service big model 22 to enhance the accurate recognition of the order query intention. It can be used to output the corresponding order information to the semantic query of user a of the user terminal 3 by the intelligent customer service big model 22. The customer service system 2 also includes a functional component library to support the intelligent customer service big model 22 to realize intelligent workflow when responding to the user's order query intention.

[0019] The adversarial generation model has two main functions in this application: generating multiple adversarial order query questions based on the prompt words, which are used to perform adversarial training on the intelligent customer service big model 22 to improve the robustness of the intelligent customer service big model 22; and performing adversarial training on the query generation of the query generation model to obtain candidate queries that are more natural and better match the first order query intent.

[0020] User terminal 3: Install a client that matches the server deployed in the business server, and use the client to communicate with the intelligent customer service model 22 in the customer service system to query order-related information. The user terminal of this application is equivalent to the terminal device and is not distinguished.

[0021] When based on Figure 1aWhen the communication system shown implements the above-mentioned method of searching order information based on the private domain big model, the intelligent customer service big model running in the business server as a plug-in can respond to the user's order query intention, and is not limited to responding according to the order condition screening rules preset by the e-commerce transaction platform, thereby significantly improving the order query efficiency. Figure 1b This is a schematic diagram of a conversation interface that pops up independently in the order system. The conversation interface is a conversation interface between the intelligent customer service model 22 and the user.

[0022] It can be understood that the customer service system in this application can be applied to the service side of the e-commerce trading platform, merchants, buyers and other roles that have the need to query order information. This application mainly takes the order information query of merchants, buyers and other people on the e-commerce trading platform as an example. Other scenarios can be used as a reference and will not be elaborated on.

[0023] The solution provided in the embodiment of the present application involves technologies such as artificial intelligence (AI), natural language processing (NLP), and machine learning (ML). In view of the conversation processing direction of the intelligent customer service big model in the field of artificial intelligence, the present application can implant a multimodal intelligent customer service big model into a business server. The intelligent customer service big model can be enhanced based on the first order query intention after initially identifying the first order query intention based on the user's query, and then using the query generation model based on the first order query intention to improve the accuracy of identifying the first order query intention through a dual mechanism, thereby improving the conversation efficiency and conversation quality.

[0024] It should be noted that the server (such as a business server, a search engine) involved in the embodiments of the present application may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The order information processing device involved in the embodiments of the present application may be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, a personal digital assistant, etc., but is not limited thereto.

[0025] In order to solve the problem of low order query efficiency in related technologies, the embodiments of the present application mainly adopt the following technical solutions: 1. Realize order information query through conversation: Introduce a multimodal intelligent customer service model into the order system. After identifying the first order query intention based on the user query, determine that the query object is the order system. According to the first order query intention, traverse the orders associated with the target store in the order system to obtain the first order information that matches the order query intention, and display it in the conversation interface for users to view intuitively, and provide the corresponding order operation entry.

[0026] Specifically, the customer service system running in the business server as a plug-in can conduct efficient conversations with users offline or online in real time. Users do not need to manually filter conditions in the order system. They only need to enter natural language or any modality of order query related information, and the intelligent customer service model can display the searched order related information to the user. This solution does not require users to reach a certain level of questioning skills or a certain level of professional description of needs in order to enable the machine to understand the first order query intention, significantly improving order query efficiency and order information query quality.

[0027] 2. Accurately identify the first order query intent: Introduce a multimodal intelligent customer service model, preliminarily identify the first order query intent based on the initial query, and then use the query generation model based on the first order query intent to strengthen the input, obtain multiple candidate queries, and further determine the first order query intent. Specifically, the customer service system running in the business server as a plug-in can conduct efficient conversations with users offline or online in real time. Users do not need to change the content of the input query multiple times to express their demands, which significantly reduces the time for users to inquire + get answers, and significantly improves quality inspection efficiency and customer service reception quality, saving labor costs.

[0028] 3. Improve the naturalness and accuracy of queries: By generating adversarial models to perform adversarial training on the query generation model, it is possible to ensure that the generated candidate queries are more natural and diverse, while also improving the effectiveness and accuracy of the candidate queries in the e-commerce order system query scenario.

[0029] The following combination Figure 2a-Figure 11 The technical solution of the embodiment of the present application is exemplified.

[0030] Since the method for searching order information based on a private domain big model in the embodiment of the present application processes the output of a pre-trained intelligent customer service big model (i.e., the answer to the order query related information input by the user), the training process of the intelligent customer service big model is introduced before introducing the method for searching order information based on a private domain big model. The training process includes: 1. Data Level 1. Data Collection 1.1 Historical customer service session data Collect historical chat records that have been stored in the customer service system, including user questions, customer service responses, and corresponding product information (such as product name, price, specifications, event information, etc.).

[0031] Extract the complete conversation context (including multiple rounds of conversation) and mark the conversations that can be clearly linked to the product or the user's ordering behavior (such as "check specifications", "add to cart", "ask about delivery time", etc.).

[0032] 1.2 Product Knowledge Base / Product Information Database Contains rich information such as product name, category, brand, specification, price, inventory status, discount information, logistics rules, etc. This data can be stored in a database or knowledge graph for real-time query in conversations.

[0033] 1.3. FAQ Question and Answer Data Summarize frequently asked questions (such as "how to return goods", "how to use coupons", "invoice issuance methods", etc.), including questions and standard answers. This part of the data can be used as part of the knowledge base and will also be combined with customer service conversation data.

[0034] 1.4. Customer behavior data (optional) If more refined user personalized services are needed, user behavior data can be integrated, including user browsing history, favorites, add-to-cart, purchase history, evaluation feedback, etc. This type of data is useful for personalized scenarios, such as multi-round related conversations such as "Have you restocked the item you bought last time?"

[0035] Pay attention to privacy and compliance: When collecting and using business data, you need to comply with relevant privacy policies and regulations, anonymize and protect sensitive information.

[0036] 2. Data cleaning and labeling 2.12 Data deduplication and noise filtering De-duplicate the collected historical conversation data and filter out conversations that are irrelevant to the business or have extremely high noise (such as obvious idle chats, insults, and test texts).

[0037] 2.2 Dialogue segmentation and structuring Divide the conversation into segments or rounds based on each user question and customer service response. Make sure you can build a multi-round data format, such as: { "role": "user", "content": "Hello, is this shoe available in black?" }, { "role": "assistant", "content": "Hello, this shoe is currently available in black and white. Which one do you want?" } At the same time, the corresponding product information is extracted, or the inventory, specifications and other information queried by the customer service in the system.

[0038] 3. Business intent and slot labeling (optional) In order to enable the model to better understand user intent and extract key information (such as product name, color, size, payment method, etc.), historical conversations can be classified and slots can be marked: Intent tags: inquiring about inventory, asking about prices, asking about delivery time, return and exchange consultation, coupon information, etc. Slots: product name, color, size, user address, return reason, etc.

[0039] This step can be completed through semi-automatic or manual annotation to ensure the quality of annotation.

[0040] 4. FAQ alignment and supplement Normalize the questions and answers in the FAQ to keep the content clear and the format consistent. For example, complete some incomplete questions or format the answers into paragraphs or lists.

[0041] 2. Model training and verification 1. Model selection and strategy 1.1 Pre-trained language model + custom fine-tuning The commonly used approach at present is to first select a large-scale pre-trained language model (such as a GPT model), and then fine-tune it based on e-commerce data.

[0042] 1.2 Retrieval vs. Generative Combination In practical applications, a common approach is to combine retrieval and generation methods.

[0043] When the questions raised by users clearly require business data support (such as dynamic data such as inventory, price, order status, etc.), the customer service system first searches the database or knowledge base to obtain the answer, and then integrates the search results into the model response.

[0044] For fixed questions such as FAQs and general questions and answers, the search mode can be used to quickly provide templates or standard answers.

[0045] For complex or open-ended questions, generative models are used to understand and generate responses.

[0046] 2. Strengthen the connection between models and business data 2.1 Knowledge base / database real-time query Before generating an answer, the big model first searches and queries the knowledge base / database for specific intents or keywords (such as "Is this product in stock?", "What is its material?", "What is the promotional price?").

[0047] Return structured information (such as inventory = 123, price = 199 yuan) to the model as context or prompt for the generative model to use. This allows the model to provide the latest and most accurate business data in its answers.

[0048] 2. 2Prompt Engineering / Model Interface Design When users interact with the intelligent customer service model, business rules, data query results, product information, etc. are injected into the model context through prompts. For example: "You are now a customer service representative of a certain e-commerce platform. Please answer the following questions based on the product information and the user's question. The product information includes: name =..., price =..., inventory =..., and the following is the user's question:..." By injecting policies into prompts or system information, the model can prioritize the use of the provided business data for answers and limit the answer format (for example, requiring the provision of links / specification information, etc.).

[0049] 2.3 Fine-tune the data to highlight key business entities In the data of fine-tuning, a large number of real "user questions + query results + final customer service answers" triples can be placed. Through fine-tuning, the large model can learn: when the user's questions contain certain keywords, how to use the corresponding business information to answer.

[0050] 3. Training Process 3.1 Data Partitioning The complete corpus is divided into training set, validation set and test set. The typical division method can be 8:1:1 or 7:2:1, etc., to ensure generalization ability.

[0051] 3.2 Fine-tuning stage When based on a generative large language model (LLM), command fine-tuning or dialogue fine-tuning is usually done first: (1) Instruction Tuning: Collect common instructions and their correct responses in e-commerce customer service scenarios, such as “list five recommended products and give prices”. (2) Conversation Tuning: Historical conversation data is input into the model as multiple rounds of samples, and the model is trained to give appropriate responses based on the user context.

[0052] At the same time, examples of FAQ questions and product information query scenarios can also be added to the training data to allow the model to learn standard answer formats.

[0053] (3) Model evaluation indicators Accuracy: whether the key business information (price, inventory, etc.) involved in the answer is accurate; Semantic matching: the matching degree between the answer and the user's question, and whether the answer is irrelevant; Manual scoring: Manual evaluation of the generated results, including conversation fluency, politeness, whether it crosses sensitive lines or violates prohibitions, etc. Business success rate: such as completing transactions, solving problems, improving customer service satisfaction, reducing secondary interactions, etc. These can be counted during the online trial run.

[0054] 4. Verification and testing 4.1 Offline testing Use the validation set and test set to evaluate the model offline and measure dialogue generation indicators such as accuracy, recall, semantic matching, and BLEU.

[0055] 4.2 Online A / B Testing Small-scale launch to real customer service scenarios, conduct A / B testing, and compare the differences between the new model and the old customer service system in key indicators such as user satisfaction and problem resolution rate. You can also do a grayscale release to open the new model service to a part of users or a part of functional scenarios first.

[0056] 4.3 Model quality inspection Several cases are extracted from real conversations and professional customer service or business experts are asked to conduct quality inspection and scoring to identify deficiencies and errors in the model and make iterative corrections.

[0057] As described in the above embodiment, after the above intelligent customer service big model is obtained through model training, the first order query intention can be accurately understood based on the intelligent customer service big model, and the order-related information matching the first order query intention can be output to the user to improve the efficiency and effect of order information query, so as to timely respond to users, handle user problems and other business needs. Specifically, see Figure 2bA method for searching order information based on a private domain big model is provided. The method is applied to an order system based on a multimodal big model. The method for searching order information based on a private domain big model is implemented based on supervised learning LLM. The method can be executed by a business server or a terminal device. The business server can be an e-commerce trading platform, a social platform, a financial platform, a government affairs platform, a retrieval system, or other platform that needs to query orders.

[0058] Among them, the terminal device is a device with an e-commerce transaction platform client installed. The device that implements the above-mentioned method of searching for order information based on the private domain big model can be run in the e-commerce transaction platform client and the e-commerce transaction platform server in a plug-in manner, or it can be directly integrated and designed in the e-commerce transaction platform client and the e-commerce transaction platform server. The specific operation method is not limited to this in the embodiments of this application. The subsequent embodiments only take the server end and the e-commerce transaction platform server end applied to the e-commerce transaction platform as examples. The method is applied to an order system based on a private domain big model. The order system currently displays a first interface, and the first interface includes a query icon. The embodiments of this application include: 201. Receive a first input of a target user to the query icon.

[0059] Among them, the target users may be the first type of users of the backend services of the e-commerce trading platform, the second type of users who open stores on the e-commerce trading platform (such as merchants, suppliers, distributors, etc.), and the third type of users on the e-commerce trading platform (such as buyers).

[0060] For the third category of users, regardless of whether they have purchased goods in the store opened by the second category of users, or deleted the historical order after purchasing goods, the third category of users can enter the conversation interface based on the private domain big model on the order page to perform semantic queries. The embodiment of this application mainly takes the semantic query of the second category of users and the third category of users as an example, and the others are similar and will not be described in detail.

[0061] The first input may be in the form of click, touch, voice input, etc., and is not limited to this.

[0062] 202. In response to the first input, display a conversation interface.

[0063] The conversation interface refers to the display interface of the conversation (either real-time or historical) established between the user and the intelligent customer service model in the e-commerce trading platform when the user logs into the client of the e-commerce trading platform. Both the target user and the intelligent customer service model can send and receive messages of any mode in the target conversation interface.

[0064] The conversation interface is implemented based on a private domain big model, that is, after the intelligent customer service big model in the embodiment of the present application receives the query of the target user, the search scope of the intelligent customer service big model before outputting the answer is based on the private domain. For example, based on the e-commerce transaction platform on which it is deployed, the search scope of the intelligent customer service big model is the e-commerce transaction platform, and it will not directly search the order system of other e-commerce transaction platforms. For the first type of users, the search scope is usually all order transaction data that occurs in the e-commerce transaction platform; for the second type of users, the search scope is usually all order transaction data that occurs in the store associated with the second type of users in the e-commerce transaction platform; for the third type of users, the search scope is usually all order transaction data that occurs in the third type of users in the e-commerce transaction platform.

[0065] 203. Receive a second input from the target user in the conversation interface, where the second input indicates order query information, and the order query information is data in at least one modality.

[0066] Among them, the second input indicates order query information, and the order query information is data of at least one modality, and the second input includes at least two messages of at least one modality. It can be understood that the second input may include multimodal media data such as text, pictures, voice, video, and expressions (dynamic or static). The intelligent customer service big model in the embodiment of the present application is based on multimodal data training, so it can understand data of any modality in the conversation, and can also send data of any modality that matches the first order query intention or solves the user's demands in the conversation interface. The embodiment of the present application takes text conversation as an example and does not make any limitation to this.

[0067] In some implementations, for the above-mentioned first input, the first input includes many messages sent by users. If the intelligent customer service model is to quickly give an answer that matches the first order query intention within a certain period of time, it is necessary to first infer the contextual relationship of each message in the first input.

[0068] The contextual relationship of each message can be realized based on in-context learning. The quality inspection model is a large language model (LLM). When training, it can be based on pre-acquired training data (including various task descriptions-corresponding answer pair structure data). The attention layers in the LLM can realize contextual reasoning. For example, the target user is asking about how to use the function of the e-commerce trading platform's delivery assistant. The intelligent customer service model quickly understands the target user's first order query intention and gives the corresponding operation strategy to ensure smooth communication throughout the conversation.

[0069] 204. In response to the second input, extract target key features of the order query information, and determine the first order query intention of the target user according to the target key features.

[0070] The target key feature may include descriptive information extracted from any modal data associated with any resource currently consulted by the user, and may also include emotional features perceived based on these modal data. The target key feature including emotional features is more helpful for the intelligent customer service model to clarify the first order query intention.

[0071] The target key features refer to the content features that can describe the various modal data in the first input. For example, if the user sends a combination of at least one modality such as text, image, audio, and video in the conversation interface, the intelligent customer service model can select the corresponding feature extraction method based on the modality of the data in the first input to obtain the content features that can describe the various modal data in the first input from the first input.

[0072] The first order query intention may include a query on all order statuses. The first order query intention may include at least one query condition for searching order information. The query condition may be time, product-related features, order status, and order-related operations that the user wants to (to) perform. It is understandable that the query condition may generally be equivalent to the aforementioned target key features. The query condition may be obtained directly or after modification based on the target key features, and is not limited to this.

[0073] In some implementations, in order to further clarify the user's first order query intention in the order system, content enhancement is performed on the target key features based on a generative model to obtain a first search field set.

[0074] The first search field set includes a plurality of search fields that are semantically associated with the target key features.

[0075] In some implementations, the generative model may be a query generation model. In an e-commerce trading platform, when faced with an initial query input by a user, a query generation model may be used to enhance the input based on the first order query intent identified by the initial query, thereby obtaining multiple candidate queries (which may include the initial query), and then performing a deep search based on answers matching the multiple candidate queries, and finally integrating the answers to all candidate queries, thereby improving the accuracy of identifying the first order query intent.

[0076] For example, if Figure 2b As shown, the initial query entered by the user is: Check the orders in the past 7 days? The initial query is enhanced based on the query generation model to obtain the following candidate queries: Check the sales orders in the past 7 days? Check the after-sales orders in the past 7 days? Check the unshipped orders in the past 7 days? Check the abnormal logistics orders in the past 7 days? … It can be seen that multi-intent queries based on the above-mentioned multiple candidate queries can further determine the first order query intention, especially when the initial query input by the user is not very comprehensive. On the one hand, it can intelligently help users expand more refined and comprehensive candidate queries. On the other hand, it can reduce the need for users to change the content of the input query multiple times to express their demands and the intelligent customer service model to guide inquiries multiple times to determine specific inquiries. Correspondingly, it significantly reduces the time for users to determine order query information and output order information, and the process is convenient, thereby improving user experience.

[0077] It is understandable that in some implementations, the query generation model in the embodiments of the present application can be trained adversarially through a generative adversarial network (GANs) to make the generated candidate queries more natural and diverse, such as Figure 2c A framework design is shown. The query generation model can be trained in conjunction with the order query task of the order system in the e-commerce trading platform, so that the query generation model can simultaneously understand and query the order query related information, so as to correctly understand the user's first order query intention and output a credible and controllable answer to the user.

[0078] Optionally, in some embodiments of the present application, such as Figure 2c As shown, the customer service system also includes an adversarial generation model and a query generation model, and the quality of candidate query generation generated by the query generation model is improved by performing adversarial training on the query generation model. Specifically, the method also includes: S1. Obtaining a target prompt word input into the adversarial generation model, where the target prompt word indicates order query information; S2. generating a target question set of an e-commerce trading platform based on the target prompt word, wherein the target question set includes a plurality of adversarial order query questions; S3. Input the target question set into the query generation model to perform adversarial training on the query generation model.

[0079] For example, S3 includes the following steps: S31: Acquire multi-dimensional order query data in the field of e-commerce transactions, and order query samples generated by the generator based on the input of the order query task.

[0080] Among them, order query data includes order data of various data types. For example, for merchants, order query data includes transaction data, product data, inventory data, marketing data, logistics data, after-sales data and financial data of each product in the target store. This order query data is a real sample of e-commerce transaction-related data. For buyers, order query data includes product data, logistics data, after-sales data and financial data purchased in different stores.

[0081] S32: Train the discriminator based on the order query data and query samples obtained in step 1, so that the discriminator distinguishes between real samples and query samples, that is, maximizes the edit distance between the real business data and the query samples based on the first loss function.

[0082] S33: Minimize the edit distance between the real business data and the query sample based on the second loss function, so that the query sample generated by the generator is misclassified as a real sample.

[0083] It can be seen that by performing adversarial training on the query generation model through the generative adversarial model, the effectiveness and accuracy of the candidate query in the e-commerce transaction order inquiry scenario can be improved while ensuring that the generated candidate query is more natural and diversified. Since the intelligent customer service big model is based on multimodality, it can understand any modal data input by the user. Therefore, on the basis of preliminarily identifying the first order query intention based on the target key features in the first input, the generative model is used to enhance the content based on the first order query intention, and multiple search fields are obtained. Since these search fields are all semantically associated with the target key features, on the one hand, multiple search fields can further determine the first order query intention, so that the target answer set obtained based on the multiple search fields can better match the first order query intention. On the other hand, when the intelligent customer service big model and the user have a real-time conversation, the user does not need to change the content of the input query multiple times to express his demands, which significantly reduces the user's inquiry + answer time, and significantly improves the quality inspection efficiency and the reception quality of the intelligent customer service big model, saving labor costs.

[0084] 205. Determine a target query object of the target user according to the first order query intention.

[0085] In the order query scenario, the target query object can be the order system (query order), user system module (query any store associated with the target user), etc.

[0086] 206. According to the first order query intent and the target query object, traverse the orders associated with the target store in the order system to obtain first order information matching the first order query intent.

[0087] The target store is at least one store associated with the target user.

[0088] 207. Display the first order information on the conversation interface.

[0089] It can be seen that, on the one hand, the target user uses semantic inquiry to realize the intelligent customer service big model to quickly locate the first order query intention of the target user, thereby significantly improving the query efficiency; on the other hand, because the order system of the e-commerce transaction platform is directly traversed based on the first order query intention, and the order information in the order system is all real transaction data recorded based on the consensus mechanism of the blockchain, the first order information output by the intelligent customer service big model is the real order information that matches the first order query intention, thereby reducing or avoiding the invalid orders output by the intelligent customer service big model due to attacks or the need to improve robustness. Usually, the invalid order is fabricated by the intelligent customer service big model itself, and the fabricated invalid order may be realistic, and it is impossible to distinguish the authenticity without careful observation, which will confuse the real order data and bring wrong reference data to the real-time query of the target user. In this way, the authenticity of the candidate answers presented to the target user in the conversation interface is guaranteed, and the query efficiency and user experience are also improved.

[0090] Optionally, in some embodiments of the present application, the user can query all order status of a certain product and the order status of a single distributor (identifying risk, supply level) in the conversation interface. Exceptions include logistics, after-sales, wrong goods, etc. Taking the after-sales list as an example, specifically, the first order query is intended to query abnormal orders associated with the target object in the first historical period. The target object is the first product of the target store or a distributor that has a product transaction relationship with the target user. The method also includes: 301. Traverse the abnormal orders associated with the target object in the order system to obtain second order information matching the first order query intention.

[0091] The second order information includes the order status of the first product or the order status of the distributor.

[0092] 302. Verify the authenticity of each order status in the second order information, and filter orders with normal order status.

[0093] In some implementations, the authenticity of each order status may be verified by the following steps: Acquire the abnormal processing data of the first abnormal order from at least one channel of a logistics assistant, an after-sales processing system, and a conversation record between a target store and a product buyer; The authenticity of each order status in the second order information is verified according to the exception processing data.

[0094] For example, a supplier sold a certain model of computer to multiple distributors. Although it was shown as shipped, the logistics showed abnormalities and had not been updated. Then the supplier entered "query a certain model of computer orders with logistics abnormalities in the past 7 days" in the conversation interface of the order system. Before giving the searched order information, the intelligent customer service model can query the historical chat records between the supplier and the distributor with the logistics abnormalities of the model of computer. It is found that the supplier has reissued the model of computer to the distributor in the chat interface yesterday and also sent the express number, so the order can be filtered.

[0095] 303. On the conversation interface, visually output the second order information according to the order status.

[0096] The visual output may be a visual chart, which includes a dynamic switching icon; the first visual chart is obtained according to the after-sales order, after-sales time, after-sales category, after-sales status, after-sales result and after-sales personnel. The after-sales list may be obtained according to the after-sales quantity of resources and after-sales status.

[0097] It is understandable that the target user only needs to input data related to after-sales orders in the conversation interface, such as natural language, pictures, voice, video and other data in any mode, to achieve semantic query. Correspondingly, the intelligent customer service big model can intelligently output the query answer of after-sales data.

[0098] For example, if Figure 3 As shown in the figure, the second type of user (such as a distribution supplier) enters "the after-sales volume and after-sales status of the X2 model CPU in the last month" in the conversation interface. The intelligent customer service model retrieves the after-sales order information of the X2 model CPU sold by the store through various distributors within one month, generates a visual chart after status analysis and statistics, and sends it to the conversation interface for the distribution supplier to click and view at any time.

[0099] It can be seen that the target user can perform semantic queries based on the private domain intelligent customer service big model in the order system, without the need for the user to select filtering conditions to guide the after-sales data, and then download it to the local area for analysis and report making. The intelligent customer service big model can understand the user's statistical needs for after-sales data based on a natural language input by the user, and analyze and count the real order data, and make it into a visual chart output. Obviously, it effectively improves the efficiency of users in analyzing after-sales situations, is convenient, and has a high user experience.

[0100] Optionally, in some embodiments of the present application, the target user (i.e., supplier) can query the sales of some or all of the products of a distributor in the conversation interface at any time. Specifically, the first order query is intended to query the distribution order associated with the first distributor, and the first distributor has a commodity transaction relationship with the target store; the method also includes: The order system is traversed to obtain third order information associated with the first distributor, wherein the third order information includes distribution orders of part or all of the commodities; and the third order information is visually outputted in the conversation interface according to the commodity sales status.

[0101] For example Figure 4 As shown in the figure, the second type of user (such as a distribution supplier) enters "the sales situation of the X2 model CPU in the last month" in the conversation interface. The intelligent customer service model retrieves the sales order information of the X2 model CPU sold by the store through various distributors within one month, generates a visual chart after status analysis and statistics, and sends it to the conversation interface for the distribution supplier to click and view at any time.

[0102] It can be seen that the target user can perform semantic queries based on the private domain intelligent customer service big model in the order system. The intelligent customer service big model can understand the user's statistical needs for the sales data of the distributed products based on the natural language input by the user, and analyze and count the real distribution order data, and make a visual chart output. Obviously, it effectively improves the efficiency of users in analyzing after-sales situations, is convenient, and has a high user experience.

[0103] Optionally, in some embodiments of the present application, after confirming the user's first order query intention, considering that the first order query intention may be an intention that requires at least two or more functional operations to be realized, in order to more conveniently respond to the user's first order query intention, a combined operation can be dynamically and real-time generated to realize the user's first order query intention. Specifically, the method further includes: The first order query intention is analyzed, and if it is determined that the first order query intention matches at least two preset process nodes, an execution logic is generated based on the at least two preset process nodes, and the execution logic includes calling corresponding functional components in the conversation interface according to the execution order of each preset process node. An operation instruction of the target user is received to run the execution logic.

[0104] The execution logic can be dynamically generated based on the user's first order query intention rather than fixed rules, so that target users can get diverse answers.

[0105] It is understandable that in the embodiment of the present application, in the scenario where the user's first order query intention is realized through the above-mentioned combination operation in the conversation interface, in order to ensure the adaptation to the first order query intention with variable combination operations, the order query business logic can also be divided into multiple functional components in advance according to the query business function type. After identifying the user's first order query intention, different functional components can be called to combine to form the execution logic. Specifically, the file definition can be pre-configured to implement the order query process in the conversation interface to dynamically load at least two corresponding functional components, and each functional component can be reused.

[0106] In some implementations, for users who want to customize the deployment of the intelligent customer service large model in this application, the loading of the functional components required for the above-mentioned combined operations can be implemented based on containers or low-code. When implemented with low code, users can quickly build applications by dragging and dropping corresponding functional components, setting parameters, and configuring processes. This approach reduces the need for traditional coding and allows business personnel and non-professional developers to participate in application development. Users can also customize their own requirements for order queries based on changes in actual order query needs or business needs.

[0107] Among them, drag components: in the platform's visual interface, drag the required interface components (such as forms, buttons, tables, etc.) to the design area of ​​the conversation interface.

[0108] Configure business logic: Use the logic configuration tools provided by the platform (such as workflow, rule engine, etc.) to set the interaction logic and business rules between components.

[0109] Connect data sources: Configure the data sources required by the application, including databases, APIs, etc., to achieve data storage and interaction.

[0110] It can be seen that in the embodiment of the present application, after identifying the user's first order query intention, it is split into at least two process nodes. Since the above-mentioned execution logic is dynamically generated based on at least two process nodes that match the user's first order query intention, rather than the fixed rules of the conventional workflow, the target user only needs to make semantic queries to obtain diverse answers, and can also perform corresponding related operations more conveniently.

[0111] In some implementations, an example is taken of helping users to simultaneously complete order search and print confirmation in response to a semantic inquiry by the user. In this application scenario, the target store of the target user (i.e., the supplier) will generate many commodity transaction orders within a certain period of time, and the target user needs to perform logistics and delivery processing for each commodity transaction order separately. Therefore, in order to reduce the cumbersome operations of the target user, the embodiment of the present application also connects to a delivery assistant in the intelligent customer service model to help the target user complete quick searches for delivery orders, as well as confirm printing operations of delivery orders and other matters. Specifically, the method also includes: 401. Receive a second input from the target user in the conversation interface, where the second input indicates printing of the abnormal delivery order.

[0112] It is understandable that since the order query intent corresponding to the second input is to instruct to print the abnormal shipment order, after analyzing it, it is determined that to realize the order query intent, two process nodes are required: abnormal shipment order search and abnormal shipment order printing. Then, an execution logic can be dynamically generated based on these two process nodes, including executing the abnormal shipment order search first and then executing the abnormal shipment order printing.

[0113] 402. In response to the second input, determine and visualize the abnormal shipping orders from the order system, and display a print icon.

[0114] 403: Receive a print instruction from the target user for the print icon, and execute the print instruction.

[0115] For example, if Figure 5 As shown in fig1, the supplier opens the conversation interface based on the private domain big model in the order system interface, and sends a message "Please help me print all the orders that have not been shipped in the past few days". After receiving the supplier's instruction, the intelligent customer service big model understands that the supplier's intention is to print the unshipped orders. To complete the instruction, the intelligent customer service big model splits the instruction into two operations to execute in order to achieve a combined operation. That is, the following are completed in sequence: Operation 1: Help the supplier search for unshipped orders; Operation 2: Call the shipping assistant and other software to provide the supplier with a "print" operation entrance, which can be a "print icon for a single order, a print icon for all unshipped orders" for the target user to choose freely. The printed orders automatically become an inoperable area in the conversation interface or are displayed as printed, for example Figure 5 As shown in Fig2, for example, the printed state changes to gray.

[0116] It can be seen that after understanding the user's semantic inquiry, the user's semantic input is decomposed into instructions, and then the semantic inquiry of the target user is completed in one go by combining at least two operations. The entire process is completed intelligently, which not only simplifies the process operation, but also saves the user a lot of operation time, frees up their hands, and improves the user experience.

[0117] Optionally, in some embodiments of the present application, considering that the target user (i.e., supplier or merchant) is very busy, or accustomed to simple and fast communication, the target user may have queried the intelligent customer service model for relevant order information in the conversation interface at a certain time period, and is currently querying the same relevant order information. In this case, if the target user asks for relevant order information at another time immediately after the previous historical conversation record, the intelligent customer service model can understand the context and locate that the target user's first order query intention at this time is actually based on the previous order query. It is highly likely that only the time interval of the order query has changed, so it can be combined with the historical query records, and the current simple time parameter input can complete the same type of query output.

[0118] Specifically, if Figure 6a As shown, the embodiment of the present application also includes: 501. Receive a target input from the target user in the conversation interface.

[0119] 502. In response to the target input, obtain keywords in the target input.

[0120] 503. If it is determined that the keyword matches the target query condition in the first order query intent, the target query condition in the first order query intent is updated to the keyword to obtain a second order query intent corresponding to the target input.

[0121] The target query condition is part or all of the query conditions in the first order query intent, and this application takes part of the query conditions as an example. The target query condition is a structured language used for searching. The matching of keywords with the target query condition means that the attributes of the keywords are the same or similar to the target query condition.

[0122] 504. Referring to the first order query intention, the first order information is updated according to the keyword, and the fourth order information is obtained and output visually.

[0123] It can be seen that after receiving the target input from the target user, the intelligent customer service model queries the last order query record. Based on the third input this time, it can be concluded through judgment that compared with the first order query intention last time, the keywords of this query are the same type or related semantics as one or several target query conditions for searching order information in the first order query intention. Therefore, it can be inferred that the user's second order query intention is consistent with the first order query intention, and only some query conditions I and target query conditions are changed. After updating the target query conditions in the first order query intention, the corresponding order query information can be directly output. It can be seen that by comparing the order query intention and updating some query conditions of the historical query, the target user can be quickly responded to.

[0124] Taking the time query condition update in the order query intent as an example, the other query condition updates are similar and will not be described in detail. The method also includes: S1. Receive a third input from the target user in the conversation interface.

[0125] The third input indicates querying the orders to be shipped in the third historical period.

[0126] S2. In response to the third input, information on orders to be shipped within a third historical period is obtained from the order system and displayed in the conversation interface.

[0127] S3. Receive a fourth input from the target user in the conversation interface.

[0128] The fourth input indicates a fourth historical period.

[0129] S4. In response to the fourth input, context understanding is performed based on the historical conversation records of the third historical period, and the order information to be shipped within the fourth historical period is obtained from the order system and displayed in the conversation interface.

[0130] For example, if Figure 6b The following is a schematic diagram of an interface for semantic query of order-related information. Figure 6b In Fig. 1, the target user enters a query in the independent dialog window of the order system interface: Check how many green-labeled orders I have waiting to be shipped in the last 7 days. It can be seen that this query is a semantic query, and after semantic understanding, it is clear that the user's order query intention includes three conditions: waiting to be shipped, in the last 7 days, and green label.

[0131] It can be seen that the semantic inquiry function can replace the traditional operation that requires manual selection of multiple conditions such as pending shipment, within the past 7 days, and green label.

[0132] Correspondingly, the intelligent customer service model visualizes the "orders with green labels to be shipped within 7 days" found in the dialogue window. Figure 6b Shown as a flag with no color indicated.

[0133] Next, when the target user enters "yesterday's" during the next conversation, the intelligent customer service model will retain the other conditions from the previous round of inquiries (i.e., pending shipment, green label), and automatically replace the time condition "nearly 7 days" with "yesterday", making it more flexible and convenient to use. Figure 6b As shown in fig2.

[0134] It can be seen that by implanting the semantic query function of the intelligent customer service big model in the order system, it can replace the traditional inefficient method of manually selecting various conditions to conduct effective order information query. This solution has high query efficiency and is flexible to use. Especially when it is inconvenient for users to use the order query function based on screening conditions in the traditional order system, users can directly type or send a voice.

[0135] Optionally, in some embodiments of the present application, in order to increase the commodity transaction volume of the target store, the entire platform can also be searched for popular commodities, and intelligent distribution of commodities can be carried out according to the hot commodities. Figure 7a As shown, the method also includes: 601. Receive a fifth input from the target user in the conversation interface, where the fifth input indicates to distribute hot commodities; 602. In response to the fifth input, obtain sales distribution data of hot commodities in the first historical period from the order system; 603. Determine the target hot commodity and the platform to be distributed according to the sales distribution data, and set a first distribution strategy for the target hot commodity on the platform to be distributed within a preset time period; 604. Displaying the first distribution strategy in the conversation interface; 605. Receive a sixth input from the target user on an execution operation icon of the first distribution strategy in the conversation interface; 606. In response to the sixth input, execute the first distribution strategy to distribute the target hot commodity on the platform to be distributed.

[0136] For example Figure 7bAs shown in the figure, the supplier enters in the conversation interface: Help me to distribute the top 10 best-selling New Year gift packages. After searching the order system of the e-commerce trading platform, the intelligent customer service model analyzes the order volume of the New Year gift packages in the order system, obtains the heat analysis results, and then compares the products of the supplier's store. Finally, it is determined that 8 of the top 10 best-selling New Year gift packages on this e-commerce trading platform are from the supplier's store. Then, the intelligent customer service model gives a one-click distribution answer, including the distribution channels and the number of products to be distributed for each hot product. The supplier can also edit it in the interface and then click one-click distribution.

[0137] It can be seen that by searching for popular products on the entire platform and intelligently distributing them according to the hot products, the target store's transaction volume can be increased. The operation is convenient and the distribution is reasonable. Users only need to make semantic inquiries, which provides a high user experience and high distribution efficiency.

[0138] Optionally, in some embodiments of the present application, before outputting the query answer of the product to be distributed, the transaction situation is confirmed first and then the product is distributed reasonably. Specifically, the first input is to query the transaction data of the target user on the distributed products on different e-commerce trading platforms during the second historical period; the method also includes: The order system is traversed to obtain first transaction data of the already distributed goods in the target store on different e-commerce trading platforms; first description information of the goods to be distributed is determined according to the first transaction data, wherein the first description information includes information of the goods to be distributed and the e-commerce trading platform to be distributed; a second distribution strategy is determined according to the first description information; and the second distribution strategy is visualized in the conversation interface, wherein the second visualization chart includes a smart distribution application icon and information of the goods to be distributed on different e-commerce trading platforms.

[0139] It can be seen that there is no need for users to first pull the transaction data of goods that have been distributed on different e-commerce trading platforms, nor is there a need to manually select screening conditions to export, and then download to the local area for analysis and report making. With this solution, the intelligent customer service model can understand the user's statistical needs for distribution data based on a natural language input by the user, and analyze and count the real order data and goods to be distributed, and make them into visual charts for output. Obviously, it effectively improves the order processing process, selects the appropriate e-commerce trading platform for classified distribution, and solves problems such as inventory synchronization. The overall distribution efficiency is high, and informatization and intelligence improve the rationality of distribution, which is convenient and has a high user experience. Intelligent distribution is realized based on the user's semantic inquiry, and the distribution volume and distribution channels are more reasonable, saving labor time, improving operational results and reducing operating costs.

[0140] Optionally, in some embodiments of the present application, a store's financial statements can be visually outputted through semantic queries in the order system. Specifically, the first input is to query the financial data of the target store's transaction orders on different e-commerce trading platforms during the third historical period; the method also includes: Acquire second transaction data of commodity transactions of target users on different e-commerce transaction platforms; determine first financial data of historical transaction commodities based on the second transaction data; and visualize the first financial data in the conversation interface.

[0141] For example, if Figure 8 As shown in the figure, the second type of user (for example, a store seller role) enters "transaction financial data of the store in the last three months" in the conversation interface. The intelligent customer service model retrieves order-related information of all products that have transactions, after-sales, etc. in all stores opened by the store seller on this e-commerce trading platform within three months, generates a visual chart after status analysis and statistics, and sends it to the conversation interface for the user to click and view at any time.

[0142] It can be seen that users do not need to select filtering conditions to export their own financial data on the e-commerce trading platform, and then download it to the local area for analysis and report making. With this solution, the intelligent customer service model can understand the user's statistical needs for financial data based on a natural language input by the user, and analyze and count the actual order data, and make it into a visual chart output. Obviously, it effectively improves the efficiency of users analyzing income and expenditure, is convenient, and has a high user experience.

[0143] Optionally, in some embodiments of the present application, semantic queries can be made in the order system to reasonably schedule the goods to be distributed according to the financial data of the store. Figure 9a As shown, the method also includes: 701. Receive a seventh input from the target user in the conversation interface.

[0144] The seventh input indicates that the goods are distributed according to second financial data, and the second financial data includes sales distribution data of the goods to be distributed on different trading platforms during a fourth historical period.

[0145] 702. In response to the seventh input, generate a third distribution strategy according to the financial data and the products to be distributed; 703. Displaying a third distribution strategy in the conversation interface; 704. Receive an eighth input from the target user on an execution operation icon of the third distribution strategy in the conversation interface; 705. In response to the eighth input, execute the third distribution strategy to distribute the product to be distributed on the platform to be distributed.

[0146] It is understandable that the second financial data may be the first financial data. Figure 8 On the basis of the corresponding embodiment, after obtaining the visual output of the first financial data through semantic inquiry, the target user can also be guided whether to schedule the goods to be distributed according to the financial data. Of course, this distribution scenario is also applicable to the method of realizing the first order query intention based on the combination operation in the execution logic in the above embodiment, and similarly, it will not be repeated.

[0147] Figure 9a The corresponding embodiments can be implemented by semantic query alone or in combination with Figure 8 The corresponding embodiments are further implemented without limitation thereto.

[0148] For example, if Figure 9b As shown in the figure, the supplier makes a semantic inquiry about the distributed product XX to the intelligent customer service big model in the conversation interface, and the intelligent customer service big model outputs the visualized financial data of the product XX to the supplier. Then, the corresponding distribution strategy is given based on the financial data of the product XX. The supplier can directly click "one-click distribution" in the conversation interface to complete the intelligent distribution.

[0149] It can be seen that users do not need to first pull the transaction data of goods that have been distributed on different e-commerce trading platforms, nor do they need to manually select screening conditions to export, and then download to the local area for analysis and report making. With this solution, the intelligent customer service model can understand the user's statistical needs for distribution data based on a natural language input by the user, and analyze and count the actual order data and goods to be distributed, and make them into visual charts for output. Obviously, it effectively improves the order processing process, selects the appropriate e-commerce trading platform for classified distribution, and solves problems such as inventory synchronization. The overall distribution efficiency is high, and informatization and intelligence improve the rationality of distribution, convenience, and high user experience.

[0150] The order information processing device in the embodiment of the present application has the function of implementing the method for searching order information based on the private domain big model and the training method of the intelligent customer service big model provided in the above embodiment. The function can be implemented by hardware, or by hardware executing the corresponding software. The hardware or software includes one or more modules corresponding to the above functions, and the modules can be software and / or hardware.

[0151] See also Fig.10 , Fig.10This is a schematic diagram of the structure of an order information processing device 50 provided in an embodiment of the present application. The order information processing device 50 can be applied to a computing device in a scenario where a conversation with a user needs to be implemented based on a large model, for example, it can be applied to respond to a user's query operation in an order system. Specifically, it can be understood that the order information processing device 50 can be a separately deployed intelligent customer service large model, or it can be a computing device or computer program related to order processing that deploys an intelligent customer service large model, and the present application does not limit this. The order information processing device 50 can deploy at least one adversarial generation model (optional, not shown in the figure), an input and output module 501, a display module 502, and a processing module 503. Among them, the processing module 503 can be used to Figure 1a-Figure 9b In any embodiment, operations such as training the intelligent customer service big model, determining the target questions and emotional data of the target user, and controlling the operations of the adversarial generation model, the input and output module 501, and the display module 502 are similar to other big models. The adversarial generation model, the display module 502, and the processing module 503 can be deployed separately in a computing device or centrally in a computing device, which is not limited in this application.

[0152] In some implementations, the input-output module 501 is used to receive a first input of the query icon by a target user; The processing module 503 is configured to display a conversation interface through the display module 502 in response to the first input received by the input / output module 501, wherein the conversation interface is implemented based on the private domain big model; The input-output module 501 is further configured to receive a second input from the target user in the conversation interface, where the second input indicates order query information, and the order query information is data of at least one modality; The processing module 503 is also used to extract the target key features of the order query information in response to the second input received by the input-output module 501, determine the first order query intention of the target user according to the target key features; determine the target query object of the target user according to the first order query intention; determine the target store associated with the target user; traverse the orders associated with the target store in the order system according to the first order query intention and the target query object to obtain the first order information matching the first order query intention; The display module 502 is further configured to display the first order information obtained by the processing module 503 on the conversation interface.

[0153] Other functions of the adversarial generation model, the input and output module 501, the display module 502 and the processing module 503 can be referred to Figure 1a-Figure 9bThe corresponding embodiments are not described in detail.

[0154] Compared with the prior art, in the embodiment of the present application, since the order information processing device 50 is based on multi-private domain query orders and can process multi-modal data, it can understand any modal data input by the user. Therefore, on the basis of preliminarily identifying the first order query intention based on the target key features in the first input, on the one hand, the target user realizes the intelligent customer service big model to quickly locate the first order query intention of the target user through semantic inquiry. Since the order system of the e-commerce transaction platform is traversed directly based on the first order query intention, and the order information in the order system is all real transaction data recorded based on the consensus mechanism of the blockchain, therefore, the first order information output by the intelligent customer service big model is the real order information that matches the first order query intention, thereby reducing or avoiding the invalid orders output by the intelligent customer service big model due to attacks or the need to improve robustness. Usually, the invalid order is fabricated by the intelligent customer service big model itself, and the fabricated invalid order may be realistic, and it is impossible to distinguish the authenticity without careful observation. This will confuse the real order data and bring erroneous reference data to the real-time query of the target user. In this way, the authenticity of the candidate answers presented to the target user on the conversation interface is guaranteed, and query efficiency and user experience are improved.

[0155] The order information processing device in the embodiment of the present application is described above from the perspective of modular functional entities. The order information processing device in the embodiment of the present application is described below from the perspective of hardware processing. Fig.10 The order information processing device 50 shown in the figure may have the following features: Fig.11 The structure shown, when Fig.10 The order information processing device 50 shown has the following Fig.11 When the structure shown is Fig.11 The processor and transceiver in the order information processing device can realize the same or similar functions as the adversarial generation model, input and output module 501, display module 502 and processing module 503 provided in the device embodiment corresponding to the order information processing device. Fig.11 The memory storage processor in the computer program that needs to be called when executing the above-mentioned method for searching order information based on the private domain big model and the training method of the intelligent customer service big model.

[0156] In the above embodiments, the description of each embodiment has its own emphasis. For parts not described in detail in a certain embodiment, reference can be made to the relevant description of other embodiments. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described systems, devices and modules can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here.

[0157] In several embodiments provided in the embodiments of the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms. The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0158] In addition, each functional module in each embodiment of the present application can be integrated into a processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The above-mentioned integrated module can be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0159] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program is loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk SolidState Disk (SSD)), etc.

[0160] The technical solutions provided in the embodiments of the present application are introduced in detail above. The principles and implementation methods of the embodiments of the present application are explained by using specific examples in the embodiments of the present application. The description of the above embodiments is only used to help understand the methods and core ideas of the embodiments of the present application. At the same time, for those skilled in the art, according to the ideas of the embodiments of the present application, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as a limitation on the embodiments of the present application.

Claims

1. A method for searching order information based on a private domain big model, characterized in that: The method is applied to an order system based on a private domain big model, the order system currently displays a first interface, and the first interface includes a query icon; the method includes: Receiving a first input of the query icon by a target user; In response to the first input, displaying a conversation interface, wherein the conversation interface is implemented based on a private domain big model; receiving a second input from the target user in the conversation interface, where the second input indicates order query information, and the order query information is data in at least one mode; In response to the second input, extracting a target key feature of the order query information, and determining a first order query intention of the target user according to the target key feature, wherein the first order query intention may include at least one query condition for searching order information; Determining a target query object of the target user according to the first order query intention; According to the first order query intent and the target query object, traverse the orders associated with the target store in the order system to obtain first order information matching the first order query intent, where the target store is at least one store associated with the target user; The first order information is displayed on the session interface.

2. The method according to claim 1, characterized in that The first order query is intended to query abnormal orders associated with a target object within a first period of time, where the target object is the first product of a target store or a distributor having a product transaction relationship with the target user; The method further comprises: Traversing the abnormal orders associated with the target object in the order system to obtain second order information matching the first order query intention; the second order information includes the order status of the first product or the order status of the distributor; Verifying the authenticity of each order status in the second order information, and filtering orders with normal order status; In the conversation interface, the second order information is visually output according to the order status.

3. The method according to claim 2, characterized in that The first order query is intended to query the distribution order associated with the first distributor, and the first distributor has a commodity transaction relationship with the target store; the method further includes: Traversing the order system to obtain third order information associated with the first distributor, wherein the third order information includes distribution orders of part or all of the commodities; In the conversation interface, the third order information is visually output according to the sales status of the goods.

4. The method according to claim 3, characterized in that The method further comprises: Analyzing the first order query intention, if it is determined that the first order query intention matches at least two preset process nodes, generating execution logic based on the at least two preset process nodes, the execution logic including calling corresponding functional components in the conversation interface according to the execution order of each preset process node; Receive an operation instruction from a target user to run the execution logic.

5. The method according to claim 4, characterized in that The second input indicates printing of an abnormal delivery order; the analyzing the first order query intention, if it is determined that the first order query intention matches at least two preset process nodes, generating execution logic based on the at least two preset process nodes, and receiving an operation instruction from a target user to run the execution logic, includes: In response to the second input, determining an abnormal delivery order from the order system and visually displaying the abnormal delivery order, and displaying a print icon; A print instruction for the print icon from the target user is received, and the print instruction is executed.

6. The method according to claim 5, characterized in that The method further comprises: Receiving a target input from the target user on the conversation interface; In response to the target input, obtaining keywords in the target input; If it is determined that the keyword matches the target query condition in the first order query intent, the target query condition in the first order query intent is updated to the keyword to obtain a second order query intent corresponding to the target input; Referring to the first order query intention, the first order information is updated according to the keywords, and the fourth order information is obtained and output visually.

7. The method according to claim 6, characterized in that The method further comprises: receiving a fifth input from the target user in the conversation interface, wherein the fifth input indicates to distribute hot commodities; In response to the fifth input, obtaining sales distribution data of hot commodities in the first historical period from the order system; Determine the target hot commodity and the platform to be distributed according to the sales distribution data, and set the first distribution strategy of the target hot commodity on the platform to be distributed within a preset time period; Displaying a first distribution strategy in the conversation interface; receiving a sixth input from the target user on an execution operation icon of the first distribution strategy in the conversation interface; In response to the sixth input, the first distribution strategy is executed to distribute the target hot commodity on the platform to be distributed.

8. The method according to claim 7, characterized in that The method further comprises: receiving a seventh input from the target user in the conversation interface, the seventh input instructing to distribute the product according to second financial data, the second financial data including sales distribution data of the product to be distributed on different trading platforms in a fourth historical period; In response to the seventh input, generating a second distribution strategy according to the financial data and the product to be distributed; Displaying a third distribution strategy in the conversation interface; receiving an eighth input by the target user to an execution operation icon of the third distribution strategy in the conversation interface; In response to the eighth input, the third distribution strategy is executed to distribute the product to be distributed on the platform to be distributed.

9. An order information processing device, characterized in that: The order information processing device comprises: at least one processor and memory; The memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the method according to any one of claims 1 to 8.

10. A computer program product comprising instructions, characterized in that The computer program product comprises program instructions, and when the program instructions are executed on a computer or a processor, the computer or the processor is caused to execute the method according to any one of claims 1 to 8.

Images