Logistics processing method, electronic equipment, storage medium and program product

Abstract

The embodiment of the invention provides a logistics processing method, electronic equipment, a storage medium and a program product. The method comprises the steps that multiple pieces of description information of a target article and the initial specification of the target article uploaded by a first client are acquired, and the multiple pieces of description information comprise information of different modes; performing feature extraction on the multiple pieces of description information to obtain target article features of the description information; and inputting the target article features and the initial specifications of the description information to a target model for processing to obtain target logistics attributes of the target article and reasoning information of the target logistics attributes. According to the embodiment of the invention, by combining the cross-modal description information, high-robustness identification of the target article is realized, the accuracy and generalization ability of logistics attribute identification are improved, and the technical problem that the logistics attribute identification ability of the article is not accurate in the related technology is solved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a logistics processing method, electronic device, storage medium, and program product. Background Technology

[0002] Logistics is a core component of after-sales fulfillment, impacting the receiving experience, resource consumption, and delivery efficiency. The logistics attributes of an item, such as whether it is fragile, a knife, a dangerous good, or a cold chain item, as well as its physical attributes like weight and volume, determine the logistics methods, including resource consumption, delivery range, and packaging requirements.

[0003] However, in real-world applications, the vast majority of logistics demanders do not proactively maintain the logistics attribute information of their goods, resulting in a lack of structured and reliable data support on the server side. Furthermore, the weight and volume data transmitted back by logistics providers generally contain significant errors. Over 50% of the transmitted data deviates severely from the true values ​​due to reasons such as unremoved packaging, combined weighing of multiple items, insufficient accuracy of measuring equipment, and human error (e.g., laziness / underweighting), exhibiting significant weight fluctuations in logistics transmission and further exacerbating the uncertainty in determining logistics attributes. Summary of the Invention

[0004] This application provides a logistics processing method, electronic device, storage medium, and program product to alleviate or solve the technical problem of inaccurate identification of the logistics attributes of items in related technologies.

[0005] In a first aspect, embodiments of this application provide a logistics processing method applied to a server, comprising: Obtain multiple descriptive information of the target item and the initial specifications of the target item uploaded by the first client, wherein the multiple descriptive information includes information of different modalities; Feature extraction is performed on each of the multiple descriptive information entries to obtain the target item features of each of the descriptive information entries; Input the target item features and initial specifications of each of the aforementioned descriptive information into the target model for processing, and obtain the target logistics attributes of the target item and the inference information of the target logistics attributes; Wherein, the reasoning information represents the reasoning process by which the target model obtains the target logistics attributes based on the target item features and the initial specifications, the target logistics attributes are used to determine the logistics method of the target item, and the target model is trained based on predetermined reference item features, reference logistics attributes, and reference reasoning information of the reference logistics attributes.

[0006] In the embodiments provided in this application, the step of extracting features from the plurality of descriptive information to obtain the target item features of each of the descriptive information includes: Each sub-model extracts features based on the corresponding descriptive information to obtain the target item features corresponding to each descriptive information, and the single-modal attributes corresponding to the target item features; each sub-model is used to process the modal features of the corresponding descriptive information, and the single-modal attributes represent the single-modal logistics attributes of the target item determined based on the target item features of the corresponding descriptive information.

[0007] In the embodiments provided in this application, the target item features and the initial specifications of the input description information are processed into a target model to obtain the target logistics attributes of the target item and inference information about the target logistics attributes, including: Input the target item features corresponding to each of the aforementioned description information and the initial specifications into the target model for processing to obtain the initial logistics attributes of the target item; If the initial logistics attributes are found to contradict any attribute among the single-modal attributes obtained from each of the sub-models, follow-up questions are generated based on the contradictory attributes. Based on the follow-up information and the corresponding description information, each of the sub-models is used for processing to obtain the first attribute of each sub-model for the follow-up information; Based on the initial logistics attributes and the first attributes obtained from each of the sub-models, the target logistics attributes are determined.

[0008] In the embodiments provided in this application, the target logistics attributes include modified specifications and logistics type, wherein the modified specifications are obtained by modifying the initial specifications, and the method further includes: Once the logistics type indication is determined to be a predetermined type, the transportable area corresponding to the item of the predetermined type is obtained; Based on the revised specifications and the transportable area, determine the logistics rights of the target item; Send a first prompt message to the client, the first prompt message being used to indicate at least one of the following: the deliverable area of ​​the target item, the logistics rights, or the logistics type.

[0009] In the embodiments provided in this application, the step of extracting features from the plurality of descriptive information to obtain the target item features of each of the descriptive information includes: Based on the identification of multiple descriptive information, the target item is determined to be a combined item including multiple sub-items; Feature extraction is performed on the description information including the sub-items to obtain the sub-item features corresponding to the sub-items, which are used as the target item features.

[0010] In the embodiments provided in this application, the different modalities include text modalities and image modalities. The descriptive information of the image modal includes the display image in the display page of the target item, and the descriptive information of the text modal includes the summary information of the display page, the comment information on the target item, or the category information of the target item.

[0011] In the embodiments provided in this application, the step of extracting features from the plurality of descriptive information to obtain the target item features of each of the descriptive information includes: The displayed image is determined to contain text information. The text information is then subjected to text recognition to obtain the first item feature. Based on the displayed image, the material of the target item is identified to obtain the second item feature; Based on the first item features and the second item features, the target item features corresponding to the description information of the image modality are obtained.

[0012] Secondly, embodiments of this application provide a logistics processing method applied to a first client, comprising: Send the initial specifications of the target item to the server; Receive and display a first prompt message and inference information, wherein the first prompt message is used to indicate the logistics method of the target item; Based on the aforementioned logistics method, logistics scheduling is performed on the target items; The logistics method is determined based on the target logistics attributes of the target item, and the reasoning information represents the reasoning process by which the target model obtains the target logistics attributes based on the target item characteristics and the initial specifications of the target item.

[0013] Thirdly, embodiments of this application provide an electronic device, including a memory, a processor, and a computer program stored in the memory, wherein the processor implements any of the methods of embodiments of this application when executing the computer program.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method of any one of the embodiments of this application.

[0015] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements any of the methods described in the embodiments of this application.

[0016] Based on the logistics processing method described in the first aspect above, this application has at least the following beneficial effects or advantages: It eliminates the need for the logistics demander to maintain the logistics attributes of the target item, automatically completes missing attributes, effectively covers the perception of logistics attributes for low-frequency items, and outputs the judgment results of logistics attributes in natural language, thereby enhancing the interpretability of the model and facilitating the verification and traceability of logistics attributes. It reduces the reliance on the accuracy of specification data returned by the logistics delivery party, and utilizes the reasoning capabilities embedded in the target model to accurately infer logistics attributes even if the item description is not explicitly marked, overcoming the rigidity and misjudgment problems of traditional keyword matching or rule systems.

[0017] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application, it can be implemented according to the contents of the specification. In order to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0018] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the various drawings denote the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings depict only some embodiments according to this application and should not be construed as limiting the scope of this application.

[0019] Figure 1 A flowchart of a logistics processing method according to an embodiment of this application is shown; Figure 2 A flowchart of another logistics processing method according to an embodiment of this application is shown; Figure 3 An interactive block diagram of a logistics processing method according to an embodiment of this application is shown; Figure 4 A flowchart of a logistics processing apparatus according to an embodiment of this application is shown; Figure 5 A flowchart of another logistics processing apparatus according to an embodiment of this application is shown; Figure 6 A block diagram of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0020] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the concept or scope of this application. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0021] To facilitate understanding of the technical solutions of the embodiments of this application, the relevant technologies of the embodiments of this application are described below. The following relevant technologies are optional solutions and can be combined with the technical solutions of the embodiments of this application in any way, and all of them fall within the protection scope of the embodiments of this application.

[0022] It should be noted that the application scenarios or examples provided in this application embodiment are for ease of understanding, and this application embodiment does not specifically limit the application of the technical solution. Furthermore, the user information (including but not limited to user device information, user personal information, item description information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, item specification data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. The collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0023] Traditional methods relying on manual labeling or single data sources (such as logistics feedback) are no longer sufficient to meet the demands for high accuracy and coverage in logistics attribute identification. One existing technology provides a statistical modeling method based on logistics feedback data, relying on weighing and volume data returned by courier companies to estimate the true weight and volume of items using historical data. Some systems combine category or label information provided by the logistics requester for rule matching to determine whether the item is a special category such as cold chain or knives. However, this method is highly dependent on the quality of external data. In reality, courier feedback data suffers from systematic biases, such as the failure to deduct the weight of cardboard boxes or the mixing of various items, leading to estimation errors generally exceeding 20%, making it difficult to meet the accuracy requirements for logistics attributes.

[0024] One existing technology provides a method based on a single-modal machine learning model. This method uses a natural language processing (NLP) model to parse item titles or category text, determining if they contain keywords such as "fragile," "fresh," or "knife," or uses a computer vision (CV) model to identify visual features in item images such as glass bottles or frozen packaging. However, this technology is limited to a single modality, lacks contextual understanding, and is susceptible to ambiguity or image occlusion. For example, a "ceramic cup" may not be explicitly labeled "fragile," but it actually belongs to an item requiring additional transportation methods to ensure safe delivery; and "fresh" items may appear in descriptions of non-cold chain items, leading to misjudgments. In a prior art, a rule-based reasoning system based on knowledge graphs constructs a knowledge base that maps items to attributes. It matches categories with preset attributes through a rule engine. However, this hard matching method has poor scalability, is difficult to cover low-frequency items, and cannot handle items or bundled items that have not appeared in the knowledge base. It also requires a high level of manual maintenance. The above analysis shows that existing related technologies generally suffer from problems such as strong reliance on the accuracy of back-transmitted data, weak generalization ability, information fragmentation, and lack of semantic understanding, making it impossible to achieve automated and high-precision identification of massive, dynamic, and complex logistics attributes.

[0025] This application embodiment integrates information from multiple modalities, utilizes a large reasoning model combined with its inherent common sense knowledge and reasoning ability to achieve intelligent inference of the logistics attributes of goods, solves the problems of data missingness and noise, overcomes dependence on a single data source, improves semantic understanding and common sense reasoning ability, supports generative output and interpretability, enhances decision transparency, realizes cross-modal collaborative judgment, and promotes the paradigm shift of logistics processing from passively relying on back-transmitted data to actively generating logistics attributes.

[0026] The technical solution of this application and how it solves the aforementioned technical problems are described in detail below with specific embodiments. The listed specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0027] Figure 1 A flowchart of a logistics processing method according to an embodiment of this application is shown, such as... Figure 1 As shown, the method may include steps S101, S102 and S103.

[0028] Step S101: Obtain multiple description information of the target item and the initial specifications of the target item uploaded by the first client. The multiple description information includes information on different modalities. Step S102: Extract features from multiple descriptive information to obtain the target item features for each descriptive information; Step S103: Input the target item features and initial specifications of each descriptive information into the target model for processing to obtain the target logistics attributes of the target item and the inference information on the target logistics attributes; Among them, the inference information represents the reasoning process by which the target model obtains the target logistics attributes based on the characteristics and initial specifications of the target item. The target logistics attributes are used to determine the logistics method of the target item. The target model is trained based on the predetermined reference item characteristics, reference logistics attributes, and reference inference information of the reference logistics attributes.

[0029] The aforementioned optional execution entity is the server. Multiple descriptive information entries for the target item can be stored internally or externally on the server. These entries can be in different modalities, such as text or images, to comprehensively reflect the multi-dimensional characteristics of the target item. The first client can be a client used by the logistics and delivery provider. The initial specifications returned by the first client can optionally include data such as the weight and volume of the target item. The server performs feature extraction processing on each of the multiple descriptive entries. By adapting the feature extraction methods to each modality, it extracts target item features that characterize the target item's properties from each descriptive entry. By fusing descriptive information from different modalities with the initial specifications returned by the first client, the feature loss problem caused by a single information source is avoided, providing sufficient data support for subsequent logistics attribute determination. The server inputs the target item features corresponding to each descriptive entry and the initial specifications into the target model for processing. The target model outputs the target logistics attributes of the target item and the inference information corresponding to those attributes.

[0030] The aforementioned target model is trained based on predetermined reference item features, reference logistics attributes, and reference inference information, providing ample information support for the model training phase. Reference item features encompass item representation data from different modalities, reference logistics attributes correspond to verified and suitable logistics parameters for various items, and reference inference information records the logical derivation process from item features to logistics attributes. Together, these elements form the data foundation for model training, ensuring that the model can fully learn the correlation and inference logic between item features and logistics attributes during training. The output target logistics attributes can be directly used to determine the logistics method of the target item, improving the accuracy of logistics method matching. The output of inference information makes the determination process of logistics attributes traceable, enhancing the credibility of the processing results, and providing a valid basis for subsequent model optimization or anomaly investigation.

[0031] The target model mentioned above can be a reasoning model such as a deep learning model or a neural network model. A reasoning model refers to an artificial intelligence model that has the ability of logical reasoning, causal analysis, common sense inference, and multi-step thinking. It can not only identify the surface features of the input data, but also simulate the human thinking process through internal mechanisms to decompose, deduce, summarize, and judge complex problems, thereby generating output results with logical coherence and semantic rationality, obtaining interpretable reasoning information, and increasing the trust in the output results of the target model.

[0032] For example, the aforementioned target logistics attributes may include multiple attributes, and the same target item may have one or more attributes. For instance, target logistics attributes may include fragile items, sharp objects (knives), virtual items, cold chain items, weight range, volume range, etc. Virtual items do not require actual logistics scheduling.

[0033] For example, the initial specifications can be derived from historical feedback data of similar items to the target item. Multiple historical feedback data sets of similar items serve as the foundational data source for the initial specifications. Relying on a sub-model with built-in statistical functions on the server, targeted statistical processing is performed on the aforementioned historical feedback data of similar items. This statistical sub-model takes multiple historical feedback data sets of similar items as input, and through data cleaning, feature statistics, and interval analysis, anchors a reasonable range for the initial specifications of this category of target items. This results in the generation of statistically optimized specifications, which are then used as one of the input parameters. These specifications, along with the target item features extracted from multimodal description information, are imported into the inference model to assist the inference model in generating corrected specifications included in the target logistics attributes. By using historical feedback data of similar items as the basis for the initial specifications, combined with the analysis and processing of the statistical sub-model, the specifications input into the target model are made more closely aligned with the actual characteristics of this type of item, effectively filtering outliers in historical data and improving the quality of item specification data.

[0034] In this embodiment of the application, feature extraction is performed on multiple descriptive information in step S102 to obtain the target item features of each descriptive information, which may include the following steps: Each sub-model extracts features based on the corresponding descriptive information to obtain the target item features corresponding to each descriptive information, as well as the single-modal attributes corresponding to the target item features. Each sub-model is used to process the modal features of the corresponding descriptive information. The single-modal attribute represents the single-modal logistics attribute of the target item determined based on the target item features of the corresponding descriptive information.

[0035] In this embodiment of the application, the server is configured with multiple sub-models. Each sub-model can be set to process description information of a specific modality. When multiple different modal description information of the target item is obtained, each description information will be input into its corresponding sub-model for processing.

[0036] Each sub-model extracts features based on the received unimodal description information. It not only outputs the extracted target item features but also performs qualitative analysis based on these features to determine the unimodal attributes. These attributes represent the initial conclusions about the logistics attributes of the target item independently determined by each sub-model based on the description information from its own single data source. These conclusions may be limited by the partial description of a single information source and require comprehensive reasoning with the target model. By equipping each sub-model with a dedicated sub-model for different modalities, each sub-model can leverage its specific advantages in its feature processing modality, ensuring that data with strong representational capabilities is extracted from heterogeneous description information across different modalities. Since the judgment results for each modality are explicitly exposed, it helps to intuitively pinpoint the source of the problem. If the final judgment is flawed, it is possible to quickly identify which sub-model's unimodal attribute recognition issue is the cause.

[0037] For example, the server evaluates the quality of multiple descriptive information to obtain the data quality evaluation result. When the evaluation result of the descriptive information of a certain modality is less than a predetermined threshold, it can perform weight reduction or compensation processing based on the single modality attributes and confidence levels of other modalities to ensure the stability of the model under non-ideal data conditions.

[0038] For example, the above sub-model can be a multimodal model, or a deep learning-based artificial intelligence model, capable of simultaneously understanding and processing multiple types of data modalities, such as text, images, audio, video, structured data, etc., and establishing semantic associations and joint representations between different modalities to achieve cross-modal contextual understanding and reasoning.

[0039] In this embodiment of the application, the target item features and initial specifications of each descriptive information are input into the target model for processing to obtain the target logistics attributes of the target item and the inference information of the target logistics attributes. This may include the following steps: Input the target item characteristics and initial specifications corresponding to each description information into the target model for processing to obtain the initial logistics attributes of the target item; If the initial logistics attributes contradict any attribute among the single-modal attributes obtained from each sub-model, follow-up questions are generated based on the contradictory attributes. Based on the follow-up information and the corresponding descriptive information, each sub-model is used for processing to obtain the first attribute of each sub-model in response to the follow-up information; Based on the initial logistics attributes and the first attributes obtained from each sub-model, the target logistics attributes are determined.

[0040] In this embodiment, the target model receives the target item features and initial specifications corresponding to each modal description information. After processing, it generates the initial logistics attributes of the target item. The consistency of the initial logistics attributes and the single-modal attributes output by each sub-model is checked. If it is determined that the initial logistics attributes contradict any single-modal attribute, for example, the initial logistics attributes indicate that the item is fragile, while the single-modal attributes indicate that it is a metal product, the two attributes represent material characteristics of the same item that cannot coexist, which is considered a contradiction. Based on the contradictory attributes, targeted follow-up information is generated. The follow-up information and the corresponding modal description information are fed back to each sub-model. The sub-model combines the new follow-up information to perform secondary processing on the original description information and outputs a first attribute for the follow-up information. The initial logistics attributes and the first attributes generated by each sub-model can be determined by majority voting. As in the example above, the attribute that reaches the consensus and has the largest number of first attributes among the first number of sub-models is finally determined as the target physical attribute.

[0041] Due to potential biases or conflicts during multimodal information fusion, descriptions from any single information source may be incomplete or misinterpreted. By verifying the contradictions between initial logistics attributes and single-modal attributes, the system avoids directly adopting contradictory intermediate results that could distort the final attributes. Based on contradictory attributes, follow-up questions are generated and sub-models are called a second time. This allows for targeted supplementation of information on questionable contradictory parts of the attributes, effectively resolving attribute judgment contradictions caused by insufficient or ambiguous single-modal information and improving the accuracy of logistics attributes. This approach integrates comprehensive judgment of multimodal features and utilizes feedback from single-modal information in specific dimensions, ensuring that the output target logistics attributes closely match the actual characteristics of the goods, providing a reliable basis for determining logistics methods.

[0042] For example, the generation of follow-up questions is based on the contradictions between the initial logistics attributes and the unimodal attributes. It identifies the conflicting attributes of the contradictory attributes. Taking the contradiction where the initial logistics attribute indicates fragile goods and the unimodal attribute indicates metal products as an example, the conflicting attributes are the material of the item and its fragile nature. The follow-up questions will target this conflicting dimension, aiming to guide each sub-model to supplement information that can reconcile the contradiction. If the conflicting attribute is a binary attribute (either yes or no), the optional follow-up questions can be generated according to a predetermined prompt template, avoiding inefficient information acquisition caused by ambiguous follow-up questions. Through explicit guiding questions, each sub-model is guided to vote on the supplementary information, ensuring that the obtained first attribute is directly related to the resolution of the attribute contradiction, supporting the sub-model in generating an accurate first attribute.

[0043] For example, the consensus reached by the sub-models based on the first attribute follows a majority voting logic, avoiding misjudgments caused by single-modal description information. Furthermore, the process of reaching consensus among the sub-models based on the first attribute can be achieved by integrating majority voting logic with weighting coefficients. The server configures corresponding weighting coefficients for each sub-model, for example, based on the confidence level of the sub-model and the evaluation result of the information quality of the corresponding description information. The confidence level of the sub-model can be determined based on the model's accuracy in similar historical tasks, and the evaluation result of information quality can be determined based on the modality of the description information, such as the completeness of the text description, the clarity of the image, and the timeliness of the data.

[0044] During the voting phase, the first attribute output by each sub-model is first counted according to the majority rule. Optional weighting with their respective weight coefficients is then applied, and the voting result is determined based on the weighted composite score. For example, if two out of three sub-models support the fragile item attribute, with one high-weight sub-model supporting this attribute and one low-weight sub-model supporting it, while the remaining medium-weight sub-model supports non-fragile items, then the weighted composite score for fragile items will be higher, becoming the consensus result. The introduction of weight coefficients allows sub-models with high confidence and superior information quality to play a greater role in consensus formation, improving the accuracy of the voting results. By quantifying sub-model performance and information quality into weights, the excessive influence of single-modality random factors on the results is avoided, enhancing the reliability of the target logistics attribute.

[0045] For example, the conflicting attribute mentioned above is a non-binary attribute, that is, in the case of an attribute that needs to be quantified, such as the initial logistics attribute showing a weight range greater than a predetermined weight, and the single-modal attribute showing a weight of lightweight plastic products, the sample data can be automatically labeled, and the processing log of the sample can be generated based on the first attribute of each sub-model, follow-up information and the initial logistics attribute, in order to optimize the feature recognition capabilities of the target model and each sub-model, and the correction method of the quantified attribute is confirmed by human verification.

[0046] In the embodiments provided in this application, the target logistics attributes include modified specifications and logistics type, wherein the modified specifications are obtained by modifying the initial specifications, and the method further includes: Once the logistics type is determined to be a pre-order type, the corresponding deliverable area for the pre-order type item is obtained. Determine the logistics rights of the target item based on the revised specifications and the deliverable area; Send a first prompt message to the client. The first prompt message indicates at least one of the following: the deliverable area of ​​the target item, logistics rights, or logistics type.

[0047] In this embodiment, the server determines whether the logistics type in the target logistics attribute belongs to a predetermined type. The predetermined type can be, for example, fragile goods, dangerous goods, or items of special sizes. The predetermined type can be set by the server, the first client of the logistics delivery provider, or the second client of the logistics demander. If the logistics type is determined to be a predetermined type, the server further obtains the transportable area corresponding to that predetermined type of item. The mapping relationship between the predetermined type and the transportable area can be pre-set in the server. The transportable area can be a range determined based on logistics supervision requirements and transportation conditions to avoid transportation failures due to area compatibility issues.

[0048] By combining the revised specifications and the deliverable area, and based on preset logistics rights rules, the corresponding logistics rights for the target item are determined. These logistics rights include a commitment from the first client to implement preset conditions regarding logistics resource consumption and delivery time within the deliverable area, making the application of rights rules more aligned with the actual characteristics of the item and the transportation scenario. The server sends an initial notification message to the client, which includes at least one or more of the following: deliverable area, logistics rights, and logistics type, providing the logistics provider with key logistics-related information for the target item. This client can serve as either the first client for the logistics provider or the second client for the logistics requester, reducing user queries, optimizing the user experience, and enhancing the transparency and clarity of logistics services.

[0049] For example, the aforementioned modified specification can be a specific specification value determined based on the initial specification, or it can be a specification range, determining whether the weight or volume of the target item reaches a preset specification threshold, which is associated with logistics benefits. For instance, the aforementioned modified specification may not output the specific weight value of the target item, but it may indicate that the weight exceeds a predetermined weight threshold. If the weight exceeds the threshold, additional logistics resources need to be consumed, thereby determining the final available logistics benefits for the target item.

[0050] In related technologies, treating a composite item as a whole for judgment can easily lead to misjudgment due to the different attributes of its internal sub-items. For example, hot pot sets often include solid fuel, which is a flammable material, but the information displayed on the item itself will not contain information about solid fuel, thus leading to a misjudgment of the logistics type of the composite item. This embodiment, by deconstructing the composite and analyzing its components independently, can identify the logistics attributes of each sub-item.

[0051] In this embodiment of the application, feature extraction is performed on multiple descriptive information in step S102 to obtain the target item features of each descriptive information, which may include the following steps: Based on multiple descriptive information, the target item is identified as a composite item comprising multiple sub-items. Feature extraction is performed on the description information of sub-items to obtain the sub-item features corresponding to the sub-items, which are then used as the target item features.

[0052] In this embodiment, the server identifies whether the target item is a composite item composed of multiple sub-items based on acquired descriptive information in multiple different modalities (such as images containing overall and detailed breakdowns, text describing the composition of sub-items, structured data annotating each component, etc.). If it is determined to be a composite type including multiple sub-items, feature extraction is performed on the specific content of each sub-item in the descriptive information, such as partial image blocks of the sub-item in the display image, or individual descriptions of the sub-item in the text. The resulting sub-item features are used as target item features and input into the target model for processing. This reduces erroneous judgments caused by vague or inaccurate overall descriptive information and lowers the error rate in identifying the logistics attributes of composite items.

[0053] By identifying composite items and extracting the characteristics of sub-items, the single reliance on the overall characteristics of the item is avoided. The characteristics of each sub-component in the composite item can be identified. A composite item may include various situations such as one sub-item being fragile and another sub-item being liquid, providing fine-grained basis for the accurate determination of logistics attributes.

[0054] For example, by extracting the structured features of composite products, logistics can be broken down and packaging solutions customized. Different sub-items can be packaged according to their corresponding logistics attributes. For pre-made packages containing frozen and room temperature ingredients, it can be identified that frozen ingredients require cold chain transportation, while room temperature ingredients can be transported at room temperature, thus helping clients reduce cold chain transportation costs.

[0055] For non-combined target items, if the target logistics attributes of the same target item are multiple, the logistics method for that target item must meet the transportation requirements of all attributes. For combined target items, the target logistics attributes corresponding to each sub-item are used as the basis for subcontracting and transportation, allowing items with the same logistics attributes to be transported together.

[0056] In the embodiments provided in this application, different modalities include text modalities and image modalities. The descriptive information of the image modal includes the display image in the display page of the target item, and the descriptive information of the text modal includes the summary information of the display page, the comment information on the target item, or the category information of the target item.

[0057] In this embodiment, the different modalities covered by the multiple descriptive information of the target item may include text modality and image modality. The descriptive information of the image modality is taken from the display image on the target item's display page. The display image can always present the visual features of the item, such as its appearance, color, structural details, and material texture. The descriptive information of the text modality includes summary information, comment information, and category information from the display page. The summary information can be in the form of a title, used to provide a brief textual description of the item's functions and specifications. It can also be comment information on the target item, including user experience, actual size feedback, etc. The category information of the target item can be a pre-set category on the server, such as the item category, tags, and other structured text data.

[0058] The combination of textual and image modal information enables the perception of semantic descriptions and visual information such as material texture of target objects. Textual information can convey the abstract attributes of objects and feedback from actual use, while image information can supplement intuitive morphological features. The descriptive information of different modalities corroborates and complements each other, avoiding the limitations of single-modal information.

[0059] For example, descriptive information of the same modality can be extracted using different sub-models according to different information types. For instance, descriptive information of the same text modality, such as summary information and comment information, can use different sub-models. Even descriptive information of the same modality with different forms has different expression structures and different tendencies in description methods. For example, the form of summary information is relatively fixed, while comment information is generated by different commenters, with diversified description methods and richer semantic and emotional features. By processing different information types separately, the model's ability to generalize to different descriptive information can be improved.

[0060] In the embodiments provided in this application, feature extraction is performed on multiple descriptive information to obtain the target item features of each descriptive information, which may include the following steps: Once it is determined that the displayed image contains text information, text recognition is performed on the text information to obtain the first item feature; Based on the displayed image, the material of the target item is identified, and the second item feature is obtained; Based on the first item features and the second item features, the target item features corresponding to the image modality description information are obtained.

[0061] In the embodiments of this application, the server analyzes the displayed image of the target item. If it determines that the image contains text information, it extracts and parses the text information using Optical Character Recognition (OCR) technology to obtain the first item feature. The text information can take the form of markings on the item's surface, explanatory text on packaging, labels in the image, and item parameter information in image form. Based on the visual features of the displayed image, including texture, gloss, color distribution, and morphological structure, the material properties of the target item are identified using an image recognition algorithm to obtain the second item feature. The first item feature and the second item feature are then fused together to form the target item feature corresponding to the image modality description information.

[0062] By recognizing textual information in images, hidden semantic information is extracted from the images, enabling feature extraction of image modalities to go beyond visual morphology and also obtain some textual descriptive information. This achieves the synergistic utilization of textual semantics and visual morphological information in images, making the target object features corresponding to the image modalities more accurate and providing reliable visual modal data support for the reasoning of the target model.

[0063] Figure 2 A flowchart of a logistics processing method according to an embodiment of this application is shown, such as... Figure 2 As shown, the method may include steps S201, S202 and S203.

[0064] Step S201: Send the initial specifications of the target item to the server; Step S202: Receive and display the first prompt information and the inference information. The first prompt information is used to indicate the logistics method of the target item. Step S203: Based on the logistics method, perform logistics scheduling for the target item; Among them, the logistics method is determined based on the target logistics attributes of the target item, and the reasoning information represents the reasoning process by which the target model obtains the target logistics attributes based on the target item characteristics and initial specifications of the target item.

[0065] The aforementioned executing entity can be the first client used by the logistics and distribution provider. The first client sends the initial specifications of the target item to the server, providing basic data for the server's subsequent determination of logistics attributes. The first client receives initial prompt information and inference information from the server. The initial prompt information indicates the logistics method of the target item, which can be various methods such as cold chain transportation or fragile goods transportation. The inference information describes the reasoning process by which the target model derives the target logistics attributes based on the characteristics and initial specifications of the target item. Based on the logistics method in the initial prompt information, the logistics and distribution provider's first client can perform corresponding logistics scheduling operations on the target item. Logistics scheduling can optionally include allocating suitable special transport vehicles and arranging specialized packaging matching the logistics method. Executing scheduling based on the logistics method determined by the target logistics attributes makes logistics resource allocation more aligned with the actual needs of the item, reducing transportation losses and compliance risks, and improving the accuracy and reliability of logistics and distribution.

[0066] In the embodiments provided in this application, the first notification information is also used to indicate the logistics rights of the target item, and to perform logistics scheduling on the target item based on the logistics method, including: By comparing the preset method with the logistics method, and the preset benefits with the logistics benefits, the comparison results are obtained. The preset method and preset benefits are set by the second client. If the comparison results indicate that the preset method and the logistics method are different, and / or the preset rights and the logistics rights are different, a second prompt message is sent to the second client. The second prompt message is used to guide the second client to confirm the preset method and / or the preset rights based on the reasoning information. Receive confirmation information and, based on the logistics method and rights indicated in the confirmation information, execute logistics scheduling for the target items.

[0067] In this embodiment of the application, the first prompt information is also used to indicate the logistics rights of the target item. Step S203: Based on the logistics method, perform logistics scheduling for the target item, which may include the following steps: By comparing the preset method with the logistics method, and the preset benefits with the logistics benefits, the comparison results are obtained. The preset method and preset benefits are set by the second client. If the comparison results indicate that the preset method and the logistics method are different, and / or the preset rights and the logistics rights are different, a second prompt message is sent to the second client. The second prompt message is used to guide the second client to confirm the preset method and / or the preset rights based on the reasoning information. Receive confirmation information and, based on the logistics method and rights indicated in the confirmation information, execute logistics scheduling for the target items.

[0068] In this embodiment, the first client compares the logistics method and logistics rights determined by the server with the preset method and preset rights set by the second client of the logistics demander, generating a comparison result containing the differences between the two. If the comparison result shows a difference between the preset method and the logistics method, and / or a difference between the preset rights and the logistics rights, the first client sends a second prompt message to the second client. This second prompt message, combined with the reasoning information provided by the server, guides the second client to confirm the differences. Once the first client receives the confirmation message returned by the second client, the confirmation message may indicate agreement on the adjusted logistics method and rights, or agreement reached after proposing modifications. Based on the finally determined logistics method and logistics rights in the confirmation message, specific logistics scheduling operations are executed.

[0069] By comparing preset information with information determined by the server and triggering a confirmation mechanism, it ensures that the execution of logistics methods and rights is consistent with the expectations of the second client. The second prompt information, combined with reasoning information, guides the second client to understand the rationality of the adjustment of logistics methods and rights, making it easier for the second client to complete the confirmation process, reducing disputes caused by information asymmetry, and improving service transparency.

[0070] Based on the above embodiments and optional embodiments, this application also provides an optional implementation method, such as... Figure 3 As shown, multimodal descriptive information, including the target item's display image, title, category, reviews, and returned specifications (including weight), serves as multidimensional input for server-side processing. Joint inference is performed using a pre-trained multimodal model and an inference model on the server. The server sets up separate multimodal models as sub-models for feature extraction based on different descriptions. Specification information returned by the first client used by the logistics delivery party is input into the server's statistical model for processing, producing preliminary estimates of the target item's weight and volume parameters, which are then input into the inference model for further processing.

[0071] The aforementioned display image is input into the first sub-model for processing, which yields the visual features and text features displayed in the image, serving as the target item features. The item title, comment information, and category are then input into the second, third, and fourth sub-models for processing, respectively, yielding the semantic features of the title, comments, and category, which serve as the corresponding target item features.

[0072] The outputs of the first, second, third, and fourth sub-models, as well as the statistical model, are used inference models. These models integrate key information extracted from multiple multimodal models with weight and volume ranges from the statistical model to perform comprehensive reasoning. The resulting classification of the target item as fragile, sharp, virtual, or cold chain, along with corrected specifications such as estimated weight range and estimated volume range, are used as the target logistics attributes for the item. If conflicting attributes exist between different models, a majority vote can be used to remove conflicting attributes, with the majority-consistent logistics attribute becoming the final logistics attribute.

[0073] Compared to simply outputting labeled classification results, generative modeling is used to improve the interpretability of the model, directly outputting natural language descriptions, which facilitates manual review and system traceability. For example, the generated inference information is "Based on the image showing a sealed bag of liquid, and multiple comments mentioning 'refrigerated storage,' it is determined that cold chain transportation is required," to enhance the transparency of decision-making.

[0074] By adopting the above optional implementation method, the accuracy of logistics attribute identification is improved to over 92%, while related technologies can only achieve 65%-75% accuracy. Furthermore, since no pre-set knowledge base or mapping rules are required, it can cover all items, support automatic identification of low-frequency items and newly emerging items, improve the accuracy of item rights allocation for target items by 2%, and reduce the maintenance requirements of the second client of the logistics demand party to close to 0, greatly reducing the workload of manual maintenance and achieving fully automated attribute completion.

[0075] This paper applies a multimodal generative model to the automated generation of logistics attributes for goods. By fusing multi-dimensional information such as images, titles, categories, comments, and logistics feedback, it achieves end-to-end intelligent inference of key attributes such as whether the item is fragile, sharp, cold-chain, virtual, weight, and volume, breaking the dependence on a single data source. A joint judgment mechanism based on large-model common-sense reasoning and multimodal alignment is constructed. Utilizing the physical common sense (such as material properties and storage conditions) and cross-modal semantic understanding capabilities embedded in the large model, it achieves highly robust recognition of fuzzy descriptions, noisy data, and low-frequency items, improving the accuracy and generalization ability of attribute judgment. A generative structured output framework is proposed to output interpretable logistics attribute results in a natural language generation manner.

[0076] like Figure 4 As shown, corresponding to the application scenarios and methods provided in the embodiments of this application, the embodiments of this application also provide a logistics processing device, applied on a server side, including: The first acquisition module 401 is used to acquire multiple descriptive information of the target item and the initial specifications of the target item uploaded by the first client. The multiple descriptive information includes information of different modalities. The first feature extraction module 402 is used to extract features from multiple descriptive information to obtain the target item features of each descriptive information. The first reasoning module 403 is used to input the target item features and initial specifications of each descriptive information into the target model for processing, so as to obtain the target logistics attributes of the target item and the reasoning information on the target logistics attributes. Among them, the inference information represents the reasoning process by which the target model obtains the target logistics attributes based on the characteristics and initial specifications of the target item. The target logistics attributes are used to determine the logistics method of the target item. The target model is trained based on the predetermined reference item characteristics, reference logistics attributes, and reference inference information of the reference logistics attributes.

[0077] In the embodiments provided in this application, the first feature extraction module includes: The second feature extraction module uses each sub-model to extract features based on the corresponding descriptive information, obtaining the target item features corresponding to each descriptive information, as well as the single-modal attributes corresponding to the target item features. Each sub-model is used to process the modal features of the corresponding descriptive information. The single-modal attribute represents the single-modal logistics attribute of the target item determined based on the target item features of the corresponding descriptive information.

[0078] In the embodiments provided in this application, the first inference module includes: The second reasoning module is used to input the target item characteristics and initial specifications corresponding to each descriptive information into the target model for processing, so as to obtain the initial logistics attributes of the target item; The contradiction detection module is used to determine whether there is a contradiction between the initial logistics attributes and any attribute in the single-modal attributes obtained from each sub-model, and to generate follow-up questions based on the contradictory attributes. The follow-up questioning module is used to process the follow-up question information and the corresponding description information using each sub-model to obtain the first attribute of each sub-model in response to the follow-up question information. The voting module is used to determine the target logistics attribute based on the initial logistics attributes and the first attributes obtained from each sub-model.

[0079] In the embodiments provided in this application, the target logistics attributes include modified specifications and logistics type, the modified specifications are obtained by modifying the initial specifications, and the device further includes: The first determining module is used to determine that the logistics type indication is a pre-defined type and to obtain the transportable area corresponding to the pre-defined type of item. The second determination module is used to determine the logistics rights of the target item based on the revised specifications and the transportable area; The first sending module is used to send a first prompt message to the client. The first prompt message is used to indicate at least one of the following: the deliverable area of ​​the target item, logistics rights, or logistics type.

[0080] In the embodiments provided in this application, the first feature extraction module includes: The first identification module is used to identify, based on multiple descriptive information, the target item as a combination item including multiple sub-items; The second feature extraction module is used to extract features from the description information of sub-items to obtain the sub-item features corresponding to the sub-items, which are then used as the target item features.

[0081] In the embodiments provided in this application, the above-mentioned device is configured to have different modes, including text mode and image mode. The descriptive information of the image mode includes the display image in the display page of the target item, and the descriptive information of the text mode includes the summary information of the display page, the comment information on the target item, or the category information of the target item.

[0082] In the embodiments provided in this application, the first feature extraction module includes: The second recognition module is used to determine that the displayed image contains text information, perform text recognition on the text information, and obtain the first item feature; The third recognition module is used to identify the material of the target item based on the displayed image and obtain the second item features; The fusion module is used to obtain the target item features corresponding to the image modality description information based on the first item features and the second item features.

[0083] like Figure 5 As shown, corresponding to the application scenario and method provided in the embodiments of this application, the embodiments of this application also provide a logistics processing device, applied to a first client, including: The second sending module 501 is used to send the initial specifications of the target item to the server. Display module 502 is used to receive and display first prompt information and inference information, wherein the first prompt information is used to indicate the logistics method of the target item; The first scheduling module 503 is used to perform logistics scheduling on target items based on logistics methods; Among them, the logistics method is determined based on the target logistics attributes of the target item, and the reasoning information represents the reasoning process by which the target model obtains the target logistics attributes based on the target item characteristics and initial specifications of the target item.

[0084] In the embodiments provided in this application, the first prompt information is also used to indicate the logistics rights of the target item, and the aforementioned scheduling module includes: The comparison module is used to compare the preset method with the logistics method, as well as the preset benefits with the logistics benefits, and to obtain the comparison results. The preset method and preset benefits are set by the second client. The third sending module is used to send a second prompt message to the second client when the comparison result indicates that the preset method and the logistics method are different, and / or the preset rights and the logistics rights are different. The second prompt message is used to guide the second client to confirm the preset method and / or the preset rights based on the reasoning information. The second scheduling module is used to receive confirmation information and, based on the logistics method and logistics rights indicated in the confirmation information, to perform logistics scheduling for the target items.

[0085] The functions of each module in each device in the embodiments of this application can be found in the corresponding description in the above method, and they have corresponding beneficial effects, which will not be repeated here.

[0086] Figure 6 This is a block diagram of an electronic device used to implement embodiments of this application. For example... Figure 6 As shown, the electronic device includes a memory 601 and a processor 602. The memory 601 stores a computer program that can run on the processor 602. When the processor 602 executes the computer program, it implements the method described in the above embodiments. The number of memories 601 and processors 602 can be one or more. In a specific implementation, the electronic device may also include a communication interface 603 for communicating with external devices and exchanging data.

[0087] In practical implementation, if the memory 601, processor 602, and communication interface 603 are implemented independently, they can be interconnected via a bus to communicate with each other. This bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 6 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0088] Optionally, in a specific implementation, if the memory 601, processor 602 and communication interface 603 are integrated on a single chip, the memory 601, processor 602 and communication interface 603 can communicate with each other through an internal interface.

[0089] This application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method provided in this application.

[0090] This application provides a computer program product, including a computer program that, when executed by a processor, implements the method provided in this application.

[0091] This application also provides a chip including a processor for calling and executing instructions stored in a memory, causing a communication device with the chip installed to perform the method provided in this application.

[0092] This application also provides a chip, including: an input interface, an output interface, a processor, and a memory. The input interface, output interface, processor, and memory are connected through an internal connection path. The processor is used to execute code in the memory. When the code is executed, the processor is used to execute the method provided in the application embodiment.

[0093] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. General-purpose processors can be microprocessors or any conventional processor. It is worth noting that the processor can be a processor supporting Advanced Reduced Instruction Set Machines (ARM) architecture.

[0094] Further, optionally, the aforementioned memory may include read-only memory and random access memory. The memory may be volatile memory or non-volatile memory, or may include both. Non-volatile memory may include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which serves as an external cache. By way of example, but not limitation, many forms of RAM are available. Examples include Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Sync Link DRAM (SLDRAM), and Direct Rambus RAM (DR RAM).

[0095] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another.

[0096] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0097] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.

[0098] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process. Furthermore, the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functionality involved.

[0099] The logic and / or steps described in the flowchart or otherwise herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0100] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. All or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware, the program being stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiments.

[0101] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. This storage medium can be a read-only memory, a disk, or an optical disk, etc.

[0102] The above are merely exemplary embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope described in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A logistics processing method, characterized in that, Applied to the server side, including: Obtain multiple descriptive information of the target item and the initial specifications of the target item uploaded by the first client, wherein the multiple descriptive information includes information of different modalities; Feature extraction is performed on each of the multiple descriptive information entries to obtain the target item features of each of the descriptive information entries; Input the target item features and initial specifications of each of the aforementioned descriptive information into the target model for processing, and obtain the target logistics attributes of the target item and the inference information of the target logistics attributes; Wherein, the reasoning information represents the reasoning process by which the target model obtains the target logistics attributes based on the target item features and the initial specifications, the target logistics attributes are used to determine the logistics method of the target item, and the target model is trained based on predetermined reference item features, reference logistics attributes, and reference reasoning information of the reference logistics attributes.

2. The method according to claim 1, characterized in that, The step of extracting features from each of the multiple descriptive information pieces to obtain the target item features for each of the descriptive information pieces includes: Each sub-model extracts features based on the corresponding descriptive information to obtain the target item features corresponding to each descriptive information, and the single-modal attributes corresponding to the target item features; each sub-model is used to process the modal features of the corresponding descriptive information, and the single-modal attributes represent the single-modal logistics attributes of the target item determined based on the target item features of the corresponding descriptive information.

3. The method according to claim 2, characterized in that, The target item features and initial specifications described in the input description information are processed into the target model to obtain the target logistics attributes of the target item and inference information about the target logistics attributes, including: Input the target item features corresponding to each of the aforementioned description information and the initial specifications into the target model for processing to obtain the initial logistics attributes of the target item; If the initial logistics attributes are found to contradict any attribute among the single-modal attributes obtained from each of the sub-models, follow-up questions are generated based on the contradictory attributes. Based on the follow-up information and the corresponding description information, each of the sub-models is used for processing to obtain the first attribute of each sub-model for the follow-up information; Based on the initial logistics attributes and the first attributes obtained from each of the sub-models, the target logistics attributes are determined.

4. The method according to claim 1, characterized in that, The target logistics attributes include modified specifications and logistics type, wherein the modified specifications are obtained by modifying the initial specifications, and the method further includes: Once the logistics type indication is determined to be a predetermined type, the transportable area corresponding to the item of the predetermined type is obtained; Based on the revised specifications and the transportable area, determine the logistics rights of the target item; Send a first prompt message to the client, the first prompt message being used to indicate at least one of the following: the deliverable area of ​​the target item, the logistics rights, or the logistics type.

5. The method according to claim 1, characterized in that, The step of extracting features from each of the multiple descriptive information pieces to obtain the target item features for each of the descriptive information pieces includes: Based on the identification of multiple descriptive information, the target item is determined to be a combined item including multiple sub-items; Feature extraction is performed on the description information including the sub-items to obtain the sub-item features corresponding to the sub-items, which are used as the target item features.

6. The method according to any one of claims 1 to 5, characterized in that, The different modalities include text modalities and image modalities. The image modalities include the display images on the display page of the target item, and the text modalities include the summary information of the display page, the comments on the target item, or the category information of the target item.

7. The method according to claim 6, characterized in that, The step of extracting features from each of the multiple descriptive information pieces to obtain the target item features for each of the descriptive information pieces includes: The displayed image is determined to contain text information. The text information is then subjected to text recognition to obtain the first item feature. Based on the displayed image, the material of the target item is identified to obtain the second item feature; Based on the first item features and the second item features, the target item features corresponding to the description information of the image modality are obtained.

8. A logistics processing method, characterized in that, Applied to the first client, including: Send the initial specifications of the target item to the server; Receive and display a first prompt message and inference information, wherein the first prompt message is used to indicate the logistics method of the target item; Based on the aforementioned logistics method, logistics scheduling is performed on the target items; The logistics method is determined based on the target logistics attributes of the target item, and the reasoning information represents the reasoning process by which the target model obtains the target logistics attributes based on the target item characteristics and the initial specifications of the target item.

9. The method according to claim 8, characterized in that, The first prompt information is also used to indicate the logistics rights of the target item, and the step of performing logistics scheduling on the target item based on the logistics method includes: The preset method and the logistics method, as well as the preset benefits and the logistics benefits, are compared to obtain the comparison results. The preset method and the preset benefits are set by the second client. If the comparison result indicates that the preset method is different from the logistics method, and / or the preset rights are different from the logistics rights, a second prompt message is sent to the second client. The second prompt message is used to guide the second client to confirm the preset method and / or preset rights based on the reasoning information. Upon receiving confirmation information, and based on the logistics method and logistics rights indicated by the confirmation information, perform logistics scheduling for the target item.

10. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory, wherein the processor, when executing the computer program, implements the method of any one of claims 1 to 9.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 9.

12. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 9.

Images