Object processing method, electronic device, storage medium, and program product

Abstract

This application provides an object processing method, electronic device, storage medium, and program product, relating to the fields of artificial intelligence technology and data processing. The method includes: acquiring contextual information of an object to be processed, wherein the object to be processed represents an object exhibiting anomalies, and the contextual information represents information related to the anomaly; performing anomaly identification on the object to be processed based on the contextual information to obtain anomaly identification results, wherein the anomaly identification results reflect the cause of the anomaly; parsing the anomaly identification results to construct object processing information for the object to be processed; and processing the object to be processed based on the object processing information. This application solves the technical problem of low efficiency in inspecting goods exhibiting anomalies in related technologies.

Description

Technical Field

[0001] This application relates to artificial intelligence technology and data processing, and more specifically, to an object processing method, electronic device, storage medium, and program product. Background Technology

[0002] In current platform operation and governance practices, to ensure the authenticity, compliance, and consumer rights of product information, processing systems typically need to continuously inspect a massive number of products on the platform to identify violations such as "inaccurate descriptions," "false advertising," "lack of qualifications," and "invalid evidence." However, traditional product inspection methods mainly rely on manual sampling and rule-engine-driven keyword matching mechanisms. While these methods can identify anomalies to some extent, their efficiency in inspecting products is generally low due to limitations in manpower and cognitive load, or restrictions on keywords and fixed fields.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This application provides an object processing method, electronic device, storage medium, and program product to at least solve the technical problem of low efficiency in inspecting abnormal goods in related technologies.

[0005] According to one aspect of the embodiments of this application, an object processing method is provided, comprising: obtaining context information of an object to be processed, wherein the object to be processed is used to characterize an object with an anomaly, and the context information is used to characterize information that is associated with the anomaly; performing anomaly identification on the object to be processed based on the context information to obtain anomaly identification result, wherein the anomaly identification result is used to reflect the reason for the anomaly; parsing the anomaly identification result to construct object processing information of the object to be processed; and processing the object to be processed based on the object processing information.

[0006] According to one aspect of the embodiments of this application, another object processing method is also provided, including: responding to an input instruction applied to an operation interface, displaying context information of an object to be processed on the operation interface, wherein the object to be processed is used to characterize an object with an anomaly, and the context information is used to characterize information related to the anomaly; responding to a processing instruction applied to the operation interface, displaying object processing information on the operation interface, wherein the object to be processed is processed, and the object processing information is obtained by parsing anomaly identification results, the anomaly identification results are obtained by identifying anomalies in the object to be processed through the context information, and the anomaly identification results are used to reflect the reason for the existence of the anomaly.

[0007] According to another aspect of the embodiments of this application, an object processing apparatus is also provided, comprising: an information acquisition module, configured to acquire context information of an object to be processed, wherein the object to be processed is used to characterize an object with an anomaly, and the context information is used to characterize information related to the anomaly; an anomaly identification module, configured to perform anomaly identification on the object to be processed based on the context information, and obtain an anomaly identification result, wherein the anomaly identification result is used to reflect the reason for the anomaly; a result parsing module, configured to parse the anomaly identification result and construct object processing information of the object to be processed; and an object processing module, configured to process the object to be processed based on the object processing information.

[0008] According to another aspect of the embodiments of this application, another object processing apparatus is also provided, including: a first display module, configured to respond to an input command applied to an operation interface and display context information of an object to be processed on the operation interface, wherein the object to be processed is used to characterize an object with an anomaly, and the context information is used to characterize information related to the anomaly; a second display module, configured to respond to a processing command applied to the operation interface and display object processing information on the operation interface, wherein the object to be processed is processed, and the object processing information is obtained by parsing anomaly identification results, the anomaly identification results are obtained by identifying anomalies in the object to be processed through context information, and the anomaly identification results are used to reflect the reason for the existence of an anomaly.

[0009] According to another aspect of the embodiments of this application, a computing device is also provided, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods in various embodiments of this application when it runs.

[0010] According to another aspect of the embodiments of this application, an electronic device is also provided, including: a memory storing an executable program; and a processor connected to the memory via a bus for running the program, wherein the program executes the methods in various embodiments of this application when it runs.

[0011] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of this application.

[0012] According to another aspect of the embodiments of this application, a computer program product is also provided, including a computer program that, when executed by a processor, implements the methods of various embodiments of this application.

[0013] According to another aspect of the embodiments of this application, a computer program product is also provided, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the methods in various embodiments of this application.

[0014] According to another aspect of the embodiments of this application, a computer program is also provided, which, when executed by a processor, implements the methods of the various embodiments of this application.

[0015] In this embodiment, the following methods are employed: obtaining context information of the object to be processed; performing anomaly identification on the object to be processed based on the context information to obtain anomaly identification results; parsing the anomaly identification results to construct object processing information for the object to be processed; and processing the object to be processed based on the object processing information. By constructing object processing information for rectifying anomalies based on the anomaly identification results obtained from the anomaly identification of the object to be processed, and processing the object to be processed according to the object processing information, an automated process for inspecting and rectifying the object to be processed can be realized, thereby solving the technical problem of low efficiency in inspecting abnormal goods in related technologies.

[0016] The above general description and the following detailed description are for illustrative and explanatory purposes only and do not constitute a limitation thereof. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0018] Figure 1 This is a schematic diagram illustrating an application scenario of an object processing method according to an embodiment of this application;

[0019] Figure 2 This is a flowchart illustrating an object processing method according to an embodiment of this application;

[0020] Figure 3 This is a schematic diagram illustrating a commodity inspection process according to an embodiment of this application;

[0021] Figure 4 This is a schematic diagram illustrating a rectification suggestion generation process according to an embodiment of this application;

[0022] Figure 5 This is a schematic diagram illustrating an abnormal rectification process according to an embodiment of this application;

[0023] Figure 6 This is a schematic diagram of a processing system for inspecting goods according to an embodiment of this application;

[0024] Figure 7 This is a flowchart illustrating another object processing method according to an embodiment of this application;

[0025] Figure 8 This is a structural block diagram of an object processing apparatus according to an embodiment of this application;

[0026] Figure 9 This is a structural block diagram of another object processing apparatus according to an embodiment of this application;

[0027] Figure 10 This is a structural block diagram of a computing device according to an embodiment of this application;

[0028] Figure 11 This is a structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0029] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some, not all, of the embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative effort should fall within the scope of protection of the present application.

[0030] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in other orders. Wherein, "other orders" refers to orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such as a process, method, system, product, or apparatus that comprises a series of steps or units, not necessarily limited to those explicitly listed, but may include other steps or units not explicitly listed, or inherent to such processes, methods, products, or apparatus.

[0031] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation portals are provided for users to choose to authorize or refuse.

[0032] The technical solution provided in this application is mainly implemented using a deep learning model. Deep learning models can be widely applied in fields such as Natural Language Processing (NLP), computer vision, and speech processing. Specifically, they can be applied to computer vision tasks such as Visual Question Answering (VQA), Image Captioning (IC), and image generation, as well as to natural language processing tasks such as text-based sentiment classification, text summarization, and machine translation. Therefore, the main application scenarios of this application include, but are not limited to, digital assistants, intelligent robots, search, online education, office software, e-commerce, and intelligent design. In this application embodiment, the scenario of analyzing anomalies in goods, using an object processing model for data processing, is used as an example for explanation.

[0033] First, some nouns or terms that appear in the description of the embodiments of this application shall be interpreted as follows:

[0034] Transaction period freeze: The funds corresponding to the order are converted into an unavailable balance for potential compensation protection.

[0035] BC Chat: Communication records between Buyer and Customer Service.

[0036] According to embodiments of this application, an object processing method is provided. The steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowcharts, in some cases the steps shown or described may be performed in a different order than that shown here.

[0037] The technical solutions provided in this application can employ deep learning models with relatively large parameter scales, such as large models containing billions or even more model parameters. Here, "large model" is just one example; this application does not limit the number of model parameters supported by the deep learning model used, aiming to meet actual needs. The deep learning models involved in this application can be artificial intelligence-based language models (LM) or multimodal models (MM).

[0038] Considering the limited computing resources of mobile terminals, the methods described above in this application embodiment can be applied to, for example... Figure 1 The application scenarios shown. Figure 1 This is a schematic diagram illustrating an application scenario of an object processing method according to an embodiment of this application. Figure 1 In the application scenario shown, the deep learning model is deployed on server 10. Server 10 can connect to one or more client devices 20 via a local area network (LAN), wide area network (WAN), internet connection, or other types of data network. Client devices 20 may include, but are not limited to, smartphones, tablets, laptops, PDAs, personal computers, smart home devices, and in-vehicle devices. Client devices 20 can interact with users through a graphical user interface to invoke the deep learning model, thereby implementing the method provided in this embodiment.

[0039] In this embodiment, the system consisting of a client device and a server can perform the following steps: the client device obtains the context information of the object to be processed; and processes the object to be processed based on the object processing information. The server performs anomaly identification on the object to be processed based on the context information, obtains the anomaly identification result, parses the anomaly identification result, and constructs the object processing information of the object to be processed.

[0040] With the rapid development of high-performance computing units, the methods provided in this application embodiment can also be applied to model-in-the-loop machines in other application scenarios. In one optional embodiment, the model-in-the-loop machine has multiple built-in models. Users can select a model to adjust as needed to obtain their own model. The high-performance computing unit built into the model-in-the-loop machine can then directly call the adjusted model to execute the methods provided in this application embodiment. In another optional embodiment, the deep learning model-in-the-loop machine has a pre-trained model built-in. The high-performance computing unit built into the model-in-the-loop machine can then directly call this model to execute the methods provided in this application embodiment.

[0041] Furthermore, when users need to train their own models, they can upload their own datasets via the client. This dataset is sent from the client to the server. The server can then use this dataset to fine-tune the pre-trained model, resulting in the user's customized model, which can then be deployed to the production environment. To facilitate user adjustments, the server provides complete adjustment tools, development frameworks, and processes, supporting various adjustment strategies. This allows the adjusted model to better adapt to different application domains and achieve a high degree of customization.

[0042] Under the aforementioned operating environment, this application provides the following: Figure 2 The object handling methods shown. Figure 2 This is a flowchart illustrating an object processing method according to an embodiment of this application. For example... Figure 2 As shown, the method may include the following steps:

[0043] Step S202: Obtain the context information of the object to be processed.

[0044] Among them, the object to be processed is used to represent the object that has an anomaly, and the context information is used to represent the information that is related to the anomaly.

[0045] The aforementioned objects to be processed can refer to objects identified by the processing system as having anomalies and requiring rectification, including but not limited to: abnormal goods, virtual services, digital content, etc.

[0046] For example, in a product inspection scenario, the aforementioned pending items could be products that have triggered buyer complaints or refunds more than or equal to a preset threshold within a historical time period. The processing system needs to perform anomaly detection on these products and rectify them based on the anomaly detection results. In a product rectification scenario, the aforementioned pending items could be products marked as "pending rectification" by the processing system and requiring information correction within a limited time to lift product control restrictions (such as demotion or transaction period freeze). The processing system needs to monitor the rectification process for these products to ensure that the rectification process can be carried out accurately and stably. For ease of understanding, the following explanation uses the scenario of inspecting products on the platform as an example.

[0047] The aforementioned anomalies may refer to behaviors that violate the platform's SOP specifications and mislead the interaction between buyers and the objects to be processed.

[0048] For example, suppose product A's details page displays it as a "New Year limited edition," but the buyer actually receives the regular version, and the details page doesn't mention "random shipping," "non-limited edition," or "packaging may differ," then product A is considered to have a discrepancy between its description and the actual product, and is implicitly misleading. Similarly, suppose product B's promotional image shows it will be wrapped in gold paper, but the buyer receives silver paper, and customer service doesn't proactively explain that multiple versions of product B's packaging exist, instead guiding the buyer to change their refund reason to "dislike," then product B is considered to have a discrepancy between the actual product and the promotional image, and is suspected of inducing a change in the refund reason.

[0049] The aforementioned contextual information may refer to multimodal data that is directly related to the above-mentioned anomalies and can support the detection, identification, attribution, and rectification guidance of the anomalies. This may include, but is not limited to: evidence materials provided by buyers regarding the anomalies and text of refund application reasons, complete chat logs between buyers and customer service, text and image content of the original product details page, historical violation records of the product, and the platform's SOP (Standard Operating Procedure) specifications for the category to which the product belongs.

[0050] In one optional solution of this embodiment, considering that there are a large number of products displayed on the platform, if all products are subjected to indiscriminate and full-coverage manual inspection or rule scanning, it may lead to huge consumption of computing resources, extremely high manual review costs, and excessively long response cycles, which may affect the efficiency of inspecting products on the platform and the timeliness of detecting and rectifying anomalies in products.

[0051] Therefore, in order to improve the efficiency of inspecting goods on the platform, during the inspection process, the processing system can select objects with abnormalities from the order database according to preset screening rules, namely the above-mentioned objects to be processed, and then conduct targeted inspections on the objects to be processed, thereby realizing the transformation from manual sampling to intelligent pre-screening and improving the efficiency and comprehensiveness of inspecting goods on the platform.

[0052] The screening rules may include, but are not limited to: the same product has 3 or more refund orders in the past 30 days, and the refund reason contains keywords such as "description does not match" or "the actual product is inconsistent with the picture", the number of evidence pictures uploaded by the buyer is ≥2, the BC chat history contains high-risk semantic patterns such as "random delivery not specified" or "inducing modification of refund reason", or the product's historical rectification recurrence rate is greater than the threshold.

[0053] Furthermore, considering that anomalies in products on the platform are often somewhat hidden and scenario-dependent, the essence of such anomalies is not triggered by a single keyword, but is caused by a combination of multimodal behaviors such as semantic conflicts between images and text, inducements from customer service scripts, and avoidance of category rules. If we simply analyze and attribute the anomalies in products through keyword matching by the rule engine or manual sampling, such as analyzing only the refund text or only identifying image defects to analyze the cause of the anomaly, it is easy to make misjudgments.

[0054] For example, if a buyer uploads a picture of a moldy product when applying for a refund, but fills in "incorrect size" as the reason for the refund, and the processing system determines that the product is not abnormal simply because the refund reason is "not a quality issue," it may overlook a real food safety problem.

[0055] If a buyer explicitly states in the BC chat that "I received item A, but ordered item B", but fails to upload a photo of the actual item as evidence, the processing system may deem the buyer's evidence invalid due to the lack of visual description, thus ignoring the issue of the seller sending the wrong item.

[0056] If a merchant labels "contains trace amounts of preservatives" in small print on the product details page, but uses the phrase "0 added" in the product's main image and title, the processing system may not trigger an alert because the keyword "preservatives" is not matched, thus ignoring the issue of the merchant publishing false information.

[0057] If customer service guides buyers to change their refund reason in the BC chat, such as "Don't write 'the description doesn't match,' just write 'I don't like it,'" the processing system may overlook the issue of the merchant's inducement to evade responsibility because such wording is not included in the rule base.

[0058] Therefore, in order to accurately locate and analyze anomalies in the objects to be processed, and to construct interpretable, verifiable, and category-customized anomaly identification results, after identifying the objects to be processed that have anomalies, the processing system can obtain the context information of the objects to be processed, and perform anomaly analysis based on the context information, thereby ensuring the efficiency and accuracy of identifying anomalies in products.

[0059] Step S204: Based on the context information, perform anomaly identification on the object to be processed to obtain the anomaly identification result.

[0060] Among them, the anomaly identification results are used to reflect the reasons for the existence of anomalies.

[0061] In one optional solution of this embodiment, in order to accurately rectify the anomalies in the goods and avoid errors in the rectification direction due to relying solely on a single text rule or vague manual judgment, which would affect the efficiency of the goods inspection, after obtaining the context information of the object to be processed, the processing system can identify the anomalies of the object to be processed based on the context information and obtain the anomaly identification result.

[0062] In order to ensure the accuracy of rectifying anomalies in products when identifying anomalies in the objects to be processed, the processing system can start from dimensions such as "anomaly attribution goal orientation, multi-source evidence fusion analysis, and dynamic task rule constraints" to clarify "why to conduct anomaly analysis", "how to conduct anomaly analysis", and "what factors need to be considered when conducting anomaly analysis". Based on the obtained context information, the system can actively analyze the anomalies in the objects to be processed, determine the cause of the anomalies, and obtain the above-mentioned anomaly analysis results.

[0063] In one optional embodiment, when performing anomaly identification on the object to be processed using context information, the processing system can first understand the buyer's current after-sales needs based on the after-sales order initiated by the buyer. Then, based on information such as BC chat records, evidence images uploaded by the buyer, text and image descriptions on the product details page, and platform SOP specifications for the product category, the system uses a pre-trained anomaly analysis model to jointly encode and semantically align the evidence images and BC chat text, identify the conflict relationship between the product details page and the physical characteristics, and, in conjunction with the platform SOP specifications, determine whether the conflict constitutes a compliance anomaly as defined by the platform, thereby analyzing the root cause of the product's anomaly.

[0064] For example, suppose a buyer initiates a refund request for a product, stating that "the received product does not match the product displayed by the seller." The processing system can determine the abnormal characteristic of the product based on the buyer's uploaded product image showing a regular version, while the main image on the product details page shows a "New Year limited edition." Combined with the buyer's explicit mention in the BC chat log that "the page said there was a New Year edition, so why was the old version sent?", the system determines that the product's abnormal characteristic is "the advertised version is inconsistent with the actual shipped version." Subsequently, the system calls the apparel category knowledge base and finds that 87% of similar complaints in this category stem from sellers failing to indicate "random shipping" or "mixed versions" on the details page, and this product already has two previous records of the same complaint. Therefore, the system determines that the root cause of this abnormality is: to reduce the return rate, the seller deliberately uses a highly attractive version in the promotional image but fails to clearly state on the details page that multiple versions exist and are randomly shipped, constituting a systemic misleading behavior.

[0065] Step S206: Analyze the anomaly identification results and construct object processing information for the object to be processed.

[0066] The aforementioned object processing information may refer to information on how to rectify any exceptions existing in the object to be processed.

[0067] In one optional embodiment, in order to accurately rectify anomalies in the product, after analyzing the anomaly identification result based on the context information, the processing system can further parse the anomaly identification result to determine the anomaly type, evidence anchor point, cause of existence, and other information, and construct object processing information for rectifying the anomaly based on a structured and standardized rectification instruction template.

[0068] For example, suppose the anomaly identification results show that the goods sold by the merchant are not as described, that is, the goods purchased by the buyer do not match the goods actually received. The processing system can determine the information related to the problem based on the anomaly identification results: (1) the physical image uploaded by the buyer shows "normal" packaging; (2) the main image and title of the product details page advertise "New Year limited edition"; (3) the buyer clearly pointed out in the BC chat history, "The page says it is a limited edition, but why is it a normal edition?" Then, by calling the knowledge base related to the product category, it is determined whether "not clearly marked multiple versions and random delivery" is a common violation under this category. If so, the processing system can use the information related to the problem determined above to automatically fill the preset "version confusion" rectification template and generate a rectification instruction to rectify the problem of the goods not matching the description.

[0069] For issues such as discrepancies between the goods ordered and received, the rectification instructions may include:

[0070] Corrective action: Add a prominent note to the first screen of the product details page and below the main image: "This product includes two packaging options, shipped randomly, not due to quality issues."

[0071] Evidence required: Upload clear comparison photos of the two products (limited edition and regular edition);

[0072] Verification method: After rectification, you need to upload a screenshot of the modified details page in the merchant backend. The system will automatically verify whether the text and images are complete.

[0073] Consequence warning: If rectification is not completed within a certain period of time, transaction account freeze and flow restriction will be automatically triggered.

[0074] Based on this, the processing system can generate accurate rectification suggestions, namely the aforementioned object processing information, so that merchants can clearly understand "why they were penalized, how to rectify, and how to verify after rectification", thereby improving the completion rate and initiative of rectifying anomalies in products, reducing the appeal rebound rate, and realizing the platform's transformation from a "post-event punishment" to a "pre-event guidance" governance model.

[0075] Step S208: Process the object to be processed based on the object processing information.

[0076] In one optional embodiment, after constructing the object processing information of the object to be processed, the processing system can process the object to be processed according to the object processing information to rectify the anomalies in the product.

[0077] For example, if a merchant reports that they can rectify the anomalies in their product according to the rectification requirements included in the object processing information within a preset time period, the processing system can send the object processing information to the merchant, allowing the merchant to proactively adjust the product's anomalies. During the adjustment process, the processing system can also monitor the rectification process in real time to automatically determine whether the merchant's rectification actions meet the rectification requirements, thereby improving the accuracy of rectifying product anomalies.

[0078] If a merchant fails to rectify the product anomalies according to the rectification requirements included in the object processing information within the preset time period, or if the merchant still fails to meet the rectification requirements after multiple rectifications, the processing system can automatically rectify the product anomalies based on the object processing information, provided the merchant has authorized rectification. This can include adjusting the description text on the product details page, inserting the phrase "This product includes two packaging options, shipped randomly, not a quality issue" into the main image text, and automatically embedding a system-generated comparison image of the two products on the first screen of the details page. Upon completion of the rectification, the system will send a notification to the merchant indicating that rectification has been completed, along with the relevant rectification operations performed. This improves the efficiency and accuracy of rectifying product anomalies while ensuring the merchant's right to know and right to appeal.

[0079] If a merchant does not authorize the processing system to rectify anomalies in their product listings, and the merchant fails to rectify the anomalies according to the rectification requirements, the processing system can automatically trigger the platform's control mechanisms for the merchant or product. These mechanisms may include freezing the transaction funds for the corresponding order, restricting the product's participation in platform promotional activities, reducing its ranking weight on search and recommendation pages, and sending a "rectification failed" notification to the merchant along with the specific reasons for the failure. This allows the platform to shift its governance from "passive penalties" to "proactive prevention + intelligent fallback," thus protecting the rights of buyers.

[0080] In this embodiment, the following methods are employed: obtaining context information of the object to be processed; performing anomaly identification on the object to be processed based on the context information to obtain anomaly identification results; parsing the anomaly identification results to construct object processing information for the object to be processed; and processing the object to be processed based on the object processing information. By constructing object processing information for rectifying anomalies based on the anomaly identification results obtained from the anomaly identification of the object to be processed, and processing the object to be processed according to the object processing information, an automated process for inspecting and rectifying the object to be processed can be realized, thereby solving the technical problem of low efficiency in inspecting abnormal goods in related technologies.

[0081] In this embodiment, the context information includes text information and image information. Anomaly identification is performed on the object to be processed based on the context information to obtain anomaly identification results, including: encoding the text information and image information respectively to obtain text encoding results and image encoding results; based on the text encoding results and image encoding results, semantic matching is performed between the text words contained in the text information and the image blocks contained in the image information to obtain semantic matching results, wherein the semantic matching results are used to characterize the association between the semantic features of the text information and the semantic features of the image information; the anomaly type is determined based on the semantic matching results; and attribution analysis is performed on the anomaly based on the anomaly type and context information to obtain anomaly identification results.

[0082] In one optional solution of this embodiment, in order to achieve accurate attribution of product violations and overcome the bottleneck of high misjudgment rate in traditional single-modal recognition in the case of conflicting text and image information, the obtained context information may include at least the text information and image information corresponding to the object to be processed.

[0083] For example, text information may include reasons such as "received goods do not match the description" filled in by the buyer in the refund application, or text in the BC chat history with customer service such as "the page said it was a limited edition for the New Year, but I received a regular version?" Image information may include images such as photos of the actual packaging taken and uploaded by the buyer, close-up photos of the product label, and comparisons with the main image on the details page.

[0084] Based on this, in order to achieve strong alignment and consistency judgment at the multimodal semantic level when there is semantic ambiguity, vague expression or deliberate inducement in the text and image information, the processing system can first encode the text information and image information separately during the anomaly identification of the object to be processed based on the context information, and obtain the corresponding text encoding results and image encoding results.

[0085] Then, the processing system can perform semantic matching between the text words contained in the text information and the image blocks contained in the image information based on the text encoding results and the image encoding results, so as to determine the association between the semantic features of the text information and the semantic features of the image information through the semantic matching results.

[0086] After obtaining the semantic matching results, the processing system can further determine the anomaly type based on the semantic matching results, and perform attribution analysis on the anomaly based on the anomaly type and context information to obtain the corresponding anomaly identification results.

[0087] For example, when an abnormal product is identified, the processing system can first obtain text and image information related to the abnormality of the product. For instance, the buyer fills in "received product does not match the description" in the refund application, and explicitly asks in the BC chat log "the page shows a limited edition for the New Year, why did I receive a regular product?", while uploading three images: the main image of the product details page, a front view of the physical outer packaging, and a close-up image of the label inside the product.

[0088] Next, the processing system can encode the acquired text and image information. For example, it can segment the BC chat records into word sequences (such as "the page says", "New Year limited edition", "how it was sent", "regular edition"), divide the physical images uploaded by the buyer into 16×16 pixel image blocks, and extract the semantic embedding vector of each image block, thereby obtaining the text encoding results and image encoding results related to the anomalies of the product.

[0089] Then, the processing system can use a cross-modal attention mechanism to calculate the semantic similarity between each text term and each image block, obtain the semantic matching results mentioned above, and determine the anomaly type of the product based on the semantic matching results. For example, if the term "New Year limited edition" has a high matching score with the image block "text label area in the main image of the details page" but a low matching score with the image block "no label on the outer packaging of the actual product", it can be determined that the buyer's description of "the advertisement does not match the actual product" has a strong correspondence in the text and image evidence, and this difference is not an occasional photographic error, but a systematic misleading advertisement. The anomaly type of the product is currently: version confusion type description mismatch.

[0090] Based on this, the processing system further invokes pre-configured platform SOP specifications to determine the conditions that the product must meet when selling it. For example, "If a product has multiple versions (such as limited edition, regular version, and randomly shipped version), the merchant must clearly indicate the version differences and shipping rules in a prominent position on the first screen of the details page; failure to do so is considered a violation." Simultaneously, the processing system can query the product's historical complaint data for the past 30 days. If it finds that the product has received two or more complaints of the same type without rectification, the processing system can determine that the product is a "repeated, habitual violation," thus identifying the reason for the product's anomaly as follows: the merchant deliberately uses highly attractive version images for promotion to increase conversion rates, but intentionally circumvents the platform's mandatory disclosure obligations by failing to clearly indicate on the details page that the actual shipped version is the regular version, constituting a systemic misleading behavior towards consumers.

[0091] Based on this, the processing system can obtain structured and standardized anomaly identification results. These results can include: anomaly type, core conflicting evidence pairs, attribution criteria, historical behavior, and other relevant information.

[0092] In this embodiment, processing an object to be processed based on object processing information includes: generating a processing work order corresponding to the object to be processed based on the object processing information, wherein the processing work order is used to track the processing process of the object to be processed; outputting the processing work order and controlling the object to be processed to enter a control state, wherein the control state is used to represent the control of interactive behaviors that are related to the object to be processed; in response to detecting a change in the object characteristics of the object to be processed, matching the changed object characteristics with the information characteristics of the object processing information to obtain a feature matching result; in response to the feature matching result indicating that the changed object characteristics match the information characteristics, determining that the object to be processed has been successfully processed, and controlling the object to be processed to exit the control state.

[0093] The aforementioned processing work order can refer to a structured task order automatically generated by the processing system, containing information such as product ID, exception type, graphic evidence chain, category SOP violation clauses, rectification suggestions, and rectification acceptance standards. Through processing work orders, the processing system can achieve comprehensive tracking of the exception rectification process.

[0094] The aforementioned control status refers to legally binding, phased interactive restrictions imposed by the processing system on a product or merchant in accordance with management regulations, such as platform SOP specifications. These restrictions may include, but are not limited to: restricting the product's participation in platform promotional activities, reducing its ranking weight on search and recommendation pages, and freezing transaction funds for related orders. Furthermore, the control status can only be lifted after the merchant has rectified the anomaly.

[0095] In one optional embodiment, during the process of processing the object to be processed using object processing information, in order to monitor the merchant's rectification of anomalies in the goods, the processing system can generate a processing work order corresponding to the object to be processed based on the constructed object processing information, so as to track the processing process of the object to be processed in real time through the processing work order.

[0096] After generating a processing work order, the system can output the work order and control the status of the objects to be processed. For example, it can automatically remove products from the "Sales Ranking," "Homepage Recommendations," and "Top 3 Search Pages," while freezing the settlement funds for all orders of that product within the past 7 days. This prevents merchants from transferring revenue before rectification, thereby effectively curbing the behavior of merchants who "violate regulations first, then rectify, and shirk responsibility."

[0097] After receiving a processing work order from the processing system, merchants can adjust the anomalies in the pending object according to the rectification suggestions provided in the work order. If a change is detected in the adjusted object, the processing system can match the changed object characteristics with the information characteristics of the object's processing information to obtain a feature matching result. This feature matching result is used to determine whether the merchant has rectified the anomalies in the pending object according to the rectification suggestions.

[0098] If the feature matching result shows that the changed object features match the information features, it means that the merchant has successfully rectified the anomalies of the object to be processed according to the rectification suggestions. At this time, the processing system can determine that the object to be processed has been successfully processed and control the object to be processed to exit the management status, thereby automatically restoring the traffic permissions and fund settlement qualifications of the product, pushing a "rectification successful" notification to the merchant at the same time, and writing this rectification record into the merchant's compliance credit file for subsequent credit scoring, dynamic adjustment of governance strategies, and adjudication of appeals and disputes.

[0099] In this embodiment of the application, generating a processing work order corresponding to the object to be processed based on object processing information includes: evaluating the confidence level of the context information to obtain a confidence level evaluation result; generating a processing work order based on the work order generation module and the object processing information in response to the confidence level evaluation result indicating that the confidence level is greater than or equal to a preset threshold; and outputting the context information and object processing information in response to the confidence level evaluation result indicating that the confidence level is less than the preset threshold, receiving the information review result corresponding to the object processing information, and generating a processing work order based on the information review result and the work order generation module.

[0100] In one optional solution of this embodiment, considering that violations involving products on the platform often heavily rely on semantic understanding of images and text, and involve numerous ambiguous expressions, metaphorical representations, and deliberate misleading statements by merchants (e.g., using non-standard language such as "similar product," "same product with different name," or "random packaging," or directly displaying images of similar products on the product details page), directly generating automatic processing tickets could lead to increased misjudgment rates, a surge in merchant complaints, and damage to the platform's credibility.

[0101] Therefore, during the process of generating a processing work order, the processing system can first evaluate the obtained context information to obtain a confidence assessment result. The confidence assessment result is used to determine whether staff need to review the context information, ensuring the accuracy, compliance, and executability of the generated processing work order.

[0102] Correspondingly, if the confidence assessment result shows that the confidence level is greater than or equal to the preset threshold, the processing system can automatically generate the corresponding processing work order using the work order generation module and object processing information. If the confidence assessment result shows that the confidence level is less than the preset threshold, the processing system may consider that the current context information may have semantic ambiguity, image-text mismatch, or conflict with historical violation patterns. The processing system cannot make an independent decision in this case. In this situation, the processing system can output the context information and object processing information for manual review by staff to obtain the corresponding information review result. Finally, the processing system can generate the corresponding processing work order based on the information review result and the work order generation module.

[0103] For example, when the processing system detects a "children's toy" product where a buyer's evidence image shows the toy contains small parts, but the product details page does not indicate "not suitable for children under 3 years old," and the buyer's chat history states that "this thing feels like it could choke a child," without directly using keywords such as "safety hazard" or "choking risk," even though the buyer's image does contain small parts, the processing system may not be able to accurately assess whether the small parts pose a safety risk due to issues such as the shooting angle and cluttered background. Furthermore, if the product is newly listed and has no historical complaint records, the processing system can determine that the product is a "highly ambiguous sample," and cannot directly judge whether the identified anomaly is accurate.

[0104] At this point, the processing system can output contextual information such as the evidence submitted by the buyer, the B2C chat logs, and the product details page to the staff, who can then proactively determine if the product has any anomalies and whether rectification is needed. If the staff indicates that the product has an anomaly and requires rectification, the processing system can generate a corresponding processing work order based on the contextual information. If the staff indicates that the product does not have any anomalies, the processing system can assume that the identification process for the product anomaly may have been erroneous and does not require generating a corresponding processing work order.

[0105] In this embodiment of the application, the above method further includes: in response to detecting that the object features of the object to be processed have not changed within a preset time period, or that the feature matching result indicates that the changed object features do not match the information features, determining the processing stage for processing the object to be processed based on the processing work order; obtaining the object management method corresponding to the processing stage; and executing the object management method based on the context information to handle the exception.

[0106] In one optional embodiment, if a merchant, upon receiving a processing work order, fails to rectify the anomalies in the product within a preset time period, or if the rectification results do not meet the rectification requirements—that is, if the processing system detects that the object characteristics of the object to be processed have not changed within the preset time period, or the feature matching results show that the changed object characteristics do not match the information characteristics—it can be considered that the merchant has not performed the rectification task within a reasonable period, or that the rectification actions have not substantially resolved the anomalies in the product, constituting "invalid rectification" or "passive rectification." In this case, the processing system can first determine the current processing stage of the object to be processed based on the processing work order, and then process the anomaly according to the object management method corresponding to the processing stage, combined with contextual information.

[0107] Different processing stages can have different object management methods. For example, if the processing stage is the initial stage (within 3 days from the date of issuance of the work order, if the merchant fails to complete the rectification for the first time), it can be considered that the merchant is still in the rectification adaptation period, with cognitive bias or operational delay, and has not yet constituted malicious circumvention behavior. At this time, the processing system can automatically send a secondary reminder notification, with a rectification example video, category compliance template and manual customer service access, and appropriately extend the rectification period to 15 days. At the same time, the exposure weight of the product in the search results will be reduced by 10% to gently guide the rectification.

[0108] If the processing stage is the review stage (if rectification is not completed within 15 days, or if rectification is repeatedly deemed unsuccessful), it can be considered that the merchant has clearly ignored or been perfunctory in fulfilling the rectification requirements, and has entered the "repeated violation" risk zone. At this time, the processing system can implement medium-level control: freeze the transaction funds of the order related to the product for 7 days, block its participation in platform promotions and homepage recommendations, proactively push "warning" prompts to buyers, and mark the merchant as having a "low compliance credit rating".

[0109] In this embodiment of the application, obtaining the context information of the object to be processed includes: obtaining the interaction materials submitted by the first entity in response to the exception, wherein the first entity is used to represent the entity that has a need for the object to be processed; obtaining the communication materials generated by the communication between the first entity and the second entity in response to the exception, wherein the second entity is used to represent the entity that provides the object to be processed to the first entity; obtaining the object materials of the object to be processed; and summarizing the interaction materials, communication materials and object materials to obtain the context information.

[0110] In one optional embodiment, in order to accurately identify and rectify anomalies in the product, when obtaining the context information of the object to be processed, the processing system can first obtain the interactive materials submitted by the first entity that has a need for the object to be processed, namely the buyer, regarding the anomalies in the product. These materials may include structured or unstructured evidence materials such as refund application reason text, product review content, original platform complaint ticket, photos of the product uploaded by the buyer, unboxing videos, screenshots of test reports, and photos of logistics documents.

[0111] Simultaneously, the first entity and the second entity are obtained. The first entity provides the object to be processed to the second entity. Communication materials generated during communication regarding the anomaly are also obtained, such as BC chat logs generated when buyers and merchants communicate about the anomaly of the product, platform customer service intervention records, SMS / in-site messages exchanged to negotiate refund solutions, screenshots of compensation solutions proactively sent by the merchant, and two-way communication content such as after-sales telephone recordings converted to text.

[0112] To ensure the anomaly identification process is complete and adversarial in terms of "correspondence between images and text, clear subject, traceable time, and closed-loop evidence," the processing system can also obtain the object materials of the object to be processed, such as the product details page, historical version snapshots, product qualification documents, and clustering results of historical complaints about the same product.

[0113] After acquiring the aforementioned materials, the processing system can summarize these interactive materials, communication materials, and object materials to obtain the corresponding contextual information.

[0114] In this embodiment of the application, the anomaly identification result is parsed to construct object processing information for the object to be processed, including: obtaining the object type of the object to be processed; calling the processing information template based on the anomaly type and the object type, wherein the processing information template is used to indicate the constraints for constructing the object processing information; and generating object processing information according to the processing information template based on the anomaly identification result and context information.

[0115] In one optional solution of this embodiment, in order to accurately construct the object processing information of the object to be processed and avoid the generalization or invalidity of rectification suggestions due to category differences, during the process of constructing the object processing information, the processing system can first obtain the object type of the object to be processed, and call the processing information template used to indicate the constraints for constructing the object processing information according to the exception type and object type. Then, according to the exception identification result and context information, the corresponding object processing information is generated according to the processing information template.

[0116] To improve the efficiency of generating object processing information, an object processing model can be configured in the processing system. This model parses the anomaly identification results and constructs the corresponding object processing information. For example, the processing system can first define the field structure, required elements, evidence citation standards, and language style constraints using a processing information template. Then, the object processing model fills in the content based on contextual information to construct the corresponding object processing information. This ensures that the output of the object processing information conforms to the platform's SOP specifications and possesses natural language expression capabilities.

[0117] For example, when the processing system identifies a "description mismatch" anomaly in a "baby food" product—"advertised as '0 added preservatives,' but the actual packaging label uploaded by the buyer shows 'sodium benzoate'"—and the product type is "baby food," the system can automatically call a preset processing information template. Then, based on the anomaly identification results and contextual information, the system can fill in the processing information template. This template can be constructed according to mandatory regulations for food additive labeling to ensure that the generated processing information aligns with these regulations, thus improving the compliance of the processing information.

[0118] In this embodiment of the application, the above method further includes: acquiring processing data generated during the processing of the object to be processed; evaluating the anomaly based on the processing data and context information to obtain an anomaly evaluation result; updating the initial sample database based on the anomaly evaluation result, processing data, and context information to obtain a target sample database, wherein the target sample database contains multiple training samples, the multiple training samples are used to train the object processing model, the object processing model is used to parse the anomaly identification result, and construct the object processing information of the object to be processed.

[0119] In one optional embodiment, in order to continuously improve the object processing model's ability to identify and generate in complex, ambiguous, and novel violation scenarios, and to avoid "knowledge drift" or "coverage blind spots" due to lagging training data, the processing system can also acquire the processing data generated during the processing of the object to be processed, i.e., the rectification of anomalies in the product. Then, the system evaluates the anomalies based on the processing data and contextual information to obtain anomaly evaluation results. The anomaly evaluation results can be used to determine whether the entire rectification process can be regarded as a typical case.

[0120] The data generated during the entire rectification process may include, but is not limited to: the rectification materials finally submitted by the merchant, the correction opinions of the human reviewers, the rectification results automatically verified by the system, the merchant's appeal feedback text, and the changes in after-sales data of orders after rectification, as well as other structured and unstructured data.

[0121] If a case can be considered a typical example, it means that the case has high information density, strong representativeness and transferability. It is an important sample for adjusting the semantic understanding ability of the object processing model. The processing system can then update the initial sample database based on the anomaly evaluation results, combined with the processing data and contextual information, to obtain the target sample database. That is, to construct new training samples for training the object processing model, so as to improve the object processing model's ability to parse anomaly identification results and ensure the accuracy of the constructed object processing information.

[0122] To facilitate the immediate handling of any exceptions in the object being processed, Figure 3 This is a schematic diagram illustrating a product inspection process according to an embodiment of this application. Figure 3 As shown, during the product inspection process, the processing system can first select products with abnormalities, then detect the abnormalities, generate corresponding rectification suggestions, and encapsulate the detection results and rectification suggestions into an interface for staff to review and determine whether the merchant needs to make rectifications or whether the merchant needs to be penalized, thereby realizing the process of detecting product quality.

[0123] In the process of detecting anomalies in products, the processing system can acquire contextual information related to the anomalies, such as text and image information. It then performs anomaly detection based on this information, obtaining the anomaly detection results. Next, the system can utilize an object processing model, combining the processing information template and the anomaly detection results, to generate rectification suggestions for addressing the anomalies, instructing the merchant on how to rectify them.

[0124] Figure 4 This is a schematic diagram illustrating a rectification suggestion generation process according to an embodiment of this application. For example... Figure 4 As shown, the process of generating rectification suggestions can be divided into 6 parts, including: selecting orders, information summarization, template application, model processing, result cleaning, and result output.

[0125] Correspondingly, in the process of generating corresponding rectification suggestions for anomalies in products, in the order selection section, the processing system can obtain product orders with anomalies reported by buyers and mark the anomalies in these product orders.

[0126] In the information aggregation section, the processing system can obtain the corresponding context information for the marked anomalies, namely the text and image information that are related to the anomalies of the products, and perform operations such as information denoising, format standardization, text evidence extraction, image evidence extraction, and evidence storage on the obtained context information to obtain relevant data that proves that the products are anomalies.

[0127] In the template application section, the processing system can construct or obtain corresponding rectification suggestion templates based on the product type and the exception type. Then, it can use the relevant data summarized in the information summary section to populate the rectification suggestion templates and generate prompt information to indicate the construction of rectification suggestions.

[0128] In the model application section, the processing system can use a pre-configured object processing model, combined with the prompts output by the model application section and the relevant data summarized by the information summary section, to perform multimodal reasoning on the anomalies of the product, in order to determine the relevant information of the anomaly, and output rectification suggestions for rectifying the anomaly.

[0129] In the results cleaning section, the processing system can perform structured transformation and field validation on the rectification suggestions output by the model application section to ensure the readability, interpretability, and accuracy of the final output rectification suggestions.

[0130] In the output section, the processing system can call relevant interfaces to output rectification suggestions after cleaning, so that staff can review the rectification suggestions, or merchants can rectify the anomalies in the products according to the rectification suggestions.

[0131] Figure 5 This is a schematic diagram illustrating an abnormal rectification process according to an embodiment of this application. For example... Figure 5 As shown, the process of rectifying anomalies in products can be divided into four parts: suggestion generation, work order construction, anomaly rectification, and strategy application.

[0132] In the process of rectifying anomalies in products, during the suggestion generation section, the processing system can first use the object processing model, combined with contextual information related to the anomaly and rectification suggestion templates, to generate rectification suggestions for rectifying the anomalies in products. The system then evaluates the confidence level of the contextual information to obtain the confidence level evaluation result, in order to determine whether a corresponding processing work order can be directly generated based on the rectification suggestions.

[0133] In the work order construction section, if the confidence assessment result shows that a processing work order can be generated, the processing system can directly generate the partially generated rectification suggestions based on the suggestions and construct the corresponding processing work order; if the confidence assessment result shows that a processing work order cannot be generated, the processing system can output rectification suggestions, which will be reviewed and adjusted by staff to obtain the corresponding review results. Then, the processing system will construct the corresponding processing work order based on the received review results.

[0134] In the anomaly rectification section, the processing system can output constructed processing work orders to merchants and monitor in real time whether product characteristics change within a preset time period. If product characteristics change, the processing system can match the changed product characteristics with rectification suggestions, and staff can determine whether the rectified product meets the rectification requirements. Correspondingly, if the rectified product meets the rectification requirements, or the merchant removes the product from the platform, the anomaly rectification can be considered successful. If the rectified product does not meet the rectification requirements, or the product characteristics do not change within the preset time period, the processing system can determine that the current anomaly rectification has failed, and will start a new round of rectification based on the rectification suggestions until the anomaly rectification is successful, or the number of rectification rounds reaches a preset threshold.

[0135] In the strategy application section, the processing system can execute corresponding control strategies based on the current stage of rectifying anomalies in a product. For example, when generating rectification suggestions, the system can invite buyers to evaluate the suggestions and determine user satisfaction with the generated rectification recommendations. In the initial stage of rectification, the system can implement low-level controls, such as putting the product into a controlled state to manage interactions related to the product. In the review stage of rectification, the system can implement medium-level controls, such as freezing transaction funds for orders associated with the product and sending warning messages to the merchant, emphasizing the importance of rectifying the product's anomalies and the potential consequences of failure to rectify them. In the final stage of rectification, the system can terminate control over the product or merchant.

[0136] Figure 6 This is a schematic diagram of a product inspection system according to an embodiment of this application, as shown below. Figure 6 As shown, the processing system can include six main components: MQ message queue, engineering service, work order system, model interface service, merchant front-end, and merchants.

[0137] Among them, the MQ message queue can push inspection orders to the engineering service to carry out product inspection operations.

[0138] After receiving an inspection order, the engineering service can send a request to the work order system to create a problem product work order. The corresponding work order system can create the problem product work order based on the received request and return the created problem product work order and a success message to the engineering service. The problem product work order can be used to screen products with abnormalities, providing a basis for subsequent anomaly rectification.

[0139] After successfully creating a problematic product work order, the engineering service can query the context information of the product with the anomaly through the model interface service and receive the query results returned by the model interface service. After successfully retrieving the product's context information, the engineering service can generate corresponding rectification suggestions based on the context information and send the product's anomaly and corresponding rectification suggestions to the merchant's front end. The merchant's front end then pushes the anomaly and rectification suggestions to the merchant, enabling the merchant to rectify the product's anomaly according to the rectification suggestions.

[0140] For the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions. This is because, according to this application, certain steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0141] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms, or by hardware. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of this application.

[0142] According to an embodiment of this application, another object processing method is also provided. Figure 7 This is a flowchart illustrating another object processing method according to an embodiment of this application, such as... Figure 7 As shown, the method includes the following steps:

[0143] Step S702: In response to the input command applied to the operation interface, display the context information of the object to be processed on the operation interface.

[0144] Among them, the object to be processed is used to represent the object that has an anomaly, and the context information is used to represent the information that is related to the anomaly.

[0145] Step S704: In response to the processing instructions applied to the operation interface, display object processing information on the operation interface.

[0146] The process involves processing the object to be processed. The object processing information is obtained by parsing the anomaly identification results. The anomaly identification results are obtained by identifying anomalies in the object to be processed using context information. The anomaly identification results are used to reflect the reasons for the existence of anomalies.

[0147] In one optional embodiment, upon detecting an input command applied to the user interface, the processing system may first display the context information of the object to be processed on the user interface. The object to be processed may refer to an object exhibiting an anomaly, and the context information of the object may refer to information relating to the anomaly.

[0148] Upon detecting a processing command applied to the user interface, the processing system can use context information to identify anomalies in the object to be processed, obtaining an anomaly identification result reflecting the cause of the anomaly. This result is then parsed to obtain the object processing information corresponding to the object to be processed, which is displayed on the user interface. The processing system can then rectify any anomalies present in the object to be processed based on this object processing information.

[0149] According to an embodiment of this application, an object processing apparatus for implementing the above-described object processing method is also provided. Figure 8 This is a structural block diagram of an object processing apparatus according to an embodiment of this application, such as... Figure 8 As shown, the device includes: an information acquisition module 802, an anomaly identification module 804, a result parsing module 806, and an object processing module 808.

[0150] The information acquisition module 802 is used to acquire the context information of the object to be processed, wherein the object to be processed is used to represent an object with an anomaly, and the context information is used to represent information related to the anomaly; the anomaly identification module 804 is used to identify anomalies in the object to be processed based on the context information and obtain anomaly identification results, wherein the anomaly identification results are used to reflect the cause of the anomaly; the result parsing module 806 is used to parse the anomaly identification results and construct object processing information for the object to be processed; and the object processing module 808 is used to process the object to be processed based on the object processing information.

[0151] The information acquisition module 802, anomaly identification module 804, result parsing module 806, and object processing module 808 described above correspond to steps S202 to S208 in the above embodiments. The four modules and their corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in the above embodiments. The modules or units described above may be hardware or software components stored in memory and processed by one or more processors. These modules may also be part of a device and may run in the server 10 provided in the above embodiments.

[0152] In this embodiment, the context information includes text information and image information; the anomaly recognition module is further configured to: encode the text information and image information respectively to obtain text encoding results and image encoding results; based on the text encoding results and image encoding results, perform semantic matching between the text words contained in the text information and the image blocks contained in the image information to obtain semantic matching results, wherein the semantic matching results are used to characterize the association between the semantic features of the text information and the semantic features of the image information; determine the anomaly type based on the semantic matching results; and perform attribution analysis on the anomaly based on the anomaly type and context information to obtain anomaly recognition results.

[0153] In this embodiment, the object processing module is further configured to: generate a processing work order corresponding to the object to be processed based on the object processing information, wherein the processing work order is used to track the processing process of the object to be processed; output the processing work order and control the object to be processed to enter the control state, wherein the control state is used to represent the control of interactive behaviors that are related to the object to be processed; in response to detecting a change in the object characteristics of the object to be processed, match the changed object characteristics with the information characteristics of the object processing information to obtain a feature matching result; in response to the feature matching result representing that the changed object characteristics match the information characteristics, determine that the object to be processed has been successfully processed, and control the object to be processed to exit the control state.

[0154] In this embodiment, the object processing module is further configured to: evaluate the confidence level of the context information to obtain a confidence level evaluation result; generate a processing work order based on the work order generation module and the object processing information in response to the confidence level evaluation result indicating that the confidence level is greater than or equal to a preset threshold; and output the context information and object processing information in response to the confidence level evaluation result indicating that the confidence level is less than the preset threshold, receive the information review result corresponding to the object processing information, and generate a processing work order based on the information review result and the work order generation module.

[0155] In this embodiment of the application, the object processing module is further configured to: in response to detecting that the object features of the object to be processed have not changed within a preset time period, or that the feature matching result indicates that the changed object features do not match the information features, determine the processing stage for processing the object to be processed based on the processing work order; obtain the object management method corresponding to the processing stage; and execute the object management method based on the context information to handle the exception.

[0156] In this embodiment of the application, the information acquisition module is further configured to: acquire the interaction materials submitted by the first entity in response to the exception, wherein the first entity is used to represent an entity that has a need for the object to be processed; acquire the communication materials generated by the communication between the first entity and the second entity in response to the exception, wherein the second entity is used to represent an entity that provides the object to be processed to the first entity; acquire the object materials of the object to be processed; and summarize the interaction materials, communication materials and object materials to obtain context information.

[0157] In this embodiment of the application, the result parsing module is further configured to: obtain the object type of the object to be processed; call the processing information template based on the exception type and object type of the exception, wherein the processing information template is used to indicate the constraints for constructing object processing information; and generate object processing information according to the processing information template based on the exception identification result and context information.

[0158] In this embodiment of the application, the above-mentioned device further includes: a data acquisition module, used to acquire processing data generated during the processing of the object to be processed; an anomaly evaluation module, used to evaluate anomalies based on processing data and context information to obtain anomaly evaluation results; and a database update module, used to update the initial sample database based on the anomaly evaluation results, processing data, and context information to obtain a target sample database, wherein the target sample database contains multiple training samples, the multiple training samples are used to train the object processing model, the object processing model is used to parse the anomaly identification results, and construct the object processing information of the object to be processed.

[0159] According to an embodiment of this application, another object processing apparatus for implementing the above-described object processing method is also provided. Figure 9 This is a structural block diagram of another object processing apparatus according to an embodiment of this application, such as... Figure 9 As shown, the device includes: a first display module 902 and a second display module 904.

[0160] The first display module 902 is used to respond to input commands applied to the operation interface and display the context information of the object to be processed on the operation interface. The object to be processed is used to represent an object with an anomaly, and the context information is used to represent information related to the anomaly. The second display module 904 is used to respond to processing commands applied to the operation interface and display object processing information on the operation interface. The object processing information is obtained by parsing the anomaly identification result. The anomaly identification result is obtained by identifying anomalies in the object to be processed through the context information. The anomaly identification result is used to reflect the reason for the anomaly.

[0161] The first display module 902 and the second display module 904 described above correspond to steps S702 to S704 in the above embodiments. The two modules and their corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in the above embodiments. The modules or units described above may be hardware or software components stored in memory and processed by one or more processors. The modules may also be part of a device and may run in the server 10 provided in the above embodiments.

[0162] The preferred embodiments involved in the above embodiments of this application are the same as the solutions, application scenarios and implementation processes provided in the above embodiments, and will not be repeated here.

[0163] Embodiments of this application may provide a computing device. Figure 10 This is a structural block diagram of a computing device according to an embodiment of this application. Figure 10As shown, the computing device 1000 may include: one or more (one shown in the figure) processors 1002, memory 1004, memory controller, and peripheral interfaces.

[0164] The aforementioned computing device can be understood as an integrated intelligent terminal, including but not limited to servers, desktop computers, PCs (Personal Computers), and all-in-one model machines. Furthermore, the computing device may have the model described in the above embodiments of this application pre-installed.

[0165] Specifically, this computing device can pre-install various types of models, including but not limited to models in fields such as natural language processing, visual processing, speech processing, code processing, and multimodal task processing, thus providing diverse model choices. In different product forms, this computing device can support one or more model usage methods, including but not limited to model training, model invocation, model fine-tuning, model deployment, model inference, and application. In some product forms, this computing device also supports model management, including but not limited to multi-type model management (supporting the management of discriminative, generative, and other model types), model version control (supporting the control of different model versions), and model evaluation (evaluating model performance and effectiveness based on model evaluation tools). In other product forms, this computing device can also create applications based on models, providing API calling capabilities. Models can be called into created applications through API interfaces, and application management tools are provided to control and manage applications.

[0166] Furthermore, this computing device can also include data management (supporting the creation and management of model tuning datasets), a training center (providing abundant training resources to help users learn and master AI technologies), and basic control capabilities (providing enterprise-level basic control capabilities to ensure system security and efficient operation). Through these functions, it provides a comprehensive, integrated device for AI development, training, deployment, and application.

[0167] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and apparatus in the embodiments of this application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby implementing the methods in the above embodiments. The memory may include high-speed random access memory (RAM) and non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to terminal A via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks (LANs), mobile communication networks, and combinations thereof.

[0168] The processor can invoke an executable program stored in memory via a transmission device to execute any of the methods described in the above embodiments.

[0169] Embodiments of this application may provide an electronic device. Figure 11 This is a structural block diagram of an electronic device according to an embodiment of this application. Figure 11 As shown, the electronic device may include: an input / output device 1102; a memory 1104; and a processor 1106, wherein the processor 1106 is connected to the input / output device 1102 and the memory 1104 via a bus 1108.

[0170] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and apparatus in the embodiments of this application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby implementing the methods in the above embodiments. The memory may include high-speed random access memory (RAM) and non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to terminal A via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks (LANs), mobile communication networks, and combinations thereof.

[0171] The processor can invoke an executable program stored in memory via a transmission device to execute any of the methods described in the above embodiments.

[0172] Those skilled in the art will understand that, Figure 11 The structure shown is illustrative. Electronic devices can also be smartphones (such as Android phones, iOS phones, etc.), tablets, PDAs, mobile internet devices (MIDs), PADs, and other terminal devices. Figure 11 This does not limit the structure of the aforementioned electronic device. For example, the electronic device may include more or fewer components (such as network interfaces, display devices, etc.) than those shown in the figure, or have a different configuration than that shown in the figure.

[0173] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: a flash drive, read-only memory (ROM), random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0174] Embodiments of this application also provide a computer-readable storage medium. Optionally, in this embodiment, the aforementioned computer-readable storage medium can be used to store program code executed by the method provided in the above embodiments.

[0175] Optionally, in this embodiment, the storage medium may be located in a computing device or an electronic device.

[0176] Optionally, in this embodiment, the computer-readable storage medium is configured to store an executable program. When the executable program runs, it controls the device where the computer-readable storage medium is located to perform any of the methods described in the above embodiments.

[0177] Embodiments of this application also provide a computer program product. Optionally, in this embodiment, the computer program product may include a computer program. When executed by a processor, the computer program implements the methods provided in the above embodiments.

[0178] Embodiments of this application also provide a computer program product. Optionally, the computer program product may include a non-volatile computer-readable storage medium. The non-volatile computer-readable storage medium can be used to store a computer program. When the computer program is executed by a processor, it implements the method provided in the above embodiments.

[0179] Embodiments of this application also provide a computer program. Optionally, in this embodiment, when the computer program is executed by a processor, it implements the method provided in the above embodiments.

[0180] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0181] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are illustrative; for example, the division of units is a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined, integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling, direct coupling, or communication connection shown or discussed may be through some interfaces, indirect coupling of units or modules, or communication connection, and may be electrical or other forms.

[0182] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of this embodiment.

[0183] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0184] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0185] The above are preferred embodiments of this application. For those skilled in the art, various improvements and modifications can be made without departing from the principles of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. An object processing method, characterized in that, include: Obtain the context information of the object to be processed, wherein the object to be processed is used to represent an object with an anomaly, and the context information is used to represent information that is associated with the anomaly; Based on the context information, anomaly identification is performed on the object to be processed to obtain anomaly identification results, wherein the anomaly identification results are used to reflect the reason for the existence of the anomaly; The anomaly identification results are parsed to construct object processing information for the object to be processed; The object to be processed is processed based on the object processing information.

2. The method according to claim 1, characterized in that, The context information includes: text information and image information; the anomaly identification of the object to be processed based on the context information, to obtain the anomaly identification result, includes: The text information and the image information are encoded respectively to obtain text encoding results and image encoding results; Based on the text encoding result and the image encoding result, the text words contained in the text information are semantically matched with the image blocks contained in the image information to obtain a semantic matching result, wherein the semantic matching result is used to characterize the association relationship between the semantic features of the text information and the semantic features of the image information; The anomaly type is determined based on the semantic matching results; The anomaly is attributed based on the anomaly type and the context information to obtain the anomaly identification result.

3. The method according to claim 1, characterized in that, The process of processing the object to be processed based on the object processing information includes: Based on the object processing information, a processing work order corresponding to the object to be processed is generated, wherein the processing work order is used to track the processing process of the object to be processed; Output the processing work order and control the object to be processed to enter the control state, wherein the control state is used to represent the control of interactive behaviors that are related to the object to be processed; In response to detecting a change in the object features of the object to be processed, the changed object features are matched with the information features of the object processing information to obtain a feature matching result; In response to the feature matching result indicating that the changed object features match the information features, it is determined that the processing of the object to be processed is successful, and the object to be processed is controlled to exit the control state.

4. The method according to claim 3, characterized in that, The step of generating a processing work order corresponding to the object to be processed based on the object processing information includes: The confidence level of the context information is evaluated to obtain the confidence evaluation result; In response to the confidence assessment result indicating that the confidence level is greater than or equal to a preset threshold, the processing work order is generated based on the work order generation module and the object processing information; In response to the confidence assessment result indicating that the confidence level is less than a preset threshold, the context information and the object processing information are output, the information review result corresponding to the object processing information is received, and the processing work order is generated based on the information review result and the work order generation module.

5. The method according to claim 3, characterized in that, The method further includes: In response to the detection that the object features of the object to be processed have not changed within a preset time period, or that the feature matching result indicates that the changed object features do not match the information features, a processing stage for processing the object to be processed is determined based on the processing work order. Obtain the object management method corresponding to the processing stage; The object management method is executed based on the context information to handle the exception.

6. The method according to claim 1, characterized in that, The process of obtaining the context information of the object to be processed includes: Obtain the interaction materials of the first entity in response to the abnormal submission, wherein the first entity is used to characterize the entity that has a need for the object to be processed; Obtain communication materials generated by the first entity and the second entity in response to the anomaly, wherein the second entity is used to represent the entity that provided the object to be processed to the first entity; Obtain the object material of the object to be processed; The context information is obtained by summarizing the interactive materials, the communication materials, and the object materials.

7. The method according to claim 1, characterized in that, The step of parsing the anomaly identification result and constructing the object processing information of the object to be processed includes: Obtain the object type of the object to be processed; Based on the exception type and the object type, a processing information template is invoked, wherein the processing information template is used to indicate the constraints for constructing the object processing information; Based on the anomaly identification result and the context information, the object processing information is generated according to the processing information template.

8. The method according to any one of claims 1-7, characterized in that, The method further includes: Obtain the processing data generated during the processing of the object to be processed; The anomaly is evaluated based on the processed data and the context information to obtain an anomaly evaluation result; The initial sample database is updated based on the anomaly assessment results, the processing data, and the context information to obtain a target sample database. The target sample database contains multiple training samples, which are used to train an object processing model. The object processing model is used to parse the anomaly identification results and construct object processing information for the object to be processed.

9. An object processing method, characterized in that, include: In response to an input command applied to the operation interface, the context information of the object to be processed is displayed on the operation interface, wherein the object to be processed is used to represent an object with an anomaly, and the context information is used to represent information that is associated with the anomaly; In response to a processing instruction applied to the operation interface, object processing information is displayed on the operation interface, wherein the object to be processed is processed, and the object processing information is obtained by parsing the anomaly identification result. The anomaly identification result is obtained by identifying anomalies in the object to be processed through the context information, and the anomaly identification result is used to reflect the reason for the existence of the anomaly.

10. An electronic device, characterized in that, include: Memory, which stores executable programs; A processor, connected to a memory via a bus, is used to run the program, wherein the program, when running, executes the method described in any one of claims 1 to 9.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein, when the executable program is executed, it controls the device on which the computer-readable storage medium is located to perform the method according to any one of claims 1 to 9.

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

Images