Logistics exception item information processing method, device and equipment, and storage medium

Abstract

The present disclosure provides a logistics exception piece information processing method, device and equipment and a storage medium. The method comprises: obtaining text information and image information of a logistics exception piece; determining a keyword, an article name and an article category of the logistics exception piece based on the text information, and storing the keyword, the article name and the article category in a preset first database; performing feature extraction on the image information and the keyword to obtain feature data of the logistics exception piece, and storing the feature data of the logistics exception piece in a preset second database; receiving query information for the logistics exception piece, determining a first query result of the query information from the first database, and determining a second query result of the query information from the second database; and determining a logistics exception piece claim list corresponding to the query information based on the first query result and the second query result. Thus, the recall rate and the precision rate of the logistics exception piece can be considered to a greater extent.

Description

Technical Field

[0001] This invention relates to the field of logistics technology, and in particular to information processing methods, devices, equipment, and storage media for abnormal logistics items. Background Technology

[0002] During the entire logistics chain, express delivery information may be lost due to various reasons, preventing it from entering the next stage of the transportation chain and thus making the package an abnormal shipment. Common reasons include force majeure events such as wind, rain, and scratches from sharp objects.

[0003] In related technologies, the reporting of abnormal logistics shipments is commonly used to facilitate the subsequent claiming of such shipments. For example, using a mobile device, users fill in basic information about the abnormal shipment, click submit, and the information is stored in a database. When claiming the shipment, users can search for it using criteria such as the waybill number, item description, and reporting time to complete the claim process.

[0004] However, when using the above method to claim abnormal logistics items, sometimes the conditions are too broad, resulting in a massive number of results (information overload), and sometimes the conditions are too strict, making it easy to miss correct results (information loss). There is a problem that recall and precision are mutually exclusive. Summary of the Invention

[0005] To overcome the problems existing in related technologies, this disclosure provides a method, apparatus, equipment and storage medium for processing information on abnormal logistics items.

[0006] According to a first aspect of the present disclosure, a method for processing information on abnormal logistics shipments is provided, the method comprising: Obtain text and image information of abnormal shipments; Based on text information, determine the keywords, item names, and item categories of abnormal logistics items, and store the keywords, item names, and item categories in a preset first database; Feature extraction is performed on image information and keywords to obtain feature data of abnormal logistics items, and the feature data of abnormal logistics items is stored in a preset second database; Receive query information for abnormal shipments, determine the first query result from the first database, and determine the second query result from the second database; Based on the first and second query results, determine the logistics anomaly claim list corresponding to the query information.

[0007] In some optional implementations, keywords, item names, and item categories of abnormal shipments are determined based on text information, including: The text information is normalized by a large language model to obtain the processed text. The normalization process is used to eliminate typos and colloquial expressions in the text information. Based on the processed text, the keywords, item names, and item categories of the abnormal logistics shipments were determined; and The item name is determined in the following way: Based on the processed text, the target item name of the abnormal logistics item is determined, as well as the semantic information of the item name. Based on the semantic information of the items, the names of the target items are expanded with synonyms within the scope of the item categories to obtain a set of item names for abnormal logistics items.

[0008] In some optional implementations, feature extraction is performed on image information and keywords to obtain feature data of abnormal logistics items, including: Keywords, item names, and image information are combined to obtain the first combined result; The first concatenation result is subjected to cross-modal feature fusion and alignment using a multimodal vector representation model to obtain the first feature data; The image information is used to extract features through a visual feature extraction model to obtain the second feature data; The first feature data and the second feature data are determined as the feature data of the abnormal logistics item.

[0009] In some alternative implementations, the method further includes: Obtain scalar information on abnormal logistics items, wherein the scalar information includes at least one of the following: weight, dimensions, and location of discovery; Keywords, item names, and item categories are stored in a pre-defined first database, including: Keywords, item names, item categories, and scalar information are stored in a pre-defined first database in a structured data format; and The characteristic data of abnormal logistics items are stored in a pre-set second database, including: The characteristic data and scalar information of abnormal logistics items are stored in a pre-set second database in an unstructured data format.

[0010] In some optional implementations, based on the first query result and the second query result, a list of logistics anomaly claims corresponding to the query information is determined, including: If the query information contains scalar query conditions, the first query result and the second query result are deduplicated and merged based on the unique identifier of the abnormal logistics item to obtain the merged result; The results are filtered and merged based on scalar query conditions to obtain the list of logistics anomaly claims corresponding to the query information.

[0011] In some optional implementations, determining a first query result from a first database includes: If the query information contains query text, then determine whether the cache contains keywords and item names that match the query information; If the cache does not contain keywords and item names that match the query information, then determine the keywords and item names that match the query information; and based on the keywords and item names that match the query information, determine the first query result of the query information from the first database; If the cache contains keywords and item names that match the query information, then based on the keywords and item names in the cache that match the query information, the first query result for the query information is determined from the first database.

[0012] In some optional implementations, determining a second query result from a second database includes: If the query information contains both a query image and query text, then the query image and query text are concatenated to obtain a second concatenated result. Feature extraction is performed on the second splicing result to obtain the third feature data; Determine the similarity between the feature data of abnormal logistics items stored in the second database and the third feature data to obtain the first similarity. Feature extraction is performed on the query image to obtain the fourth feature data; Determine the similarity between the feature data of the abnormal logistics items stored in the second database and the fourth feature data to obtain the second similarity. Based on the first similarity and the second similarity, a second query result is determined from the second database to retrieve the query information.

[0013] According to a second aspect of the present disclosure, an information processing apparatus for abnormal logistics shipments is provided, the apparatus comprising: The first acquisition unit is configured to acquire text and image information of abnormal logistics items; The first determining unit is configured to: determine the keywords, item name and item category of the abnormal logistics item based on text information, and store the keywords, item name and item category in a preset first database; The feature extraction unit is configured to: extract features from image information and keywords to obtain feature data of abnormal logistics items, and store the feature data of abnormal logistics items in a preset second database; The second determining unit is configured to: receive query information for abnormal logistics items, determine a first query result of the query information from a first database, and determine a second query result of the query information from a second database; The third determining unit is configured to: determine the logistics anomaly claim list corresponding to the query information based on the first query result and the second query result.

[0014] In some optional implementations, keywords, item names, and item categories of abnormal shipments are determined based on text information, including: The text information is normalized by a large language model to obtain the processed text. The normalization process is used to eliminate typos and colloquial expressions in the text information. Based on the processed text, the keywords, item names, and item categories of the abnormal logistics shipments were determined; and The item name is determined in the following way: Based on the processed text, the target item name of the abnormal logistics item is determined, as well as the semantic information of the item name. Based on the semantic information of the items, the names of the target items are expanded with synonyms within the scope of the item categories to obtain a set of item names for abnormal logistics items.

[0015] In some optional implementations, feature extraction is performed on image information and keywords to obtain feature data of abnormal logistics items, including: Keywords, item names, and image information are combined to obtain the first combined result; The first concatenation result is subjected to cross-modal feature fusion and alignment using a multimodal vector representation model to obtain the first feature data; The image information is used to extract features through a visual feature extraction model to obtain the second feature data; The first feature data and the second feature data are determined as the feature data of the abnormal logistics item.

[0016] In some alternative embodiments, the device further includes: The second acquisition unit is configured to acquire scalar information of the abnormal logistics item, wherein the scalar information includes at least one of the following: weight, size, and location of discovery; Keywords, item names, and item categories are stored in a pre-defined first database, including: Keywords, item names, item categories, and scalar information are stored in a pre-defined first database in a structured data format; and The characteristic data of abnormal logistics items are stored in a pre-set second database, including: The characteristic data and scalar information of abnormal logistics items are stored in a pre-set second database in an unstructured data format.

[0017] In some optional implementations, based on the first query result and the second query result, a list of logistics anomaly claims corresponding to the query information is determined, including: If the query information contains scalar query conditions, the first query result and the second query result are deduplicated and merged based on the unique identifier of the abnormal logistics item to obtain the merged result; The results are filtered and merged based on scalar query conditions to obtain the list of logistics anomaly claims corresponding to the query information.

[0018] In some optional implementations, determining a first query result from a first database includes: If the query information contains query text, then determine whether the cache contains keywords and item names that match the query information; If the cache does not contain keywords and item names that match the query information, then determine the keywords and item names that match the query information; and based on the keywords and item names that match the query information, determine the first query result of the query information from the first database; If the cache contains keywords and item names that match the query information, then based on the keywords and item names in the cache that match the query information, the first query result for the query information is determined from the first database.

[0019] In some optional implementations, determining a second query result from a second database includes: If the query information contains both a query image and query text, then the query image and query text are concatenated to obtain a second concatenated result. Feature extraction is performed on the second splicing result to obtain the third feature data; Determine the similarity between the feature data of abnormal logistics items stored in the second database and the third feature data to obtain the first similarity. Feature extraction is performed on the query image to obtain the fourth feature data; Determine the similarity between the feature data of the abnormal logistics items stored in the second database and the fourth feature data to obtain the second similarity. Based on the first similarity and the second similarity, a second query result is determined from the second database to retrieve the query information.

[0020] According to a third aspect of the present disclosure, an electronic device is provided, comprising: Memory, used to store computer programs; A processor is configured to execute a computer program stored in the aforementioned memory, and when the aforementioned computer program is executed, to implement the method of any embodiment of the information processing method for logistics anomalies of the first aspect of this disclosure.

[0021] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the method of any embodiment of the information processing method for logistics anomalies of the first aspect described above.

[0022] According to a fifth aspect of the present disclosure, a computer program product is provided, including computer-readable code, which, when executed by a processor, implements any embodiment of the information processing method for logistics anomalies described in the first aspect.

[0023] The technical solutions provided by the embodiments of this disclosure may include the following beneficial effects: In this embodiment, text and image information of abnormal logistics shipments are acquired. Then, based on the text information, keywords, item names, and item categories of the abnormal logistics shipments are determined and stored in a preset first database. Next, feature extraction is performed on the image information and keywords to obtain feature data of the abnormal logistics shipments, which is then stored in a preset second database. Subsequently, query information for the abnormal logistics shipments is received. From the first database, a first query result is determined, enabling precise matching based on keywords, item names, and item categories. From the second database, a second query result is determined, enabling fuzzy matching based on feature data. Finally, based on the first and second query results, a list of claimed abnormal logistics shipments corresponding to the query information is determined. This approach combines the advantages of precise matching and fuzzy matching, improving both the comprehensiveness and accuracy of abnormal logistics shipment queries, thus balancing recall and precision to a greater extent.

[0024] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0025] The accompanying drawings, which are incorporated in and form part of this disclosure, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0026] Figure 1 This disclosure is a flowchart illustrating an information processing method for logistics anomalies according to an exemplary embodiment.

[0027] Figure 2 This is a flowchart illustrating another method for processing information on logistics anomalies according to an exemplary embodiment of this disclosure.

[0028] Figure 3AThis disclosure is a flowchart illustrating information on reporting abnormal logistics items according to an exemplary embodiment.

[0029] Figure 3B This disclosure is a flowchart illustrating the process of claiming a logistics anomaly according to an exemplary embodiment.

[0030] Figure 4 This disclosure is a block diagram of an information processing apparatus for logistics anomalies according to an exemplary embodiment.

[0031] Figure 5 This disclosure illustrates a block diagram for an electronic device according to an exemplary embodiment. Detailed Implementation

[0032] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0033] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The singular forms “a,” “the,” and “the” as used in this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0034] It should be understood that although the terms first, second, third, etc., may be used in this disclosure to describe various information, such information should not be limited to these terms. These terms are used only to distinguish information of the same type from one another. For example, without departing from the scope of this disclosure, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0035] The acquisition and processing of data (such as text information, image information, and scalar information of abnormal logistics shipments) in this disclosure comply with relevant laws, regulations, social ethics, and provisions concerning public interests. For example, the acquisition, storage, and processing of data in this disclosure are all carried out with the knowledge and consent of the users and have obtained the full authorization of the relevant users.

[0036] The embodiments of this disclosure will now be described in detail.

[0037] Figure 1 This is a flowchart illustrating an information processing method for abnormal logistics items according to an exemplary embodiment. This method can be applied to one or more electronic devices such as smartphones, laptops, desktop computers, portable computers, and servers. Furthermore, the execution entity of this method can be hardware or software. When the execution entity is hardware, it can be one or more of the aforementioned electronic devices. For example, a single electronic device can execute this method, or multiple electronic devices can cooperate with each other to execute this method. When the execution entity is software, this method can be implemented as multiple software programs or software modules, or as a single software program or software module. No specific limitations are made herein.

[0038] like Figure 1 As shown, the method specifically includes: Step 101: Obtain text and image information of the abnormal shipment.

[0039] In this embodiment, a logistics anomaly can be: a package that, during the operation of the logistics chain, has its waybill damaged or lost due to factors such as wind, rain, or scratches from sharp objects, resulting in the loss of express delivery information and preventing it from entering the next transportation stage.

[0040] Text information can be related information about abnormal shipments in text form. As an example, text information may include, but is not limited to, at least one of the following: item description, remaining shipping label information, etc.

[0041] Image information can be images of abnormal shipments or image-related information. As an example, image information may include, but is not limited to, at least one of the following: an image of the abnormal shipment, an image of the abnormal shipment's outer packaging, the storage address of the image of the shipment, the storage address of the image of the outer packaging, feature data of the image of the shipment, and feature data of the image of the outer packaging.

[0042] In some optional implementations, delivery personnel or those responsible for abnormal shipments can use mobile devices (such as computers or smartphones) to manually fill in the item description of the abnormal shipment on the reporting interface and take a picture of the abnormal shipment. Thus, the aforementioned entity can use the item description as the text information of the abnormal shipment and the photographed image as the image information of the abnormal shipment.

[0043] Step 102: Based on the text information, determine the keywords, item name, and item category of the abnormal logistics shipment, and store the keywords, item name, and item category in a preset first database.

[0044] In this embodiment, keywords can be core product attribute information extracted from text information. As an example, keywords may include, but are not limited to, at least one of the following: brand, color, material, style, size, core content words, and descriptive words.

[0045] The item name can be the name of a shipment with logistics irregularities. As an example, the item name can be a common name, colloquial expression, or common abbreviation for a shipment with logistics irregularities, as well as synonyms for the common name, colloquial expression, or common abbreviation.

[0046] The item category can be the result of classifying abnormal logistics items. As an example, the item category can include at least one of the following: clothing and footwear (such as tops, pants, shoes, hats, etc.), electronic products (such as mobile phones, computers, tablets, headphones, etc.), household appliances (such as refrigerators, washing machines, air conditioners, microwave ovens, etc.), home furnishings (such as bedside tables, coffee tables, sofas, lamps, etc.), food and beverages (such as snacks, drinks, grains and oils, etc.), beauty and personal care products (such as skin care products, cosmetics, toiletries, etc.), and books and audio-visual products (such as books, CDs, stationery, etc.).

[0047] The primary database can be a database used to store keywords, item names, and item categories. Furthermore, the primary database can be a database used to store structured data or a database used to store unstructured data. As examples, the primary database could be an Elasticsearch database, a MySQL database, a Redis database, etc.

[0048] In some optional implementations, a large language model can be used to determine the keywords, item names, and item categories of abnormal shipments based on text information. Then, the keywords, item names, and item categories of the abnormal shipments are associated and stored in a pre-defined first database.

[0049] In some optional implementations of this embodiment, the keywords, item names, and item categories of abnormal logistics items can also be determined based on text information in the following way: The first step is to normalize the text information using a large language model to obtain the processed text. Normalization is used to eliminate typos and colloquial expressions in the text.

[0050] The aforementioned large language model can be a general-purpose large language model. This large language model can perform normalization processing on text information, that is, eliminate typos and colloquial expressions in the text information, thereby obtaining processed text free of typos and colloquial expressions.

[0051] The processed text can be the result of text processing after removing typos and colloquial expressions from the text information.

[0052] The second step is to determine the keywords, item names, and item categories of the abnormal shipments based on the processed text.

[0053] In some optional implementations, keywords corresponding to the processed text can be determined through a large language model, thereby obtaining the keywords of the abnormal shipment; the item name corresponding to the processed text can be determined through a large language model, thereby obtaining the item name of the abnormal shipment; and the item category corresponding to the processed text can be determined through a large language model, thereby obtaining the item category of the abnormal shipment.

[0054] Based on this, the item name can be determined in the following way: The first step is to determine the target item name of the abnormal logistics shipment based on the processed text, as well as the semantic information of the target item name.

[0055] The target item name can be an item name determined based on the processed text. As an example, the target item name can be an item name contained within the processed text. In some alternative implementations, the target item name can also be referred to as the core product name.

[0056] Item semantic information can be the semantic context related to the name of the target item.

[0057] In some alternative implementations, the processed text can be identified using a large language model to obtain the target item name of the logistics anomaly, as well as the item semantic information that determines the target item name through the large language model.

[0058] The second step is to expand the target item name with synonyms within the item category based on the semantic information of the item to obtain a set of item names for abnormal logistics items.

[0059] The item name set can contain multiple item names. For example, the item name set can include colloquial terms, informal expressions, and common abbreviations for items with logistical issues. However, it's important to note that all item names in the item name set must belong to the same category (i.e., product category). In other words, cross-category terms are not allowed in the item name set, and cross-category synonym expansion is prohibited.

[0060] It is understandable that standardizing text information through a large language model can eliminate noise (such as typos and colloquial expressions) in the original text. Based on the processed text, keywords, item names, and item categories for abnormal logistics shipments can be determined, improving the accuracy of these identifications. By determining the target item name and its semantic information based on the processed text, and then expanding the target item name with synonyms within the item category based on this semantic information, a set of item names for abnormal logistics shipments can be obtained. This effectively covers different expressions (synonyms, colloquialisms, etc.) that users might use during queries without introducing cross-category interference, thereby improving the recall rate of item name-based retrieval.

[0061] Step 103: Extract features from image information and keywords to obtain feature data of abnormal logistics items, and store the feature data of abnormal logistics items in a preset second database.

[0062] In this embodiment, the feature data of the abnormal logistics shipment can be the result of feature extraction from image information and keywords. In some optional implementations, the feature data of the abnormal logistics shipment can be represented in vector form.

[0063] The second database can be a database used to store characteristic data of abnormal shipments. Furthermore, the second database can be a database used to store structured data or a database used to store unstructured data. As examples, the second database could be a Milvus vector database, a Pinecone database, a Weaviate database, etc.

[0064] In some optional implementations, image feature data can be generated by extracting features from image information using a visual feature extraction model. This visual feature extraction model may include, but is not limited to, at least one of the following: Scale-Invariant Feature Transform (SIFT) model, Speeded-Up Robust Features (SURF) extraction model, etc. Furthermore, text feature data can be obtained by extracting features from keywords using a keyword feature extraction model. This keyword feature extraction model may include, but is not limited to, at least one of the following: Convolutional Neural Network (CNN), Rapid Automatic Keyword Extraction (RAKE) model, etc. Subsequently, the image feature data and text feature data can be fused (e.g., concatenated) to obtain the feature data of abnormal logistics shipments.

[0065] In some optional implementations, image information and keywords can be concatenated and then input into a multimodal vector representation model for cross-modal feature fusion and alignment to generate a unified multimodal feature vector, thereby obtaining the feature data of the abnormal logistics shipment. Alternatively, keywords and item names can be concatenated, then concatenated with image information according to a preset format, and the concatenated result can be input into a multimodal vector representation model for cross-modal feature fusion and alignment to generate a unified multimodal feature vector, thereby obtaining the feature data of the abnormal logistics shipment.

[0066] In some optional implementations, the feature data of abnormal logistics shipments may include at least one of the following: image feature data, text feature data, and multimodal feature vectors. For example, the feature data of abnormal logistics shipments may include image feature data and multimodal feature vectors, thereby allowing the image feature data and multimodal feature vectors to be associated and stored in a pre-defined second database.

[0067] In some optional implementations of this embodiment, feature extraction can be performed on image information and keywords to obtain feature data of abnormal logistics items: The first step is to combine keywords, item names, and image information to obtain the first combined result.

[0068] The first splicing result can be obtained by splicing keywords, item names, and image information.

[0069] In some optional implementations, keywords and item names can be connected by commas to obtain a keyword list; and the keyword list and image information can be concatenated according to a preset format (e.g., keyword list + preset image placeholder + image URL (Uniform Resource Locator) + end marker) to obtain the first concatenation result.

[0070] The second step involves performing cross-modal feature fusion and alignment on the first splicing result using a multimodal vector representation model to obtain the first feature data.

[0071] The multimodal vector representation model can be a deep learning model used to simultaneously process and understand multiple data modalities (such as text, images, etc.). As an example, a deep learning model that simultaneously processes and understands multiple data modalities can include, but is not limited to, at least one of the following: Vision-Language Multi-modal Model (VLMo), Visual Language BERT (Bidirectional Encoder Representations from Transformers) model, etc.

[0072] The first feature data can be multimodal feature data generated by a multimodal vector representation model. In some optional implementations, the first feature data can be represented by vectors.

[0073] The third step is to extract features from the image information using a visual feature extraction model to obtain the second feature data.

[0074] The second feature data can be pure image feature data generated by a pure visual feature extraction model. In some optional implementations, the second feature data can be represented as a vector.

[0075] The fourth step is to determine the first and second feature data as the feature data of the abnormal logistics item.

[0076] It is understandable that by concatenating keywords, item names, and image information, and then using a multimodal vector representation model to perform cross-modal feature fusion and alignment on the first concatenation result to obtain the first feature data, a feature representation that integrates textual and image semantic information can be generated. By using a visual feature extraction model to extract features from the image information to obtain the second feature data, only visual appearance features can be retained. By identifying the first and second feature data as the feature data of abnormal logistics items, feature data containing joint textual and image semantics and pure visual information can be constructed. Thus, in subsequent multimodal retrieval, it is possible to understand textual and image semantics while capturing subtle visual differences, thereby improving retrieval accuracy in complex scenarios.

[0077] Step 104: Receive query information for abnormal logistics shipments, determine the first query result of the query information from the first database, and determine the second query result of the query information from the second database.

[0078] In this embodiment, the query information can be used to locate abnormal shipments. In some optional implementations, the query information can be entered by a user or other object on the claim interface.

[0079] In some optional implementations, the query information may include at least one of the following: scalar information, text information, image information, etc.

[0080] The first query result can be the result retrieved from the first database.

[0081] The second query result can be the result retrieved from the second database.

[0082] In some optional implementations, the query information entered by the user regarding the abnormal logistics item can be obtained on the claim interface. Then, a precise match is performed on keywords, item name, and item category in a first database to obtain the first query result. Finally, a similarity search is performed on feature data in a second database to determine the second query result.

[0083] In some optional implementations, the first query result and the second query result may each contain information related to multiple abnormal logistics items.

[0084] In some optional implementations of this embodiment, the first query result of the query information can be determined from the first database in the following manner: The first step is to determine whether the cache contains keywords and item names that match the query information if the query information contains query text.

[0085] The query text can be the text information contained in the query information.

[0086] Here, if the query information contains query text, it can be determined whether the cache of the aforementioned executing entity (e.g., the server) contains keywords and item names that match the query information.

[0087] The cache contains keywords and item names that match the query information. This usually means that the same query text has been queried and processed before, and the results have been saved so that the same query text can be responded to quickly the next time, thus improving query efficiency.

[0088] In some optional implementations, the query information input by users or other objects can first be normalized to obtain processed query information. Normalization is used to eliminate typos and colloquial expressions in the query information input by users or other objects. Next, it is determined whether the processed query information contains query text. If the processed query information contains query text, it is then determined whether the cache contains keywords and item names that match the processed query information.

[0089] The second step is to determine the keywords and item names that match the query information if the cache does not contain such keywords and item names; and to determine the first query result of the query information from the first database based on the keywords and item names that match the query information.

[0090] The keywords that match the query information can be keywords from the query text. The item names that match the query information can be item names from the query text.

[0091] In some optional implementations, an exact match (i.e., the keywords and item names are the same) can be performed in a first database based on keywords and item names that match the query information to obtain the first query result. For example, keywords matching the query information can be retrieved from the first database to obtain the first result corresponding to multiple abnormal logistics shipments, and item names matching the query information can be retrieved from the first database to obtain the second result corresponding to multiple abnormal logistics shipments. Then, the sets (e.g., the union) of the multiple abnormal logistics shipments corresponding to the first and second results are used as the first query result.

[0092] The third step is to determine the first query result of the query information from the first database based on the keywords and item names in the cache that match the query information.

[0093] Here, if the cache contains keywords and item names that match the query information, the first query result of the query information can be determined directly from the first database based on the keywords and item names in the cache that match the query information.

[0094] It is understandable that when the query information contains the query text, it can be determined whether the cache contains keywords and item names that match the query information. If the cache does not contain keywords and item names that match the query information, then by determining the keywords and item names that match the query information, and based on the keywords and item names, the first query result of the query information can be determined from the first database, which can ensure the accuracy and semantic coverage of the retrieval. If the cache already contains keywords and item names that match the query information, then the first query result of the query information can be directly determined from the first database based on the results in the cache, without having to determine the keywords and item names that match the query information through large models or other methods. Thus, the text retrieval process can be accelerated, the system's computational overhead can be reduced, and the response speed of user queries can be improved.

[0095] In some optional implementations of this embodiment, the second query result of the query information can be determined from the second database in the following manner: The first step is to concatenate the query image and query text if the query information includes both a query image and query text, thus obtaining the second concatenation result.

[0096] The second concatenation result can be obtained by concatenating the query image and the query text.

[0097] In some optional implementations, the query information input by users or other objects can first be normalized to obtain processed query information. Normalization is used to eliminate typos and colloquial expressions in the query information input by users or other objects. Next, it is determined whether the processed query information contains query images and query text. If the processed query information contains query images and query text, the query images and query text in the processed query information are concatenated to obtain a second concatenated result.

[0098] The query text can be the text information contained in the query information.

[0099] The image to be queried can be image information contained in the query information, such as image features.

[0100] In some optional implementations, the query image and query text can be concatenated according to a preset format to obtain a second concatenated result.

[0101] The second step is to extract features from the second splicing result to obtain the third feature data.

[0102] The third feature data can be obtained by extracting features from the second concatenation result. In some optional implementations, the third feature data can be represented as a vector.

[0103] The third step is to determine the similarity between the feature data of the abnormal logistics items stored in the second database and the third feature data to obtain the first similarity.

[0104] The first similarity can represent the similarity between the feature data of the abnormal logistics items stored in the second database and the third feature data.

[0105] In some alternative implementations, for each abnormal logistics shipment stored in the second database, the similarity between the feature data of the abnormal logistics shipment and the third feature data (e.g., at least one of Euclidean distance, cosine similarity, and Manhattan distance) can be calculated to obtain the first similarity corresponding to the abnormal logistics shipment.

[0106] The fourth step is to extract features from the query image to obtain the fourth feature data.

[0107] The fourth feature data can be the result of feature extraction from the query image. In some optional implementations, the fourth feature data can be represented as a vector.

[0108] The fifth step is to determine the similarity between the feature data of the abnormal logistics items stored in the second database and the fourth feature data to obtain the second similarity.

[0109] The second similarity can be the similarity between the feature data of the abnormal logistics items stored in the second database and the fourth feature data.

[0110] In some optional implementations, for each abnormal shipment stored in the second database, the similarity between the feature data of the abnormal shipment and the fourth feature data (e.g., at least one of Euclidean distance, cosine similarity, and Manhattan distance) is calculated to obtain the second similarity corresponding to the abnormal shipment.

[0111] The sixth step is to determine the second query result of the query information from the second database based on the first similarity and the second similarity.

[0112] Here, we can identify the logistics anomaly items in the second database that have a high first similarity (e.g., the first similarity is greater than a preset first threshold, or a preset first number of first similarities selected in descending order), thus obtaining a first set of logistics anomaly items. We then identify the logistics anomaly items in the second database that have a high second similarity (e.g., the second similarity is greater than a preset second threshold, or a preset second number of second similarities selected in descending order), thus obtaining a second set of logistics anomaly items. The union of the first and second sets of logistics anomaly items is input into a multimodal re-ranking model (e.g., a Rerank model) to obtain a sequence of logistics anomaly items, which is then used as the second query result for the query information.

[0113] It is understandable that if the query information includes a query image and query text, then by concatenating the query image and query text, a second concatenation result is obtained. Feature extraction is then performed on the second concatenation result to obtain third feature data, which can generate query data that integrates the semantics of the query image and text. By determining the similarity between the feature data of the abnormal logistics items stored in the second database and the third feature data, candidate results that are semantically similar to the query information can be recalled. By extracting features from the query image to obtain fourth feature data, and determining the similarity between the feature data of the abnormal logistics items stored in the second database and the fourth feature data, candidate results that are visually similar to the query can be recalled. By determining the second query result of the query information from the second database based on the first and second similarities, a second query result that balances semantic relevance and visual fidelity can be obtained by fusing joint semantic similarity and pure visual similarity.

[0114] Step 105: Based on the first query result and the second query result, determine the logistics anomaly claim list corresponding to the query information.

[0115] In this embodiment, the logistics anomaly claim list can be a list of logistics anomalies that are available for claiming and displayed to the user.

[0116] In some optional implementations, the union of multiple abnormal logistics items corresponding to the first query result and the second query result can be sorted to obtain the logistics abnormal item claim list corresponding to the query information.

[0117] In some alternative implementations, the following method can be used to sort the union of multiple logistics anomaly items corresponding to the first and second query results respectively: The first step involves calculating the score for each abnormal logistics item in the union of multiple abnormal logistics items corresponding to the first and second query results. This score is based on the item's target similarity, the duration between its reporting time and the current time, and the number of abnormal logistics items reported within the target area (e.g., a circular area with the reporting location as the center and a preset length as the radius). Specifically, if the first similarity for the abnormal logistics item has been calculated but the second similarity has not, then the first similarity is determined as the target similarity. If the second similarity has been calculated but the first similarity has not, then the second similarity is determined as the target similarity. If both the first and second similarities have been calculated, then the sum of the first and second similarities is determined as the target similarity.

[0118] As an example, the following formula can be used to calculate the score of the abnormal logistics shipment based on the target similarity Sim, the duration Δt between the reporting time and the current time, and the number N_region of abnormal logistics shipments reported within the target area where the reporting location is located: Score=w1×Sim+w2×exp(-|Δt-T0|)+w3×[1-(logN_region / logN_max)]; Where exp() represents the abbreviation for exponential function, which means exponential operation with the natural constant e as the base; T0 represents the preset base duration (e.g., 1 week), and its unit is consistent with Δt, for example, its unit can be weeks; N_max represents the actual maximum number of reports in all regions of the system, or a set threshold (e.g., 100); the bases of the two logs can be any positive number other than 1, and the bases of the two logs can be equal; w1, w2, and w3 are preset weight values, which are positive.

[0119] The second step is to sort the abnormal logistics items in the above-mentioned set according to their scores from highest to lowest, thus obtaining a list of abnormal logistics items to be claimed corresponding to the query information. [Agent 1] It should be noted that, where there is no conflict, the technical features described in different alternative implementations can be included in the same embodiment. For the sake of brevity, they will not be elaborated here.

[0120] Based on the embodiments of this disclosure, text and image information of abnormal logistics shipments are obtained. Then, based on the text information, keywords, item names, and item categories of the abnormal logistics shipments are determined, and these keywords, item names, and item categories are stored in a preset first database. Next, feature extraction is performed on the image information and keywords to obtain feature data of the abnormal logistics shipments, and this feature data is stored in a preset second database. Subsequently, query information for the abnormal logistics shipments is received. From the first database, a first query result for the query information is determined, which can achieve precise matching based on keywords, item names, and item categories. From the second database, a second query result for the query information is determined, which can achieve fuzzy matching based on feature data. Finally, based on the first and second query results, a list of abnormal logistics shipments to be claimed corresponding to the query information is determined. This approach combines the advantages of precise matching and fuzzy matching, improving both the comprehensiveness and accuracy of abnormal logistics shipment queries, thereby balancing recall and precision to a greater extent.

[0121] Figure 2 This is a flowchart illustrating another method for processing information on logistics anomalies according to an exemplary embodiment of this disclosure. For example... Figure 2 As shown, the method specifically includes: Step 201: Obtain text information, image information, and scalar information of the abnormal logistics item. The scalar information includes at least one of the following: weight, dimensions, and location of discovery.

[0122] In this embodiment, scalar information can represent the attributes of logistics anomalies quantified by numerical values ​​or enumeration values.

[0123] In some optional implementations, users and other entities can fill in basic information about abnormal logistics items on the reporting interface, such as the location where the abnormal logistics item was discovered, the quantity of the item, weight, and other scalar information.

[0124] In addition, step 201 and Figure 1 Step 101 in the corresponding embodiment is basically the same, and will not be repeated here.

[0125] Step 202: Based on the text information, determine the keywords, item names, and item categories of the abnormal logistics items, and store the keywords, item names, item categories, and scalar information in a structured data format in a preset first database.

[0126] In this embodiment, the structured data can be data stored according to a preset format. As an example, each piece of data in the structured data can have the same fields and data type.

[0127] Here, keywords, item names, and item categories can be used as keys to determine the values ​​of keywords, item names, and item categories from text information. This allows the specific content of keywords, item names, item categories, and scalar information for each logistics anomaly to be stored as a record in a pre-defined first database in a structured data format.

[0128] In addition, step 202 and Figure 1 Step 102 in the corresponding embodiment is basically the same, and will not be repeated here.

[0129] Step 203: Extract features from image information and keywords to obtain feature data of abnormal logistics items, and store the feature data and scalar information of abnormal logistics items in a pre-set second database in an unstructured data format.

[0130] In this embodiment, the format of each piece of unstructured data can be different.

[0131] Here, the characteristic data and scalar information of abnormal logistics items can be linked and stored in a pre-set second database in an unstructured data format.

[0132] In addition, step 203 and Figure 1 Step 103 in the corresponding embodiment is basically the same, and will not be repeated here.

[0133] Step 204: Receive query information for abnormal logistics shipments, determine the first query result of the query information from the first database, and determine the second query result of the query information from the second database.

[0134] In this embodiment, step 204 and Figure 1 Step 104 in the corresponding embodiment is basically the same, and will not be repeated here.

[0135] Step 205: Based on the first query result and the second query result, determine the logistics anomaly claim list corresponding to the query information.

[0136] In this embodiment, step 205 and Figure 1 Step 105 in the corresponding embodiment is basically the same, and will not be repeated here.

[0137] In some optional implementations of this embodiment, the logistics anomaly claim list corresponding to the query information can be determined in the following manner, based on the first query result and the second query result: The first step is to merge the first and second query results based on the unique identifier of the abnormal logistics item, if the query information contains scalar query conditions, to obtain the merged result.

[0138] Among them, scalar query conditions can be constraints on scalar information entered by users or other objects when querying abnormal logistics items.

[0139] Unique identifiers can be used to uniquely identify items with logistics issues. Different items with logistics issues can be assigned different unique identifiers.

[0140] The merged result can be obtained by deduplicating and merging the first query result and the second query result.

[0141] Here, if the first query result and the second query result contain logistics anomalies represented by the same unique identifier, it means that the first query result and the second query result contain the same logistics anomaly. Therefore, duplicate logistics anomalies can be removed, and the first query result and the second query result can be merged so that the logistics anomalies contained in the merged result are different.

[0142] The second step is to filter and merge the results based on the scalar query conditions to obtain the logistics anomaly claim list corresponding to the query information.

[0143] Here, scalar query conditions can be used to filter each abnormal logistics item in the merged result to obtain a filtered result that meets the scalar query conditions, and the list of each abnormal logistics item in the filtered result can be used as the logistics abnormal item claim list corresponding to the query information.

[0144] It is understandable that if the query information contains scalar query conditions, then by deduplicating and merging the first and second query results based on the unique identifier of the abnormal logistics item, duplicate results from different databases can be eliminated, avoiding the display of redundant information to the user. By filtering and merging the results based on scalar query conditions, candidate results that do not meet the scalar query conditions specified by the user (such as weight or discovery location) can be eliminated, thereby obtaining a list of abnormal logistics items that integrates multi-source search results and meets hard constraints.

[0145] It should be noted that, in addition to the contents described above, this embodiment may also include... Figure 1 The corresponding technical features described in the corresponding embodiments, thereby achieving Figure 1 For details on the technical effectiveness of the information processing method for abnormal logistics shipments shown, please refer to [link / reference needed]. Figure 1 The relevant descriptions are presented concisely and will not be elaborated upon here.

[0146] Based on the embodiments of this disclosure, by acquiring the scalar information of abnormal logistics items and storing keywords, item names, item categories, and scalar information in a structured data format in a preset first database, keywords, item names, item categories, and scalar information can be used as strong constraints to achieve more efficient and accurate matching and filtering. By storing the feature data and scalar information of abnormal logistics items in an unstructured data format in a preset second database, the feature data and scalar information of abnormal logistics items can be expressed through feature data, thereby ensuring that the final search results satisfy both semantic or visual similarity and user-specified constraints.

[0147] The following uses an example of a missing item in logistics to illustrate the embodiments of this disclosure. However, it should be noted that the following content is only used to understand the technical solutions of the embodiments of this disclosure and does not constitute a limitation on the protection scope of the embodiments of this disclosure.

[0148] As express delivery operates through the entire logistics chain, force majeure events such as wind, rain, or scratches from sharp objects can cause the waybill affixed to the surface of the package to be damaged or even lost. This results in the loss of express delivery information, preventing the package from entering the next stage of the transportation chain. These packages are called undeliverable packages.

[0149] When handling missing items, staff generally follow two steps: reporting missing items and claiming missing items.

[0150] The general process for reporting missing items is as follows: After entering the missing item reporting interface using a mobile device, first fill in the basic information of the missing item (such as the type of missing item, the location where it was found, the quantity of the item, weight, and other scalar information); if there is any remaining label information, manually identify and manually enter the remaining label number, three-segment code, etc.; then manually fill in the item description of the missing item, take a photo of the missing item, click submit, and store it in the database.

[0151] The general process for claiming unclaimed items is as follows: Using a mobile device, users can access the unclaimed item claim interface, search by tracking number, item description, reporting time, etc., manually compare the items with the images and descriptions displayed on the interface, and finally complete the claim.

[0152] Several factors affect the efficiency of claiming lost items during the actual handling of lost items by staff: 1. Unstructured information such as item name and brand needs to be manually entered by branch staff, which is cumbersome and inefficient for reporting missing items.

[0153] 2. The query relies heavily on manually constructed query conditions. Broad conditions result in a massive amount of results (information overload), while strict conditions are prone to missing correct results (information loss). Furthermore, the text descriptions are greatly influenced by personal experience, making it unfriendly to beginners.

[0154] 3. The search and claim process is time-consuming, with an average claim time of up to 125 hours (about 5 days), which seriously affects timeliness assessment and customer satisfaction.

[0155] 4. Low claim efficiency leads to a large number of packages being converted into claims or abandoned, resulting in direct economic losses.

[0156] The inefficiency of handling unclaimed packages is a common pain point in the current express delivery industry. The market urgently needs an intelligent solution to reduce reliance on manual labor while balancing recognition accuracy and coverage. Specifically, the following market demands exist: 1. Urgent need for cost reduction, efficiency improvement and asset recovery: It is necessary to significantly shorten the claim cycle, increase the average daily claim volume, reduce claims expenditures, and realize the asset recovery of stranded packages.

[0157] 2. Need for intelligent parsing and standardized input of unstructured information: Manual input of item descriptions is extremely inefficient and prone to errors. There is an urgent need to introduce large model technology to achieve automatic error correction, key information extraction and structured standardization of colloquial item descriptions, and achieve "second-level reporting".

[0158] 3. The need for precise and fuzzy semantic retrieval: Item descriptions have problems such as synonyms and confusion between hierarchical concepts. The system needs to be able to understand semantics and make reasonable expansions, and also make precise constraints based on specific fields to avoid cross-category interference; at the same time, it needs to take into account the precise filtering of scalar attributes (such as weight and time).

[0159] Currently, mainstream products and related technical solutions on the market can be mainly divided into the following categories, but all of them have obvious limitations: Traditional logistics information management systems (CRUD systems based on relational databases): rely on SQL (Structured Query Language) for exact matching or fuzzy queries, cannot understand semantics, and recall and precision are mutually exclusive; semantic retrieval systems that introduce traditional NLP (Natural Language Processing): lack professional knowledge in the express delivery field, and feature extraction is inaccurate; it is difficult to achieve multimodal alignment, and cannot map "package appearance image" and "item description text" to the same vector space; and have poor generalization ability when dealing with long-tail item descriptions.

[0160] In summary, the current market urgently needs a new generation of undeliverable processing system that can break through the limitations of a single modality, possess professional semantic understanding capabilities in the express delivery field, and achieve "cross-modal addressability and precise filtering".

[0161] This solution provides a system for reporting and claiming unclaimed express parcels based on large model feature extraction and multimodal fusion, which mainly consists of two processes: reporting unclaimed parcels and claiming unclaimed parcels.

[0162] See Figure 3A The missing item reporting process is as follows: When a branch staff member discovers a missing item and reports it, the system obtains the scalar information (attributes of the missing item that can be quantified with specific numerical values, such as weight, size, and location of discovery), text information (including manually entered residual shipping label information and item description), and image information (images of the contents and outer packaging), and then proceeds to the following processing branches: 1. Text correction and structured branching: 1) Text correction: The general language model is used to automatically correct colloquial and misspelled text in the user's input of item descriptions; 2) Hierarchical Feature Extraction: The corrected text is structurally extracted, outputting three parts of features: Keyword extraction: Extract core attribute information of products, including brand, color, material, style, size, core content words and descriptive words, etc. Product name (i.e., the name of the above-mentioned items) recognition: accurately identify the core product name, and reasonably expand synonyms, colloquial names, spoken expressions and common abbreviations based on the semantics of the product, without introducing cross-category words; Product categorization: Categorize items according to general e-commerce industry classification standards to obtain item names. If multiple categories are involved, all categories must be listed.

[0163] 3) ES Database Storage: The extracted keywords, product names, classification results, and other structured data, along with the scalar information input by the user, are stored in the ES (Elasticsearch) database (i.e., the first database mentioned above) for subsequent precise matching retrieval and scalar filtering.

[0164] 2. Text + Image Multimodal Feature Extraction and Vectorization Branch: 1) Error correction + text simplification: Connect the "keywords" obtained from error correction and extraction in step 1 with the "product name (i.e. the above item name)" with a comma to simplify it into a keyword list; and concatenate them according to the preset multimodal interleaving format (e.g.: text keywords + preset image placeholders + image URL + end marker) to generate a multimodal input sequence (i.e. the above first concatenation result). 2) Image model vectorization: Use a pure visual feature extraction model (image model) to extract features from the uploaded objectless image and generate a pure image feature vector (i.e., the second feature data mentioned above). 3) Text + Image Multimodal Model Vectorization: The generated multimodal input sequence is input into the multimodal vector representation model to perform cross-modal feature fusion and alignment, generating a unified multimodal feature vector (i.e., the first feature data mentioned above). 4) Milvus database storage: The generated pure image feature vectors and multimodal feature vectors, along with the scalar information input by the user, are stored in the Milvus vector database (i.e., the second database mentioned above) for subsequent cross-modal semantic similarity retrieval and scalar filtering.

[0165] See Figure 3B The process for claiming unclaimed items is as follows: When a staff member or customer attempts to claim an unclaimed item on the claim interface, the system obtains the user's input query conditions (including scalar conditions, text queries, and / or image queries), then enters the corresponding search branch based on the input modality, and finally converges on the unclaimed item list: 1. Text search branch (for text search criteria): 1) Cache detection: The system first determines whether the input text query conditions hit the cache, that is, whether the cache contains keywords and item names that match the query information; 2) If the cache is not hit: Call the large model to perform text correction, accurate classification (classify words into goods_name / goods_category / other_words), and restricted synonym expansion (execute a strictly controlled expansion strategy according to the category, prohibiting cross-category expansion), and generate classification and synonym results; 3) If the cache is hit: directly retrieve the category and synonym results corresponding to the text; 4) Database Retrieval: Based on classification and synonym results, hierarchical targeted mapping matching retrieval is performed in the Elasticsearch database. Specifically, if the vocabulary category is "goods_category," its synonyms are matched against the product category field in Elasticsearch using absolute string exact matching; if it's "goods_name," it matches against the product name field; and if it's "other_words," it matches against the keyword field. Simultaneously, scalar conditions are used for filtering to obtain the text retrieval results.

[0166] 2. Text + Image Search Branch (for text + image search criteria): 1) Text processing: Align with the reporting process, call the large model to perform "error correction + text simplification" on the query text, and generate a list of query keywords; 2) Image model vectorization and recall: The query image is vectorized using an image model, and similarity retrieval is performed in the Milvus database (i.e., the second database mentioned above) to recall visually similar candidate data. 3) Vectorization and Recall of Text + Image Multimodal Model: The multimodal input sequence of “concise query text + query image” is jointly vectorized using a multimodal vector representation model. Similarity retrieval is performed in the Milvus database (i.e. the second database mentioned above) to recall candidate data that are visually and semantically similar. 4) Multimodal Rerank Model Re-ranking: The candidate results retrieved by Milvus search are input into the multimodal re-ranking model, and the results are refined and scored by combining multi-dimensional features of images and text to improve the accuracy.

[0167] 3. Scalar Conditions and Result Aggregation Branch: The text retrieval branch (ES database retrieval results, i.e., the first query result), the multimodal retrieval branch (Milvus results rearranged by the Rerank model, i.e., the second query result), and the scalar conditions input by the user (i.e., scalar query conditions) are aggregated; the multi-source results are deduplicated and merged based on the unique identifier (unclaimed item ID), and finally filtered in combination with the scalar conditions to generate the final unclaimed item list (i.e., the logistics abnormal item claim list) and display it to the user.

[0168] It should be noted that, in addition to the contents described above, this embodiment may also include the technical features described in the above embodiments, thereby achieving the technical effect of the information processing method for abnormal logistics items shown above. Please refer to the above description for details. For the sake of brevity, it will not be elaborated here.

[0169] Based on the embodiments of this disclosure, a preprocessing method for unclaimed multimodal data for cross-modal retrieval alignment is proposed. In the reporting and claiming stages, a large language model is introduced to extract a keyword list from colloquial long text through "error correction + simplification," and this list is vectorized by jointly inputting it with images according to a preset multimodal interleaving format. This solves the problem of image-text alignment offset caused by redundant context in logistics scenarios, significantly improving the hit rate of cross-modal vector retrieval. Based on a precise retrieval method using large-model restricted synonym expansion and hierarchical directional mapping, a cascaded processing logic of "error correction - boundary classification - restricted synonym expansion" is designed for unstructured logistics text, and this process is accelerated through a caching mechanism. During claiming retrieval, the synonym set is directionally mapped to the corresponding field in the ES database according to the category of the query term for absolute string precision matching. This achieves both semantic-level query expansion and avoids cross-category interference through field-level precise matching, balancing recall and precision. An intelligent claiming architecture is adopted, integrating ES structured hierarchical precise retrieval and Milvus multimodal vector similarity retrieval. This architecture parallelizes ES precise matching based on hierarchical mapping and Milvus multimodal vector retrieval based on high-dimensional semantic space. The ES engine ensures precise filtering of strong features, while the Milvus engine compensates for the fuzzy matching capability of weak features, achieving a complementarity between precision and fuzziness. A multimodal retrieval mechanism based on weighted calculation of pure visual features and image-text fusion features, along with Rerank fine-tuning, is employed. Through dual-track vectorized recall using image models and text + image multimodal models, combined with multimodal Rerank model fine-tuning, the weights of visual and semantic features can be more effectively balanced, thereby improving precision in complex long-tail scenarios. A convergence and filtering mechanism using multi-source heterogeneous data and scalar conditions allows for three-dimensional convergence of ES text retrieval results, Milvus multimodal rearrangement results, and user-input scalar conditions. Deduplication is based on unique identifiers, and final filtering is performed using scalar conditions, ensuring that the claiming results meet both semantic / visual similarity and the hard constraints of physical attributes (such as weight and time). By using a large model-driven method for automatic error correction and standardized classification of unstructured express delivery text, and leveraging the semantic understanding capabilities of the large model, the system can replace manual work in correcting item descriptions, extracting keywords, expanding product names, and mapping categories. This enables "second-level entry" for reporting undelivered items, reducing the operational threshold for branch staff.

[0170] Figure 4 This is a block diagram illustrating an information processing apparatus for logistics anomalies according to an exemplary embodiment of the present disclosure. The apparatus specifically includes: The first acquisition unit 401 is configured to acquire text information and image information of abnormal logistics items; The first determining unit 402 is configured to: determine the keywords, item name and item category of the abnormal logistics item based on text information, and store the keywords, item name and item category in a preset first database; The feature extraction unit 403 is configured to: extract features from image information and keywords to obtain feature data of abnormal logistics items, and store the feature data of abnormal logistics items in a preset second database; The second determining unit 404 is configured to: receive query information for abnormal logistics items, determine a first query result of the query information from a first database, and determine a second query result of the query information from a second database; The third determining unit 405 is configured to: determine the logistics abnormal item claim list corresponding to the query information based on the first query result and the second query result.

[0171] In some optional implementations, keywords, item names, and item categories of abnormal shipments are determined based on text information, including: The text information is normalized by a large language model to obtain the processed text. The normalization process is used to eliminate typos and colloquial expressions in the text information. Based on the processed text, the keywords, item names, and item categories of the abnormal logistics shipments were determined; and The item name is determined in the following way: Based on the processed text, the target item name of the abnormal logistics item is determined, as well as the semantic information of the item name. Based on the semantic information of the items, the names of the target items are expanded with synonyms within the scope of the item categories to obtain a set of item names for abnormal logistics items.

[0172] In some optional implementations, feature extraction is performed on image information and keywords to obtain feature data of abnormal logistics items, including: Keywords, item names, and image information are combined to obtain the first combined result; The first concatenation result is subjected to cross-modal feature fusion and alignment using a multimodal vector representation model to obtain the first feature data; The image information is used to extract features through a visual feature extraction model to obtain the second feature data; The first feature data and the second feature data are determined as the feature data of the abnormal logistics item.

[0173] In some alternative embodiments, the device further includes: The second acquisition unit (not shown in the figure) is configured to acquire scalar information of the abnormal logistics item, wherein the scalar information includes at least one of the following: weight, size, and location of discovery; Keywords, item names, and item categories are stored in a pre-defined first database, including: Keywords, item names, item categories, and scalar information are stored in a pre-defined first database in a structured data format; and The characteristic data of abnormal logistics items are stored in a pre-set second database, including: The characteristic data and scalar information of abnormal logistics items are stored in a pre-set second database in an unstructured data format.

[0174] In some optional implementations, based on the first query result and the second query result, a list of logistics anomaly claims corresponding to the query information is determined, including: If the query information contains scalar query conditions, the first query result and the second query result are deduplicated and merged based on the unique identifier of the abnormal logistics item to obtain the merged result; The results are filtered and merged based on scalar query conditions to obtain the list of logistics anomaly claims corresponding to the query information.

[0175] In some optional implementations, determining a first query result from a first database includes: If the query information contains query text, then determine whether the cache contains keywords and item names that match the query information; If the cache does not contain keywords and item names that match the query information, then determine the keywords and item names that match the query information; and based on the keywords and item names that match the query information, determine the first query result of the query information from the first database; If the cache contains keywords and item names that match the query information, then based on the keywords and item names in the cache that match the query information, the first query result for the query information is determined from the first database.

[0176] In some optional implementations, determining a second query result from a second database includes: If the query information contains both a query image and query text, then the query image and query text are concatenated to obtain a second concatenated result. Feature extraction is performed on the second splicing result to obtain the third feature data; Determine the similarity between the feature data of abnormal logistics items stored in the second database and the third feature data to obtain the first similarity. Feature extraction is performed on the query image to obtain the fourth feature data; Determine the similarity between the feature data of the abnormal logistics items stored in the second database and the fourth feature data to obtain the second similarity. Based on the first similarity and the second similarity, a second query result is determined from the second database. The information processing device for abnormal logistics items provided in this embodiment can execute the corresponding steps of the information processing methods for abnormal logistics items described above, thereby achieving the technical effects of the information processing methods for abnormal logistics items described above. The information processing device and the information processing methods for abnormal logistics items can refer to and cite each other in specific implementation and technical effects, but for the sake of brevity, they will not be elaborated here.

[0177] Figure 5 This disclosure illustrates a block diagram for an electronic device according to an exemplary embodiment. Figure 5 The illustrated electronic device 500 includes at least one processor 501, a memory 502, at least one network interface 504, and a user interface 503. The various components in the electronic device 500 are coupled together via a bus system 505. It is understood that the bus system 505 is used to implement communication between these components. In addition to a data bus, the bus system 505 also includes a power bus, a control bus, and a status signal bus. However, for clarity, ... Figure 5 The general designated all buses as Bus System 505.

[0178] The user interface 503 may include a display, keyboard, or clicking device (e.g., mouse, trackball, touchpad, or touchscreen).

[0179] It is understood that the memory 502 in this embodiment of the present disclosure may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory may be 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. The volatile memory may be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), and Direct Rambus RAM (DRRAM). The memory 502 described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0180] In some implementations, memory 502 stores elements, executable units or data structures, or subsets thereof, or extended sets thereof: operating system 5021 and application program 5022.

[0181] The operating system 5021 includes various system programs, such as the framework layer, core library layer, and driver layer, used to implement various basic business functions and handle hardware-based tasks. The application program 5022 includes various applications, such as a media player and a browser, used to implement various application functions. The program implementing the method of this embodiment can be included in the application program 5022.

[0182] In this embodiment, by calling the program or instructions stored in memory 502, specifically the program or instructions stored in application program 5022, processor 501 executes the method steps provided in each method embodiment, including, for example: Obtain text and image information of abnormal shipments; Based on text information, determine the keywords, item names, and item categories of abnormal logistics items, and store the keywords, item names, and item categories in a preset first database; Feature extraction is performed on image information and keywords to obtain feature data of abnormal logistics items, and the feature data of abnormal logistics items is stored in a preset second database; Receive query information for abnormal shipments, determine the first query result from the first database, and determine the second query result from the second database; Based on the first and second query results, determine the logistics anomaly claim list corresponding to the query information.

[0183] The methods disclosed in the above embodiments of this disclosure can be applied to or implemented by processor 501. Processor 501 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware or by instructions in the form of software in processor 501. The processor 501 may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this disclosure. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this disclosure can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software units in the decoding processor. The software units may be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 502. Processor 501 reads the information in memory 502 and, in conjunction with its hardware, completes the steps of the above method.

[0184] It is understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described above, or combinations thereof.

[0185] For software implementation, the techniques described herein can be implemented by units that perform the functions described above. The software code can be stored in memory and executed by a processor. The memory can be implemented within the processor or external to the processor.

[0186] The electronic device provided in this embodiment may be as follows: Figure 5 The electronic device shown can execute all the steps of the information processing methods for each abnormal logistics item described above, thereby achieving the technical effects of the information processing methods for each abnormal logistics item described above. For details, please refer to the relevant descriptions above. For the sake of brevity, further details are not provided here.

[0187] This disclosure also provides a storage medium (computer-readable storage medium). This storage medium stores one or more programs. The storage medium may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as read-only memory, flash memory, hard disk, or solid-state drive; the memory may also include combinations of the above types of memory.

[0188] When one or more programs in the storage medium can be executed by one or more processors to implement the above-mentioned information processing method for abnormal logistics items executed on the electronic device side.

[0189] The processor described above is used to execute the information processing program for abnormal logistics items stored in the memory, so as to implement the following steps of the information processing method for abnormal logistics items executed on the electronic device side: Obtain text and image information of abnormal shipments; Based on text information, determine the keywords, item names, and item categories of abnormal logistics items, and store the keywords, item names, and item categories in a preset first database; Feature extraction is performed on image information and keywords to obtain feature data of abnormal logistics items, and the feature data of abnormal logistics items is stored in a preset second database; Receive query information for abnormal shipments, determine the first query result from the first database, and determine the second query result from the second database; Based on the first and second query results, determine the logistics anomaly claim list corresponding to the query information.

[0190] Furthermore, the computer program product provided in this disclosure embodiment may include computer-readable code that, when executed on a device, causes a processor in the device to implement the steps of the following information processing method for logistics anomalies executed on the electronic device side: Obtain text and image information of abnormal shipments; Based on text information, determine the keywords, item names, and item categories of abnormal logistics items, and store the keywords, item names, and item categories in a preset first database; Feature extraction is performed on image information and keywords to obtain feature data of abnormal logistics items, and the feature data of abnormal logistics items is stored in a preset second database; Receive query information for abnormal shipments, determine the first query result from the first database, and determine the second query result from the second database; Based on the first and second query results, determine the logistics anomaly claim list corresponding to the query information.

[0191] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this disclosure.

[0192] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0193] It should be understood that the terminology used herein is for the purpose of describing particular exemplary embodiments only and is not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “described” as used herein may also include the plural forms. The terms “comprising,” “including,” “containing,” and “having” are inclusive and therefore indicate the presence of the stated features, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, and / or combinations thereof. The method steps, processes, and operations described herein are not construed as requiring them to be performed in a particular order described or illustrated unless the order of performance is explicitly indicated. It should also be understood that additional or alternative steps may be used.

[0194] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for processing information on abnormal logistics shipments, characterized in that, The method includes: Obtain text and image information of abnormal shipments; Based on the text information, the keywords, item names, and item categories of the abnormal logistics shipments are determined, and the keywords, item names, and item categories are stored in a preset first database; Feature extraction is performed on the image information and the keywords to obtain feature data of the abnormal logistics shipment, and the feature data of the abnormal logistics shipment is stored in a preset second database; Receive query information for abnormal logistics shipments, determine a first query result for the query information from the first database, and determine a second query result for the query information from the second database; Based on the first query result and the second query result, a logistics anomaly claim list corresponding to the query information is determined.

2. The method according to claim 1, characterized in that, The process of determining the keywords, item name, and item category of the abnormal shipment based on the text information includes: The text information is normalized using a large language model to obtain processed text, wherein the normalization process is used to eliminate typos and colloquial expressions in the text information. Based on the processed text, the keywords, item names, and item categories of the abnormal logistics shipments are determined; and The name of the item is determined in the following manner: Based on the processed text, the target item name of the abnormal logistics item is determined, as well as the item semantic information of the target item name is determined; Based on the semantic information of the items, the names of the target items are expanded using synonyms within the scope of the item category to obtain a set of item names for abnormal logistics items.

3. The method according to claim 1, characterized in that, The step of extracting features from the image information and the keywords to obtain feature data for abnormal logistics items includes: The keywords, the item name, and the image information are concatenated to obtain a first concatenation result; The first concatenation result is subjected to cross-modal feature fusion and alignment using a multimodal vector representation model to obtain the first feature data; The image information is used to extract features using a visual feature extraction model to obtain second feature data; The first feature data and the second feature data are determined as feature data of abnormal logistics items.

4. The method according to claim 1, characterized in that, The method further includes: Obtain scalar information of abnormal logistics items, wherein the scalar information includes at least one of the following: weight, dimensions, and location of discovery; The step of storing the keywords, the item name, and the item category into a preset first database includes: The keywords, item names, item categories, and scalar information are stored in a pre-defined first database in a structured data format; and The step of storing the characteristic data of abnormal logistics items into a preset second database includes: The characteristic data of the abnormal logistics items and the scalar information are stored in a preset second database in an unstructured data format.

5. The method according to claim 4, characterized in that, The step of determining the logistics anomaly claim list corresponding to the query information based on the first query result and the second query result includes: If the query information contains scalar query conditions, then based on the unique identifier of the abnormal logistics item, the first query result and the second query result are deduplicated and merged to obtain a merged result; The merged results are filtered based on the scalar query conditions to obtain the logistics anomaly claim list corresponding to the query information.

6. The method according to any one of claims 1-5, characterized in that, Determining the first query result of the query information from the first database includes: If the query information contains query text, then determine whether the cache contains keywords and item names that match the query information; If the cache does not contain keywords and item names that match the query information, then keywords and item names that match the query information are determined; and based on the keywords and item names that match the query information, a first query result of the query information is determined from the first database. If the cache contains keywords and item names that match the query information, then based on the keywords and item names in the cache that match the query information, a first query result for the query information is determined from the first database.

7. The method according to any one of claims 1-5, characterized in that, The step of determining the second query result of the query information from the second database includes: If the query information includes a query image and query text, then the query image and the query text are concatenated to obtain a second concatenation result; Feature extraction is performed on the second splicing result to obtain the third feature data; Determine the similarity between the feature data of the abnormal logistics items stored in the second database and the third feature data to obtain a first similarity. Feature extraction is performed on the query image to obtain fourth feature data; Determine the similarity between the feature data of the abnormal logistics items stored in the second database and the fourth feature data to obtain a second similarity. Based on the first similarity and the second similarity, a second query result of the query information is determined from the second database.

8. An information processing device for abnormal logistics items, characterized in that, The device includes: The first acquisition unit is configured to acquire text and image information of abnormal logistics items; The first determining unit is configured to: determine the keywords, item name and item category of the abnormal logistics item based on the text information, and store the keywords, item name and item category in a preset first database; The feature extraction unit is configured to: extract features from the image information and the keywords to obtain feature data of the abnormal logistics shipment, and store the feature data of the abnormal logistics shipment in a preset second database; The second determining unit is configured to: receive query information for abnormal logistics shipments, determine a first query result of the query information from the first database, and determine a second query result of the query information from the second database; The third determining unit is configured to: determine the logistics abnormal item claim list corresponding to the query information based on the first query result and the second query result.

9. An electronic device, comprising: Memory, used to store computer programs; A processor for executing a computer program stored in the memory, wherein when the computer program is executed, it implements the method described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1-7.

Images