Logistics question answering method and equipment based on large language model, and medium

Abstract

The invention discloses a logistics question answering method and device based on a large language model and a medium, and the method comprises the steps: obtaining input information corresponding to a target logistics question, carrying out the natural language processing of the input information, and obtaining the analysis data of the target logistics question; inputting the analysis data into a pre-trained logistics language model, wherein the logistics language model is a large language model which is finely adjusted by using a logistics industry corpus; retrieving the analysis data through a logistics knowledge graph associated with the logistics language model to obtain logistics knowledge corresponding to the analysis data; and returning the logistics knowledge to the logistics language model, and generating a question answer of the target logistics question. By constructing the logistics language model, professional answers for the logistics industry can be quickly understood and generated, and the speed and accuracy of information query are greatly improved. The knowledge graph special for the logistics industry is constructed, and more complex and diversified business problems can be processed in combination with a deep learning algorithm.

Description

Technical Field

[0001] The present application relates to the field of logistics management, and specifically to a logistics question answering method, device and medium based on a large language model. Background Art

[0002] Traditional logistics information systems mainly focus on the storage and simple query of structured data. For example, a warehouse management system built on a relational database can accurately record the quantity and location of goods entering and leaving the warehouse, but its ability to understand questions in natural language is extremely limited, and it is unable to accurately grasp the true intention and semantic connotation of the question, especially questions involving complex logical relationships, implicit information or professional terms. If you ask questions such as "Which mode of transportation is most suitable for transporting fragile goods and has a lower cost?" in natural language, it is difficult for traditional systems to give an effective response.

[0003] In addition, traditional logistics information systems have difficulty integrating logistics knowledge from different sources and in different formats, such as internal enterprise operation data, industry standard documents, external news information, etc. Each data source is often independent of each other and lacks an effective fusion mechanism, resulting in one-sided information when answering questions. Summary of the invention

[0004] In order to solve the above problems, this application proposes a logistics question answering method, device and medium based on a large language model, wherein the method includes:

[0005] Acquire input information corresponding to a target logistics problem, perform natural language processing on the input information, and obtain parsed data for the target logistics problem; input the parsed data into a pre-trained logistics language model, where the logistics language model is a large language model fine-tuned using logistics industry corpus; retrieve the parsed data through a logistics knowledge graph associated with the logistics language model to obtain logistics knowledge corresponding to the parsed data; return the logistics knowledge to the logistics language model to generate an answer to the target logistics problem.

[0006] In one example, the method of obtaining input information corresponding to the target logistics problem specifically includes: determining an access port of the user, the access port including at least one of a web page, a mobile terminal, and a voice assistant; based on the access port, providing a user interaction interface to the user, and obtaining the user's initial input; if the initial input information is of a voice type, performing voice recognition on the initial input information to obtain text-type input information.

[0007] In one example, before retrieving the parsed data through the logistics knowledge graph associated with the logistics language model, the method further includes: extracting entities and relationships from an existing logistics industry corpus, the logistics industry corpus including at least one of a logistics enterprise database, an industry standard document, and logistics news information; based on the entities and relationships, generating multiple tuples to constitute the logistics knowledge graph; and / or, retrieving the parsed data through the logistics knowledge graph associated with the logistics language model, specifically including: constructing a query statement based on keywords contained in the parsed data; executing the query statement in the logistics knowledge graph to obtain entities and relationships related to the parsed data.

[0008] In one example, returning the logistics knowledge to the logistics language model to generate an answer to the target logistics question specifically includes: converting the parsed data into a word vector representation containing context information through a bidirectional transformer model in the logistics language model; processing the word vector representation through a BiLSTM layer to capture long-distance dependencies in the parsed data; and globally optimizing the output of the BiLSTM through a conditional random field layer to ensure that the generated label sequence meets grammatical and semantic requirements.

[0009] In one example, the method also includes: recording the user's input information, the input information including question information, query information, answer selection information and feedback information; when the data volume corresponding to the target user's input information is higher than a preset threshold, generating a user portrait of the target user based on the input information; based on the user portrait, and in combination with the user's geographic location information, weather information and holiday information, correcting the answer to the question.

[0010] In one example, the user portrait includes: frequent shippers whose shipping frequency is higher than a first preset threshold, occasional shippers whose shipping frequency is lower than a second preset threshold, non-shippers whose shipping frequency is zero, frequent consignees whose receiving frequency is higher than a third preset threshold, occasional consignees whose receiving frequency is lower than a fourth preset threshold, non-consignees whose receiving frequency is zero, recent shippers whose historical shipping records are within a preset time period, and recent consignees whose historical receiving records are within a preset time period.

[0011] In one example, the method also includes: generating the target user's question expectations based on the user portrait, the user's geographic location information, weather information, and holiday information; determining historical users similar to the target user's question expectations in a historical database; obtaining historical question data corresponding to the historical users; selecting a question template based on the historical question data, and presenting the question template to the target user.

[0012] In one example, after returning the logistics knowledge to the logistics language model to generate an answer to the target logistics question, the method further includes: sending the answer to the question to a user and receiving feedback information from the user, wherein the feedback information includes satisfaction and error correction; updating the logistics knowledge graph and adjusting the logistics language model through the feedback information and a self-learning mechanism.

[0013] The present application also provides a logistics question answering device based on a large language model, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute: obtaining input information corresponding to a target logistics problem, performing natural language processing on the input information, and obtaining parsed data of the target logistics problem; inputting the parsed data into a pre-trained logistics language model, wherein the logistics language model is a large language model fine-tuned using logistics industry corpus; retrieving the parsed data through a logistics knowledge graph associated with the logistics language model to obtain logistics knowledge corresponding to the parsed data; returning the logistics knowledge to the logistics language model to generate an answer to the target logistics problem.

[0014] The present application also provides a non-volatile computer storage medium storing computer executable instructions, characterized in that the computer executable instructions are configured to: obtain input information corresponding to a target logistics problem, perform natural language processing on the input information, and obtain analytical data of the target logistics problem; input the analytical data into a pre-trained logistics language model, wherein the logistics language model is a large language model fine-tuned using logistics industry corpus; retrieve the analytical data through a logistics knowledge graph associated with the logistics language model to obtain logistics knowledge corresponding to the analytical data; return the logistics knowledge to the logistics language model to generate an answer to the target logistics problem.

[0015] The method proposed in this application can bring the following beneficial effects: by building a logistics language model, it can quickly understand and generate professional answers for the logistics industry, greatly improving the speed and accuracy of information query. At the same time, a knowledge graph specifically for the logistics industry is built, and combined with a deep learning algorithm, the system can handle more complex and diverse business problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] 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:

[0017] Figure 1 A flow chart of a logistics question answering method based on a large language model in an embodiment of the present application;

[0018] Figure 2 A flow chart of another method for answering logistics questions based on a large language model in an embodiment of the present application;

[0019] Figure 3 This is a schematic diagram of a data flow based on a model in an embodiment of the present application;

[0020] Figure 4 This is a schematic diagram of a logistics language model in an embodiment of the present application. DETAILED DESCRIPTION

[0021] 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.

[0022] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0023] Figure 1 A flowchart of a logistics question answering method based on a large language model is provided for one or more embodiments of this specification. The process can be executed by a computing device in the corresponding field, and some input parameters or intermediate results in the process can be manually intervened to help improve accuracy.

[0024] The analysis method involved in the embodiments of the present application can be implemented by a terminal device or a server, and the present application does not impose any special restrictions on this. For the convenience of understanding and description, the following embodiments are described in detail by taking a server as an example.

[0025] It should be noted that the server can be a single device or a system composed of multiple devices, that is, a distributed server, and this application does not make any specific limitations on this.

[0026] like Figure 1 and Figure 2 As shown, the embodiment of the present application provides a logistics question answering method based on a large language model, including:

[0027] S101: Obtain input information corresponding to a target logistics problem, perform natural language processing on the input information, and obtain analytical data of the target logistics problem.

[0028] like Figure 3 As shown in the figure, the server provides an open API interface to support integration with other logistics management systems. For example, enterprises can embed the system's question-and-answer function into existing transportation management systems, warehouse management systems, and other systems to achieve seamless integration. After access, users will provide input information corresponding to the target logistics problem through the user interface. After the server performs natural language processing on the input information, it obtains the parsed data of the target logistics problem.

[0029] Traditional technologies are mostly based on simple text matching rules, without in-depth analysis of language structure and semantic logic. As a result, the existing logistics customer service system lacks deep natural language understanding capabilities and cannot accurately grasp the true intention and semantic connotation of the problem, especially those involving complex logical relationships, implicit information or professional terms.

[0030] S102: Input the parsed data into a pre-trained logistics language model, where the logistics language model is a large language model fine-tuned using logistics industry corpus.

[0031] After natural language processing of user questions, the parsed data is input into a pre-trained logistics language model, which is obtained by pre-selecting a high-performance large language model and fine-tuning it using a large amount of logistics industry corpus. These corpora include historical records, operating guidelines, laws and regulations, etc. to ensure that the model can understand and generate professional answers for the logistics industry. The trained model is deployed to the cloud or local server to ensure that it can run stably in the actual environment. Through a continuous feedback mechanism, the performance of the model is continuously optimized to improve the accuracy and speed of the answers.

[0032] S103: Retrieve the parsed data through the logistics knowledge graph associated with the logistics language model to obtain logistics knowledge corresponding to the parsed data.

[0033] This application constructs a knowledge graph specifically for the logistics industry, organically organizes scattered knowledge points to form a complete knowledge system. The large model uses the knowledge graph structured knowledge to enhance semantic understanding and reasoning, retrieves the graph to obtain entity association information when dealing with problems, and optimizes answer generation. It can also reversely update the graph to improve knowledge. For example, if a new transportation route is found, the transportation relationship of the graph can be updated, improving the quality of the system's knowledge reserve and the accuracy and depth of questions and answers, helping users to gain in-depth insights into logistics business logic. After the logistics language model obtains the parsed data, it will retrieve the parsed data through the logistics knowledge graph to obtain the logistics knowledge corresponding to the parsed data.

[0034] S104: Return the logistics knowledge to the logistics language model to generate an answer to the target logistics question.

[0035] The logistics knowledge retrieved from the knowledge graph is returned to the logistics language model, and the assistant logistics language model generates answers to the target logistics questions.

[0036] When generating answers, the server first converts the parsed data into a word vector representation containing contextual information through the bidirectional transformer model in the logistics language model; the bidirectional transformer model is a pre-trained language model that can capture the proper nouns and complex contexts in the logistics field. The word vector representation is then processed through the BiLSTM layer to capture the long-distance dependencies in the parsed data, which is very important for understanding complex logistics scenarios, such as multi-stage transportation, cross-border logistics, etc. The multi-head attention mechanism can help the model better focus on different parts of the sentence, especially those keywords related to logistics entities; this helps to improve the recognition of nested entities (such as "Beijing" and "Shanghai" in "Transportation route from Beijing to Shanghai"). Finally, the conditional random field layer globally optimizes the output of the BiLSTM to ensure that the generated label sequence meets the grammatical and semantic requirements. The conditional random field layer can avoid some common labeling errors, such as inconsistent labels or inaccurate boundaries.

[0037] In one embodiment, when obtaining input information, the server first determines the user's access port, which includes a web page, a mobile terminal, and a voice assistant. Then, based on the access port, different user interaction interfaces are provided to the user, and the user's initial input is obtained. Since natural language processing is required in the subsequent process, the initial input needs to be converted into a text type. Specifically, if the initial input information is a voice type, voice recognition is performed on the initial input information to obtain text-type input information.

[0038] In one embodiment, when constructing a logistics knowledge graph, entities and relationships need to be extracted from the existing logistics industry corpus, which includes logistics enterprise databases, industry standard documents, logistics news information, etc. Then, based on the entities and relationships, multiple tuples are generated to constitute the logistics knowledge graph, such as triples such as "cargo-transportation mode-railway transportation" and "warehouse-storage goods-electronic products".

[0039] like Figure 4 As shown, in one embodiment, the method disclosed in the present application can be implemented through a logistics language model. In this case, the logistics language model includes an input layer, a processing layer, an output layer, and a feedback layer, and implements corresponding functions.

[0040] In one embodiment, the present application can also help the logistics language model understand the user's interests and common query types by recording each question, query content, selected answer, and feedback (such as satisfaction) asked by the user in the logistics question-and-answer system. When the amount of data corresponding to the input information of the target user is higher than the preset threshold, a user portrait of the target user can be generated based on the input information, and then the question answer can be corrected based on the user portrait and combined with the user's geographic location information, weather information, and holiday information. For example, during holidays or inclement weather conditions, users may be more concerned about logistics delays. By analyzing the user's historical query records, the user's preferences are extracted. According to the user's feedback on the system's answers (such as likes, comments, complaints, etc.), the user's satisfaction is evaluated and the system's answer strategy is adjusted. For example, if the user is not satisfied with an answer, the system can provide different explanations or suggestions in subsequent similar questions. Users are divided into different groups using a density peak-based clustering algorithm. By analyzing the user's historical query records, the system can predict the user's intentions and prepare relevant answers in advance. Through user portraits, the system can better understand the user's multi-round dialogue needs. For example, a user may first ask “What is the shipping time of SF Express from Beijing to Shanghai?” and then follow up with “Will the shipping time be extended if there is bad weather?” The system can provide a more coherent and detailed answer based on the user’s historical queries and current questions.

[0041] In one embodiment, each group may have similar logistics needs or behavior characteristics. For example, users may be divided into frequent shippers with a shipping frequency higher than a first preset threshold, occasional shippers with a shipping frequency lower than a second preset threshold, non-shippers with a shipping frequency of zero, frequent consignees with a receiving frequency higher than a third preset threshold, occasional consignees with a receiving frequency lower than a fourth preset threshold, non-receivers with a receiving frequency of zero, recent shippers with a historical shipping record within a preset time period, and recent consignees with a historical receiving record within a preset time period.

[0042] Furthermore, the user's geographical location, weather conditions, holidays and other external factors can be combined to further provide convenient services for the user. For example, based on the user portrait and external factors, the questions that the user may ask can be inferred, and these questions can be presented to the user through the interactive interface. Specifically, the target user's question expectations can be generated based on the user portrait, the user's geographical location information, weather information and holiday information. The question expectations here can be a feature matrix, and then historical users similar to the target user's question expectations can be determined in the historical database, and the historical question data corresponding to the historical users can be obtained; finally, a question template can be selected based on the historical question data, and the question template can be presented to the target user.

[0043] In one embodiment, the logistics language model will also record the context information of the question and answer during the question and answer process. By understanding and remembering the user's context information, it can answer the user's questions more accurately. For example, when a user asks multiple related questions in succession, the system can infer the user's intentions and needs based on the previous answers and give a more accurate answer.

[0044] At the same time, the logistics language model supports multiple languages ​​and can meet the needs of users in different countries and regions around the world. Through multilingual training, the logistics language model can understand and generate answers in multiple languages, improving the service capabilities of international logistics companies. The system has a built-in translation function, and users can choose to translate answers into different languages, facilitating communication between multinational companies and customers.

[0045] In one embodiment, after the logistics knowledge is returned to the logistics language model and the answer to the target logistics question is generated, the answer to the question needs to be sent to the user and feedback information from the user needs to be received. The feedback information here may include satisfaction and error correction. Then, based on the feedback information and the self-learning mechanism, the logistics knowledge graph can be updated and the logistics language model can be adjusted.

[0046] In addition to building a structured logistics knowledge graph, in logistics scenarios, there is a wealth of multimodal information such as cargo images, transport vehicle videos, and voice commands. For example, photos of damaged cargo can assist in determining liability and claims processes, and vehicle driving voice records can analyze transportation status. A multimodal fusion architecture can be built, and convolutional neural networks can be used to process image data to obtain cargo features. Speech recognition technology can be used to convert speech into text mining requirements. The feature fusion module can be used to fuse multimodal features with text features and input into a large model to improve its comprehensive and accurate understanding of logistics business, make up for the limitations of a single text modality, and optimize the question-answering system's ability to respond to complex multimodal logistics scenarios.

[0047] An embodiment of the present application also provides a logistics question answering device based on a large language model, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: obtain input information corresponding to a target logistics problem, perform natural language processing on the input information, and obtain parsed data of the target logistics problem; input the parsed data into a pre-trained logistics language model, wherein the logistics language model is a large language model fine-tuned using logistics industry corpus; retrieve the parsed data through a logistics knowledge graph associated with the logistics language model to obtain logistics knowledge corresponding to the parsed data; return the logistics knowledge to the logistics language model to generate an answer to the target logistics problem.

[0048] The embodiment of the present application further provides a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions are configured as follows:

[0049] Acquire input information corresponding to a target logistics problem, perform natural language processing on the input information, and obtain parsed data for the target logistics problem; input the parsed data into a pre-trained logistics language model, where the logistics language model is a large language model fine-tuned using logistics industry corpus; retrieve the parsed data through a logistics knowledge graph associated with the logistics language model to obtain logistics knowledge corresponding to the parsed data; return the logistics knowledge to the logistics language model to generate an answer to the target logistics problem.

[0050] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0051] The devices and media provided in the embodiments of the present application correspond one-to-one to the methods. Therefore, the devices and media also have similar beneficial technical effects as the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0052] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0053] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0054] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0055] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0056] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0057] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0058] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include transitory media such as modulated data signals and carrier waves.

[0059] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0060] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A logistics question answering method based on a large language model, characterized in that: include: Obtaining input information corresponding to a target logistics problem, performing natural language processing on the input information, and obtaining analytical data of the target logistics problem; Inputting the parsed data into a pre-trained logistics language model, where the logistics language model is a large language model fine-tuned using logistics industry corpus; Retrieving the parsed data through a logistics knowledge graph associated with the logistics language model to obtain logistics knowledge corresponding to the parsed data; The logistics knowledge is returned to the logistics language model to generate an answer to the target logistics question.

2. The method according to claim 1, characterized in that The step of obtaining input information corresponding to the target logistics problem specifically includes: Determine an access port of the user, where the access port includes at least one of a web terminal, a mobile terminal, and a voice assistant; Based on the access port, providing a user interaction interface to the user and obtaining an initial input from the user; If the initial input information is of voice type, voice recognition is performed on the initial input information to obtain text type input information.

3. The method according to claim 1, characterized in that Before retrieving the parsed data through the logistics knowledge graph associated with the logistics language model, the method further includes: Extracting entities and relationships from an existing logistics industry corpus, wherein the logistics industry corpus includes at least one of a logistics enterprise database, an industry standard document, and logistics news information; Based on the entities and relationships, generate multiple tuples to form the logistics knowledge graph; And / or, the retrieving the parsed data through the logistics knowledge graph associated with the logistics language model specifically includes: Constructing a query statement based on the keywords contained in the parsed data; The query statement is executed in the logistics knowledge graph to obtain entities and relationships related to the parsed data.

4. The method according to claim 1, characterized in that: The returning the logistics knowledge to the logistics language model to generate an answer to the target logistics question specifically includes: The parsed data is converted into a word vector representation containing context information through a bidirectional transformer model in the logistics language model; Processing the word vector representations through a BiLSTM layer to capture long-range dependencies in the parsed data; The output of BiLSTM is globally optimized through the conditional random field layer to ensure that the generated label sequence meets the grammatical and semantic requirements.

5. The method according to claim 1, characterized in that The method further comprises: Recording user input information, including question information, query information, answer selection information, and feedback information; When the data volume corresponding to the input information of the target user is higher than a preset threshold, generating a user profile of the target user according to the input information; Based on the user portrait and in combination with the user's geographical location information, weather information and holiday information, the answer to the question is corrected.

6. The method according to claim 5, characterized in that The user portraits include: frequent shippers whose shipping frequency is higher than a first preset threshold, occasional shippers whose shipping frequency is lower than a second preset threshold, non-shippers whose shipping frequency is zero, frequent consignees whose receiving frequency is higher than a third preset threshold, occasional consignees whose receiving frequency is lower than a fourth preset threshold, non-consignees whose receiving frequency is zero, recent shippers whose historical shipping records are within a preset time period, and recent consignees whose historical receiving records are within a preset time period.

7. The method according to claim 6, characterized in that The method further comprises: Generate the target user's question expectation based on the user portrait, the user's geographical location information, weather information, and holiday information; Determine in a historical database historical users with similar question expectations to the target user; Obtaining historical question data corresponding to the historical user; Based on the historical question data, a question template is selected and presented to the target user.

8. The method according to claim 1, characterized in that: After returning the logistics knowledge to the logistics language model to generate an answer to the target logistics question, the method further includes: Sending the answers to the questions to the user and receiving feedback from the user, the feedback including satisfaction and correction of errors; Through the feedback information and the self-learning mechanism, the logistics knowledge graph is updated and the logistics language model is adjusted.

9. A logistics question answering device based on a large language model, characterized in that: include: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to execute: Obtaining input information corresponding to a target logistics problem, performing natural language processing on the input information, and obtaining analytical data of the target logistics problem; Inputting the parsed data into a pre-trained logistics language model, where the logistics language model is a large language model fine-tuned using logistics industry corpus; Retrieving the parsed data through a logistics knowledge graph associated with the logistics language model to obtain logistics knowledge corresponding to the parsed data; The logistics knowledge is returned to the logistics language model to generate an answer to the target logistics question.

10. A non-volatile computer storage medium storing computer executable instructions, characterized in that: The computer executable instructions are configured to: Obtaining input information corresponding to a target logistics problem, performing natural language processing on the input information, and obtaining analytical data of the target logistics problem; Inputting the parsed data into a pre-trained logistics language model, where the logistics language model is a large language model fine-tuned using logistics industry corpus; Retrieving the parsed data through a logistics knowledge graph associated with the logistics language model to obtain logistics knowledge corresponding to the parsed data; The logistics knowledge is returned to the logistics language model to generate an answer to the target logistics question.

Images