Port logistics information processing method based on general big language model

Abstract

The invention relates to a port logistics information processing method based on a general large language model. The method comprises the following steps: S1, collecting logistics clue information; S2, preprocessing the logistics clue information; S3, marking the logistics clue information; the method has the advantages that port logistics information chains can be effectively integrated, information islands can be broken, more comprehensive, accurate and timely information support can be provided for port logistics services, and compared with a traditional method that an AI knowledge base processes and stores information in a text slicing mode, the method has the advantage that the information processing efficiency is improved. The method is applied to retrieval of internal information in the field of port services, the integration efficiency is remarkably improved, higher flexibility is achieved in the data application stage, and clue groups can be integrated into coherent logistics chain information by selecting a traditional series connection mode or a series connection mode based on a general large language model according to actual requirements.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data processing, in particular to a port logistics information processing method based on a general large language model. BACKGROUND

[0002] Port logistics covers many aspects such as ship transportation, wharf loading and unloading, warehouse management, and cargo distribution, and involves multiple participating parties such as wharfs, shipping companies, cargo owners, freight agents, and transportation companies. The business between the participating parties is frequent, and the logistics information interaction is huge.

[0003] Under the traditional port logistics information processing mode, the participating parties mainly rely on paper documents such as bills of lading and delivery notes, or transmit logistics information through emails and instant messaging tools. This method is not only inefficient, but also prone to information errors and delays.

[0004] In recent years, with the gradual popularization of information systems, parties have begun to exchange logistics information through messages and data interfaces. However, there are still many problems: on the one hand, due to the diversity and complexity of port logistics business, the data formats and standards used by different parties differ greatly, and there is a lack of unified standard for data exchange protocols, resulting in huge compatibility problems for participating parties in parsing and obtaining logistics information, making it difficult to achieve seamless integration and efficient sharing of information; on the other hand, port logistics information is highly dynamic and uncertain, and traditional information processing methods are difficult to collect, organize and analyze massive, multi-source and heterogeneous logistics information in real time, and cannot meet the strict requirements of port logistics business on information timeliness and accuracy.

[0005] The emergence of general large language models brings new hope to solve these problems. With its powerful organization and induction ability for unordered information, it can efficiently organize, induce and classify information in text segments with unknown arrangement rules, accurately identify and extract key information, and realize intelligent processing of logistics clue information from different channels and arbitrary formats. SUMMARY

[0006] The purpose of the present application is to overcome the above problems existing in the prior art and provide a port logistics information processing method based on a general large language model.

[0007] To achieve the above technical purposes and effects, the present application realizes the following technical solutions: The port logistics information processing method based on a general large language model comprises the following steps, S1, collecting logistics clue information: collecting various logistics clue information of port supply chain in real time by integrating multiple information collection channels; S2, logistics clue information preprocessing: processing various types of logistics clue information into long text format logistics clue information; S3, marking logistics clue information: using a general large language model to extract specified field values in long text format logistics clue information, and marking long text format logistics clue information with port logistics related marks according to the extracted specified field values; S4, store logistics clue information: store logistics clue information in a multi-table storage structure including one main table and multiple sub-tables, the main table stores original logistics clue information, and assigns a primary key to each original logistics clue information as a unique identity of logistics clue information in the entire storage system, each sub-table stores a mark and the association between the mark and the primary key of the original logistics clue information; S5, retrieve logistics clue information to form a clue group: based on the stored data, retrieve from any sub-table, the relevant marks and original logistics clue information to form a clue group; S6, integrate clue groups into logistics chain information in series: integrate clue groups into coherent logistics chain information in series using traditional series or based on general large language model series.

[0008] Among them, the information collection channel includes email, various standardized or non-standardized data interfaces, robotic process automation applications, and online customer service platforms that directly interact with customers.

[0009] Among them, the type of logistics clue information includes text files, image files, audio files, and video files.

[0010] Among them, the mark includes ship name, departure place, destination, cargo name, cargo weight, container number, cargo owner information, transportation status, cargo arrival mode, cargo arrival time, transportation time, cargo departure mode, and cargo departure time.

[0011] Among them, the use of general large language model to extract specified field values in long text format logistics clue information is: inputting long text format logistics clue information into a prompt word input general large language model based on prompt word engineering development, the general large language model uses its powerful natural language processing ability and deep understanding of text semantics to accurately extract specified field values corresponding to the prompt word from long text format logistics clue information, and outputs the results of specified field values in a structured form.

[0012] Among them, the traditional series method is to integrate the marks in the clue group into coherent logistics chain information in series according to time sequence or logical sequence.

[0013] The general large language model based serial connection method is that the clue group is re-input into the general large language model, the general large language model is used to comprehensively evaluate and correlation analyze the clue group from multiple dimensions by using the deep understanding and comprehensive analysis ability of the general large language model on information, and finally a logistics chain information containing logistics clue information and prediction information is output.

[0014] The beneficial effects of the present application are that not only the port logistics information chain can be effectively integrated, and the information island can be broken, but also more comprehensive, accurate and timely information support can be provided for the port logistics business. Compared with the method of using text slicing to process and store information in the traditional AI knowledge base, the method applied in the port business field has significantly improved the efficiency of information retrieval and integration, and has higher flexibility in the data application stage. The traditional serial connection method or the general large language model based serial connection method can be selected to integrate the clue group into coherent logistics chain information according to actual needs. BRIEF DESCRIPTION OF DRAWINGS

[0015] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and serve to explain the principles of the application. In the drawings: Figure 1 is a flow chart of the port logistics information processing method in the present application. DETAILED DESCRIPTION

[0016] The present application will be described in detail below with reference to the accompanying drawings and embodiments.

[0017] As shown in the drawings, Figure 1 The port logistics information processing method based on the general large language model includes the following steps: S1, collecting logistics clue information, S2, preprocessing logistics clue information, S3, marking logistics clue information, S4, storing logistics clue information, S5, retrieving logistics clue information to form a clue group, and S6, integrating the clue group into logistics chain information by serial connection.

[0018] Collecting logistics clue information: collecting various types of logistics clue information of the port supply chain in real time by integrating various information collection channels.

[0019] The information collection channels include electronic mail, various standardized or non-standardized data interfaces, robot process automation application programs, and online customer service platforms for direct interaction with customers. The various standardized or non-standardized data interfaces are, for example, data interfaces of scanning equipment, data interfaces of audio and video shooting equipment, and data interfaces of port gate systems. The types of logistics clue information include text files, image files, audio files, and video files.

[0020] Preprocessing of logistics clue information: various types of logistics clue information are processed into long text format logistics clue information. For example: image files use optical character recognition technology to accurately identify and convert the text content in the picture into editable and processable text form; data from databases or special format files are converted into long text suitable for general large language model processing through data parsing technology.

[0021] Labeling logistics clue information: using a general large language model to extract specified field values from long text format logistics clue information, and labeling long text format logistics clue information with port logistics related labels according to the extracted specified field values.

[0022] The labels include ship name, departure place, destination, cargo name, cargo weight, container number, cargo owner information, transportation status, cargo arrival mode, cargo arrival time, transportation time, cargo departure mode, and cargo departure time. And according to the expansion and changes of actual business needs, the names of the labels can be flexibly added or adjusted.

[0023] Using a general large language model to extract specified field values from long text format logistics clue information: inputting long text format logistics clue information into a general large language model based on prompt word engineering, the general large language model uses its powerful natural language processing ability and deep understanding of text semantics to accurately extract the specified field values corresponding to the prompt words from the long text format logistics clue information, and outputs the results of the specified field values in a structured form.

[0024] After the general large language model outputs the results of the specified field values, these field information will be normalized and standardized to ensure the format of the data is uniform and accurate, and then converted into various attribute value labels of logistics clue information. These labels comprehensively and accurately reflect the key information contained in the logistics clue information, and push them to the next storage link, laying a solid foundation for subsequent information management and application.

[0025] Storing logistics clue information: using a multi-table storage structure including one main table and multiple sub-tables to store logistics clue information. The main table stores the original logistics clue information and assigns a primary key to each original logistics clue information as the unique identity of the logistics clue information in the entire storage system. Each sub-table stores a label and the association between the label and the primary key of the original logistics clue information.

[0026] Each time a new label is added, a corresponding sub-table will be automatically created to store the information of the label and the association between the corresponding primary key.

[0027] The design of this multi-table storage structure has significant advantages. On the one hand, it can ensure the independent storage and integrity of the original information and the mark information, avoiding mutual interference and confusion between data. On the other hand, through the association between the primary key and the sub-table, multi-dimensional and flexible query and retrieval functions of the logistics clues can be realized.

[0028] For example, the user can quickly query the container number in the corresponding sub-table according to the specific container number, and obtain all the logistics clues related to the container according to the association between the primary key and the sub-table. Or according to different dimensions such as time range, cargo name, etc. to query the logistics clues, so as to meet the complex and diverse information query needs in port logistics business, and provide strong support for the depth analysis and application of logistics information.

[0029] Retrieving logistics clue information forms a clue group: based on the stored data, retrieving from any sub-table, forming a clue group with related marks and original logistics clue information.

[0030] For example, when tracking the logistics information of a certain container, the system can use the container number as the query condition in the container number sub-table to quickly retrieve all the logistics clue information related to the container in the database, and aggregate these logistics clue information to form a clue group with the container number as the dimension.

[0031] Similarly, the user can also retrieve according to other dimensions such as ship name, cargo owner information, time range, etc. to generate corresponding clue groups to meet the information query and analysis needs in different business scenarios.

[0032] Integrating clue groups into logistics chain information: using traditional concatenation method or general large language model-based concatenation method to integrate clue groups into coherent logistics chain information.

[0033] Traditional concatenation method: concatenating and integrating marks in clue groups into coherent logistics chain information according to time sequence or logical sequence.

[0034] General large language model-based concatenation method: re-inputting clue groups into general large language model, using the deep understanding and comprehensive analysis capabilities of general large language model to comprehensively evaluate and analyze clue groups from multiple dimensions, and finally outputting a logistics chain information containing logistics clue information and prediction information. For example, the general large language model can automatically infer important information such as potential risk points and estimated arrival time of the goods during transportation based on the time, location, transportation method, etc. in the clue group, and integrate it into the logistics chain information, providing more rich decision basis for business personnel.

[0035] Embodiment: Taking the container logistics business of a certain large port as an example, the practical application process and effect of the present application are specifically described. In the daily operation of the port, a large number of containers are transferred and transported by various transportation modes such as ships, railways and highways every day, involving many participants such as consignors, forwarding agents, shipping companies and transportation enterprises, and the interaction of logistics information is extremely frequent and complex.

[0036] In the data collection link, by integrating with the existing business systems of the port, the transportation plan and ship dynamic information of the shipping company are obtained in real time through data interfaces, including the port arrival time, departure time, loading and unloading operation plan, container loading list and other detailed data of the ship; at the same time, the customer sends the goods order information, special goods transportation requirements and pick-up instructions and other information through email and online customer service platform; in addition, the robot process automation application deployed in each business link of the port automatically collects the operation records of the terminal loading and unloading equipment, the warehouse in-out records and other on-site logistics data, and all kinds of data are collected to the information collection system.

[0037] In the information preprocessing and labeling stage, the system classifies and processes various types of logistics clue information collected, uses optical character recognition technology to identify the text content in the bill of lading picture, and converts it into text format; the structured data from the database are parsed and reorganized to form long texts suitable for model processing.

[0038] Then, combined with the carefully designed prompt words, these long texts are input into the general large language model, and the general large language model accurately extracts key field values such as container number, ship name, consignor information, cargo description, transportation status, etc. from them, and generates corresponding attribute value labels. For example, for a ship port operation logistics clue, the model can accurately extract the ship name as "XX round", the port arrival time as "2025-01-10 08:00", and the loading and unloading container numbers including "COSU1234567" "COSU7654321" and other detailed information. These labeled information is then stored in the main table and each sub-table of the database, the main table records the original data source and collection time of the logistics clue information and other basic information, and the sub-tables store the labeled content of the container number, ship name, time and other dimensions and their association with the main table primary key.

[0039] In the information query and stringing application, when the port business personnel need to query the complete logistics chain information of a certain container, they only need to input the container number in the database, and the database will quickly retrieve all the logistics clues related to the container in the container number sub-table, forming a clue group with the container as the dimension.

[0040] Then, by using the traditional serial method, data matching and retrieval are carried out in the time sub-table and the place sub-table through the primary key in the clue group, the whole process information of the container from the departure of the ship to the arrival at the destination port, from the unloading at the dock to the transportation to the warehouse or the final consignee is serially connected in time sequence, a detailed space-time logistics chain report is generated, and each link of the container in the logistics process is intuitively displayed, including the time node, the place position and the state change.

[0041] Meanwhile, the database can also select a serial connection mode based on a general large language model, re-input the clue group information into the general large language model, the general large language model comprehensively analyzes the multi-dimensional data of the transportation mode, the cargo information, the participant information and the like of the container in each logistics clue, and outputs an intelligent logistics chain analysis report containing potential risk analysis, predicted subsequent flow path prediction and the like, thereby providing comprehensive and in-depth support for the decision of the business personnel in optimizing the transportation scheme, arranging the loading and unloading operation in advance, improving the quality of customer service and the like.

[0042] The method of the application significantly improves the efficiency and quality of logistics information processing, and provides strong support for the operation and management of the port.

[0043] The above shows and describes the basic principles, main features and advantages of the application. It should be understood by those skilled in the art that the application is not limited by the above examples, the above examples and descriptions in the specification are only to illustrate the principles of the application, and various changes and improvements can be made without departing from the spirit and scope of the application, and these changes and improvements all fall within the scope of the claimed application.

Claims

1. A port logistics information processing method based on a general large language model, characterized in that, The method comprises the following steps: S1, collecting logistics clue information: collecting various logistics clue information of port supply chain in real time through integrating various information collection channels; S2, preprocessing logistics clue information: processing various types of logistics clue information into long text format logistics clue information; S3, marking logistics clue information: using a general large language model to extract specified field values in long text format logistics clue information, and marking long text format logistics clue information with port logistics related marks according to the extracted specified field values; S4, storing logistics clue information: storing logistics clue information by using a multi-table storage structure including one main table and multiple sub-tables, the main table stores original logistics clue information, and assigns a primary key to each original logistics clue information as a unique identity of logistics clue information in the entire storage system, each sub-table stores a mark and the association between the mark and the primary key of the original logistics clue information; S5, retrieving logistics clue information to form a clue group: based on the stored data, retrieving from any sub-table, the related marks and original logistics clue information form a clue group; S6, integrating clue groups into logistics chain information in series: integrating clue groups into coherent logistics chain information in series using a traditional series connection method or a general large language model series connection method.

2. The port logistics information processing method according to claim 1, characterized in that: The information collection channels include email, various standardized or non-standardized data interfaces, robotic process automation applications, and online customer service platforms that directly interact with customers.

3. The port logistics information processing method according to claim 1, characterized in that: The types of logistics clue information include text files, image files, audio files, and video files.

4. The port logistics information processing method of claim 1, wherein: The marks include ship name, departure location, destination, cargo name, cargo weight, container number, cargo owner information, transportation status, cargo arrival mode, cargo arrival time, transportation time, cargo departure mode, and cargo departure time.

5. The port logistics information processing method of claim 1, wherein, The general large language model extracts specified field values in long text format logistics clue information by inputting long text format logistics clue information into a prompt word input general large language model based on prompt word engineering, and the general large language model accurately extracts specified field values corresponding to prompt words from long text format logistics clue information using its powerful natural language processing capabilities and deep understanding of text semantics, and outputs the results of specified field values in a structured form.

6. The port logistics information processing method of claim 1, wherein, The traditional series connection method is to integrate the marks in the clue group into coherent logistics chain information in series according to time sequence or logical sequence.

7. The port logistics information processing method of claim 1, wherein, The general large language model series connection method is to re-input the clue group into the general large language model, use the deep understanding and comprehensive analysis capabilities of the general large language model to comprehensively evaluate and analyze the clue group from multiple dimensions, and finally output a logistics chain information containing logistics clue information and prediction information.

Images