Object label processing method and device, electronic equipment and storage medium

Abstract

The invention discloses an object label processing method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a current object, carrying out the multi-dimensional label recognition of the current object, obtaining a to-be-processed label set of the current object, carrying out the addition and deletion processing of labels in the to-be-processed label set according to label addition and deletion reference information, and obtaining a target label set, the label addition and deletion reference information is determined based on the identification accuracy of the labels of the historical objects in the historical object set in a preset historical time period, the reliability of the label addition and deletion reference information meets the preset reliability, and business processing corresponding to the preset business is carried out on the current object based on the target label set. According to the embodiment of the invention, the continuously updated label addition and deletion reference information can be obtained through a small amount of manual auditing data in the historical time period, the labels of the current object are filtered and screened, the accuracy and recall rate of the labels are improved, and high-frequency training of a label system is realized without a large amount of labels.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an object tag processing method, device, electronic device and storage medium. Background Art

[0002] The process of labeling videos is becoming increasingly important because the accuracy of labeling will affect downstream applications.

[0003] Generally speaking, label recognition models can be used to identify labels for videos. However, massive amounts of videos appear every day, and the focus of the videos is also different. Therefore, a large number of manually reviewed labels are required every day to train and update the label recognition models. Therefore, a lot of manpower and material resources are spent on data labeling tasks every day, which is time-consuming and labor-intensive. Summary of the invention

[0004] In order to solve the problems of the prior art, the embodiments of the present invention provide an object tag processing method, device, electronic device and storage medium. The technical solution is as follows:

[0005] In one aspect, a method for processing an object label is provided, the method comprising:

[0006] Get the current object;

[0007] Perform multi-dimensional label recognition on the current object to obtain a label set to be processed for the current object;

[0008] Adding and deleting tags in the tag set to be processed is performed according to the tag addition and deletion reference information to obtain a target tag set; the tag addition and deletion reference information is determined based on the recognition accuracy of the tags of the historical objects in the historical object set within a preset historical period; the reliability of the tag addition and deletion reference information meets the preset reliability;

[0009] Perform business processing corresponding to the preset business on the current object based on the target tag set.

[0010] In another aspect, an object label processing device is provided, the device comprising:

[0011] Object acquisition module, used to obtain the current object;

[0012] The label recognition module is used to perform multi-dimensional label recognition on the current object to obtain a label set to be processed for the current object;

[0013] The tag processing module is used to perform tag addition and deletion processing on tags in the tag set to be processed according to the tag addition and deletion reference information to obtain a target tag set; the tag addition and deletion reference information is determined based on the recognition accuracy of the tags of the historical objects in the historical object set within a preset historical period; the reliability of the tag addition and deletion reference information meets the preset reliability;

[0014] The tag application module is used to perform business processing corresponding to the preset business on the current object based on the target tag set.

[0015] In some possible embodiments, the device further includes a tag addition and deletion reference information determination module, which is used to:

[0016] Obtain a historical object set for a preset historical period and a historical label set corresponding to the historical object set; the historical label set includes a sub-label set for each historical object;

[0017] If the historical tag set includes a first tag whose recognition accuracy is less than or equal to a first preset threshold, adding the first tag to the first deletion reference information in the tag addition and deletion reference information;

[0018] Tag processing module for:

[0019] Obtaining first deletion reference information;

[0020] If the to-be-processed tag set includes the first tag, the first tag is deleted from the to-be-processed tag set according to the first deletion reference information to obtain a target tag set.

[0021] In some possible embodiments, each tag in the sub-tag set of each historical object carries tag identification source information;

[0022] The tag addition and deletion reference information determination module is used to:

[0023] The first tags in the historical tag set are grouped according to the tag identification source information to obtain at least one first tag group; the first tags in each first tag group correspond to the same tag identification source information;

[0024] determining the recognition accuracy of each first tag group;

[0025] Acquire the tag identification source information to be marked according to the identification accuracy of each first tag group; the tag identification source information to be marked is the tag identification source information of the first tag group whose identification accuracy is less than or equal to the first preset threshold;

[0026] The first label, the label identification source information to be labeled, and the relationship information between the first label and the label identification source information to be labeled are added to the first deletion reference information in the label addition and deletion reference information.

[0027] In some possible embodiments, each tag in the to-be-processed tag set of the current object carries tag identification source information;

[0028] Tag processing module for:

[0029] If the tag set to be processed includes the first tag, and the tag identification source information carried by the first tag in the tag set to be processed is the same as the tag identification source information carried by the first tag in the first deletion reference information, the first tag is deleted from the tag set to be processed to obtain the target tag set.

[0030] In some possible embodiments, the device further includes a tag addition and deletion reference information determination module, which is used to:

[0031] Perform label correction processing on each sub-label set in the historical label set to obtain a processed historical label set;

[0032] Determine multiple label pairs according to the labels in each sub-label set in the processed historical label set; wherein each label pair includes two labels, and the two labels included in each label pair are labels included in the same sub-label set; each label pair carries a positive correlation mapping relationship between the two labels;

[0033] Determine a positive correlation mapping probability value for each label pair in a plurality of label pairs; the positive correlation mapping probability value represents a probability value of the second label existing in a sub-label set of the first label; the first label and the second label are labels in the same label pair;

[0034] The tag pairs whose positive correlation mapping probability values ​​are greater than or equal to the second preset threshold are added into the first added reference information in the tag addition and deletion reference information.

[0035] In some possible embodiments, the tag addition and deletion reference information determination module is used to:

[0036] For each of multiple label pairs, execute:

[0037] Determine the tag pair currently being executed as the current tag pair;

[0038] Based on the first tag in the current tag pair, the tags in each sub-tag set in the processed historical tag set are traversed to determine the first historical tag set from the processed historical tag set; the first historical tag set is a set of sub-tag sets including the first tag;

[0039] Based on the second tag in the current tag pair, the tags in each sub-tag set in the first historical tag set are traversed to determine the second historical tag set from the first historical tag set; the second historical tag set is a set of sub-tag sets containing the second tag;

[0040] The positive correlation mapping probability value is determined according to the number of sub-tag sets included in the second historical tag set and the number of sub-tag sets included in the first historical tag set.

[0041] In some possible embodiments, the tag processing module is used to:

[0042] Traverse each tag pair in the first added reference information;

[0043] If in the first added reference information, the first tag in the first tag pair is the second tag included in the tag set to be processed, and the tag set to be processed does not include the second tag in the first tag pair, the second tag in the first tag pair is added to the tag set to be processed to obtain the target tag set.

[0044] In some possible embodiments, the device further includes a tag addition and deletion reference information determination module, which is used to:

[0045] Determine multiple original label pairs according to the labels in each sub-label set in the historical label set; wherein each original label pair contains two labels, and the two labels contained in each original label pair are labels contained in the same sub-label set; each original label pair carries a positive correlation mapping relationship between the two labels;

[0046] Perform label correction processing on each sub-label set in the historical label set to obtain a processed historical label set;

[0047] Determine multiple corrected label pairs according to the labels in each sub-label set in the processed historical label set; wherein each corrected label pair includes two labels, and the two labels included in each corrected label pair are labels included in the same sub-label set in the processed historical label set; each corrected label pair carries a positive correlation mapping relationship between the two labels;

[0048] Determine a first positive correlation mapping probability value for each original label pair based on the historical label set;

[0049] Determine a second positive correlation mapping probability value for each corrected label pair based on the processed historical label set;

[0050] Determining the mapping accuracy of each preset label pair based on the second positive correlation mapping probability value and the first positive correlation mapping probability value of each preset label pair; the preset label pair is a label pair that exists in both the original label pair and the corrected label pair;

[0051] The preset tag pairs whose mapping accuracy is less than or equal to the third preset threshold are added into the second deletion reference information in the tag addition and deletion reference information.

[0052] In some possible embodiments, the tag processing module is used to:

[0053] Traversing each label pair in the second deletion reference information;

[0054] If in the second deletion reference information, the first tag in the second tag pair is the third tag included in the tag set to be processed, and the tag set to be processed includes the fourth tag corresponding to the second tag in the second tag pair, the fourth tag is deleted from the tag set to be processed to obtain the target tag set.

[0055] In some possible embodiments, the tag addition and deletion reference information determination module is used to:

[0056] Get the preset object set for the preset historical period;

[0057] The preset object set is sampled based on at least one of the account information, type information, and tag quantity information corresponding to each object in the preset object set to obtain a historical object set.

[0058] In some possible embodiments, the tag identification module is used to:

[0059] Using multiple label recognition sources to perform multi-dimensional label recognition on the current object, and obtaining a label set to be processed for the current object;

[0060] Each tag in the tag set to be processed carries tag identification source information.

[0061] On the other hand, an electronic device is provided, including a processor and a memory, wherein at least one instruction or at least one program is stored in the memory, and the at least one instruction or at least one program is loaded and executed by the processor to implement the above-mentioned object tag processing method.

[0062] On the other hand, a computer-readable storage medium is provided, in which at least one instruction or at least one program is stored, and the at least one instruction or at least one program is loaded and executed by a processor to implement the object label processing method as described above.

[0063] On the other hand, a computer program product or a computer program is provided, the computer program product or the computer program comprising computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the above-mentioned object tag processing method.

[0064] The embodiment of the present invention obtains the current object, performs multi-dimensional label recognition on the current object, obtains the label set to be processed of the current object, performs label addition and deletion processing on the labels in the label set to be processed according to label addition and deletion reference information, obtains the target label set, the label addition and deletion reference information is determined based on the recognition accuracy of the labels of the historical objects in the historical object set within a preset historical period, the reliability of the label addition and deletion reference information meets the preset reliability, and performs business processing corresponding to the preset business on the current object based on the target label set. The embodiment of the present application filters and screens the labels of the current object through label addition and deletion reference information with very high reliability, such as the continuously updated label addition and deletion reference information obtained through a small amount of manual review data in the historical period, to improve the accuracy and recall rate of the labels, without the need to implement high-frequency training of the label system through a large amount of annotation, thereby further increasing the accuracy of the labels while saving a large amount of manpower and material resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0066] Figure 1 is a schematic diagram of an implementation environment provided by an embodiment of the present invention;

[0067] Figure 2 is a flow chart of an object label processing method provided by an embodiment of the present invention;

[0068] Figure 3 is a flowchart of a method for determining first deletion reference information provided by an embodiment of the present invention;

[0069] Figure 4 is a flowchart of a first method for determining added reference information provided by an embodiment of the present invention;

[0070] Figure 5 It is a flow chart of a method for determining a positive correlation mapping probability value provided by an embodiment of the present invention;

[0071] Figure 6 is a flow chart of a second method for determining reference information for deletion provided by an embodiment of the present invention;

[0072] Figure 7 is a schematic diagram of an object label processing process provided by an embodiment of the present invention;

[0073] Figure 8is a schematic diagram of an object label processing process provided by an embodiment of the present invention;

[0074] Fig. 9 is a structural block diagram of an object label processing device provided by an embodiment of the present invention;

[0075] Fig.10 The present invention provides a hardware structure block diagram of an electronic device. DETAILED DESCRIPTION

[0076] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0077] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0078] It is understandable that in the specific implementation of this application, related data such as user information is involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0079] In order to facilitate understanding of the above-mentioned technical solutions and the technical effects produced by the embodiments of the present disclosure, a brief introduction is given to the terms involved in the embodiments of the present disclosure:

[0080] Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that machines have the functions of perception, reasoning and decision-making.

[0081] Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics and other technologies. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0082] Computer Vision (CV) is a science that studies how to make machines "see". To put it more specifically, it refers to machine vision such as using cameras and computers to replace human eyes to identify and measure targets, and further perform graphic processing so that computer processing becomes an image that is more suitable for human eye observation or transmission to instrument detection. As a scientific discipline, computer vision studies related theories and technologies, and attempts to establish an artificial intelligence system that can obtain information from images or multi-dimensional data. Computer vision technology usually includes image processing, image recognition, image semantic understanding, image retrieval, OCR, video processing, video semantic understanding, video content / behavior recognition, three-dimensional object reconstruction, 3D technology, virtual reality, augmented reality, simultaneous positioning and map construction, and other technologies, as well as common biometric recognition technologies such as face recognition and fingerprint recognition.

[0083] Natural language processing (NLP) is an important direction in the fields of computer science and artificial intelligence. It studies various theories and methods that can achieve effective communication between people and computers using natural language. Natural language processing is a science that integrates linguistics, computer science, and mathematics. Therefore, research in this field will involve natural language, that is, the language people use in daily life, so it is closely related to the study of linguistics. Natural language processing technology usually includes text processing, semantic understanding, machine translation, robot question answering, knowledge graph and other technologies.

[0084] Machine Learning (ML) is a multi-disciplinary subject that involves probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all areas of artificial intelligence. Machine learning and deep learning usually include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and self-learning.

[0085] Autonomous driving technology usually includes high-precision maps, environmental perception, behavioral decision-making, path planning, motion control and other technologies. Autonomous driving technology has broad application prospects.

[0086] With the research and advancement of artificial intelligence technology, artificial intelligence technology has been studied and applied in many fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned driving, automatic driving, drones, robots, smart medical care, smart customer service, etc. I believe that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.

[0087] The embodiments of the present invention can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, smart transportation, assisted driving, etc.

[0088] Among them, cloud technology refers to a hosting technology that unifies hardware, software, network and other resources in a wide area network or local area network to realize data computing, storage, processing and sharing. Cloud technology is a general term for network technology, information technology, integration technology, management platform technology, application technology, etc. based on the cloud computing business model application, which can form a resource pool and be used on demand, flexible and convenient. Cloud computing technology will become an important support. The background services of the technical network system require a large amount of computing and storage resources, such as video websites, picture websites and more portal websites. With the high development and application of the Internet industry, in the future, each item may have its own identification mark, and all need to be transmitted to the background system for logical processing. Data of different levels will be processed separately. All kinds of industry data require strong system backing support, which can only be achieved through cloud computing.

[0089] Cloud storage is a new concept that extends and develops from the concept of cloud computing. A distributed cloud storage system (hereinafter referred to as storage system) refers to a storage system that uses cluster applications, grid technology, and distributed storage file systems to bring together a large number of storage devices of various types (storage devices are also called storage nodes) in the network to work together through application software or application interfaces to provide data storage and business access functions. At present, the storage method of the storage system is to create a logical volume. When creating a logical volume, a physical storage space is allocated for each logical volume. The physical storage space may be composed of disks of a storage device or several storage devices. The client stores data on a logical volume, that is, the data is stored on the file system. The file system divides the data into many parts, each of which is an object. The object contains not only data but also additional information such as data identification (ID, ID entity). The file system writes each object into the physical storage space of the logical volume, and the file system records the storage location information of each object. Therefore, when the client requests to access the data, the file system can allow the client to access the data according to the storage location information of each object. The process of the storage system allocating physical storage space to a logical volume is as follows: based on the estimated capacity of the objects stored in the logical volume (this estimate often has a large margin relative to the actual capacity of the objects to be stored) and the grouping of independent redundant disk arrays (RAID, Redundant array of Independent Disks), the physical storage space is divided into stripes in advance. A logical volume can be understood as a stripe, thereby allocating physical storage space to the logical volume.

[0090] In short, a database can be seen as an electronic filing cabinet - a place where electronic files are stored, and users can add, query, update, delete, and other operations on the data in the files. The so-called "database" is a collection of data that is stored together in a certain way, can be shared with multiple users, has as little redundancy as possible, and is independent of the application.

[0091] Blockchain is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0092] The underlying blockchain platform can include processing modules such as user management, basic services, smart contracts, and operational detection. Among them, the user management module is responsible for the identity information management of all blockchain participants, including maintaining public and private key generation (account management), key management, and the maintenance of the correspondence between the user's real identity and the blockchain address (authority management), etc., and, under authorization, supervises and audits the transactions of certain real identities and provides risk control rule configuration (risk control audit); the basic service module is deployed on all blockchain node devices to verify the validity of business requests, and records valid requests to storage after consensus is reached. For a new business request, the basic service first performs interface adaptation analysis and authentication processing (interface adaptation), and then encrypts the business information through the consensus algorithm (consensus management). The smart contract module is responsible for the registration and issuance of contracts, as well as contract triggering and contract execution. Developers can define the contract logic in a programming language and publish it to the blockchain (contract registration). According to the logic of the contract terms, the key or other events are called to trigger the execution and complete the contract logic. It also provides the function of contract upgrade and cancellation. The operation detection module is mainly responsible for the deployment, configuration modification, contract setting, cloud adaptation and real-time status visualization output of the product during the product release process, such as alarm, network status detection, node equipment health status detection, etc.

[0093] The platform product service layer provides the basic capabilities and implementation framework of typical applications. Developers can superimpose business features based on these basic capabilities to complete the blockchain implementation of business logic. The application service layer provides application services based on blockchain solutions for business participants to use.

[0094] See also Figure 1 , which is a schematic diagram of an implementation environment provided by an embodiment of the present invention, and the implementation environment may include a client 110, a server 120 and a database 130.

[0095] The client 110 and the server 120 as well as the server 120 and the database 130 may be connected and communicated via a network.

[0096] Among them, the client 110 includes but is not limited to mobile phones, computers, intelligent voice interaction devices, smart home appliances, car-mounted clients, aircraft, etc. An application with human-computer interaction function can be run in the client 110, and the application can launch virtual item distribution activities in different business scenarios, such as flash sales, lottery activities, reward activities for completing tasks, etc. The client 110 can obtain the current object, perform multi-dimensional tag recognition on the current object, obtain the tag set to be processed for the current object, add and delete tags in the tag set to be processed according to the tag addition and deletion reference information, and obtain the target tag set. The tag addition and deletion reference information is determined based on the recognition accuracy of the tags of the historical objects in the historical object set within the preset historical period, and the reliability of the tag addition and deletion reference information meets the preset reliability. The current object is processed according to the preset business based on the target tag set. The embodiments of the present application filter and screen the labels of the current object through highly reliable label addition and deletion reference information, such as continuously updated label addition and deletion reference information obtained through a small amount of manual review data in a historical period, to improve the accuracy and recall rate of the labels without the need for high-frequency training of the label system through a large amount of annotation. Therefore, while saving a lot of manpower and material resources, the accuracy of the labels is further increased.

[0097] The server 120 can obtain the current object, perform multi-dimensional label recognition on the current object, obtain the label set to be processed of the current object, perform label addition and deletion processing on the labels in the label set to be processed according to the label addition and deletion reference information, and obtain the target label set. The label addition and deletion reference information is determined based on the recognition accuracy of the labels of the historical objects in the historical object set within a preset historical period. The reliability of the label addition and deletion reference information meets the preset reliability. Based on the target label set, the business processing corresponding to the preset business is performed on the current object. The embodiment of the present application filters and selects the labels of the current object through label addition and deletion reference information with very high reliability, such as the continuously updated label addition and deletion reference information obtained through a small amount of manual review data in the historical period, to improve the accuracy and recall rate of the labels without the need to implement high-frequency training of the label system through a large amount of annotation. Therefore, while saving a lot of manpower and material resources, the accuracy of the labels is further increased.

[0098] The database 130 may include an in-memory database and a relational database. It should be noted that the servers, databases, nodes, etc. of the embodiments of the present invention may be independent physical servers, or they may be server clusters or distributed systems composed of multiple physical servers. They may also be cloud servers that provide basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), as well as big data and artificial intelligence platforms.

[0099] In an exemplary embodiment, the client 110, the server 120 and the database 130 can all be node devices in the blockchain system, and can share the acquired and generated information with other node devices in the blockchain system to achieve information sharing between multiple node devices. Multiple node devices in the blockchain system can be configured with the same blockchain, which is composed of multiple blocks, and the adjacent blocks have an association relationship, so that when the data in any block is tampered, it can be detected by the next block, thereby preventing the data in the blockchain from being tampered with, and ensuring the security and reliability of the data in the blockchain.

[0100] See also Figure 2 , which is a flow chart of an object label processing method provided by an embodiment of the present invention, which can be applied to Figure 1 In the implementation environment shown, the execution subject of this method can be Figure 1 The server that executes clustering result determination can also be the client or other server node that executes clustering result determination. It should be noted that this specification provides method operation steps as described in the embodiments or flow charts, but more or fewer operation steps may be included based on routine or non-creative labor. The order of steps listed in the embodiments is only one way of executing the steps among many orders, and does not represent the only order of execution. When the system or product is actually executed, it can be executed in sequence or in parallel (for example, in a parallel processor or multi-threaded processing environment) according to the method shown in the embodiments or drawings. Specifically, Figure 2 As shown, the method may include:

[0101] S201, obtaining the current object.

[0102] In the embodiment of the present application, the current object may be an object in an application or an object in a web page. Optionally, the application may be a short video application, a social application, a music application, etc.

[0103] The present application embodiment does not limit the current object, and optionally, the current object may be a text object, a picture object, a graphic object, an audio object, a video object, an audio-video object, or an audio-video-graphic object, etc. The following will be described using a video object as an example.

[0104] S203, performing multi-dimensional label recognition on the current object to obtain a label set to be processed for the current object.

[0105] In the embodiment of the present application, a label system (label recognition model) can be used to perform multi-dimensional label recognition on the current object to obtain a label set to be processed for the current object. The label system can include multiple label recognition sources, or in other words, the label system can include a multi-channel label recall subsystem.

[0106] This application does not limit the tag recognition sources included in the tag system. Optionally, the tag system may include a face tag recognition source, a search tag recognition source, a classification tag recognition source, a drama title tag recognition source, a song tag recognition source, a movie tag recognition source, etc. Each tag recognition source is responsible for recalling the corresponding vertical category tags.

[0107] In this way, the embodiment of the present application can use multiple tag identification sources to perform multi-dimensional tag identification on the current object to obtain the tags to be processed corresponding to each tag identification source. The tags to be processed corresponding to each tag identification source can be combined into a tag set to be processed for the current object. Among them, each tag in the tag set to be processed carries tag identification source information.

[0108] In the embodiment of the present application, the labeling system is constructed based on the above machine learning and deep learning.

[0109] S205, adding and deleting tags in the tag set to be processed according to the tag addition and deletion reference information to obtain a target tag set; the tag addition and deletion reference information is determined based on the recognition accuracy of tags of historical objects in the historical object set within a preset historical period; the reliability of the tag addition and deletion reference information meets the preset reliability.

[0110] In the embodiment of the present application, before adding or deleting tags in the tag set to be processed according to the tag adding or deleting reference information to obtain the target tag set, the tag adding or deleting reference information may be obtained first.

[0111] In the embodiment of the present application, the tag addition and deletion reference information can be determined according to the tags of the historical object set of the preset historical period. Optionally, the recognition accuracy represents the accuracy of the tag recognition result obtained by the tag system performing tag recognition on the historical object.

[0112] In some optional embodiments, the label addition and deletion reference information is determined based on a small amount of manually reviewed data for a preset historical period. That is, after label recognition is performed on historical objects for a preset historical period based on a label recognition model and the labels corresponding to the historical objects are obtained, the labels corresponding to the historical objects can be manually reviewed again to determine the label addition and deletion reference information.

[0113] Optionally, the reliability of the tag addition and deletion reference information represents the reliability of the method for determining the accuracy of the tag recognition result above. Since the recognition accuracy of the tag for determining the tag addition and deletion reference information is determined by manual review, the reliability of the tag addition and deletion reference information meets the preset reliability. Optionally, in order to ensure the accuracy of the manual review, the tag addition and deletion reference information is reviewed and determined by multiple manual reviewers.

[0114] Optionally, a historical object set of a preset historical period and a historical tag set corresponding to the historical object set may be obtained, wherein the historical tag set includes a sub-tag set of each historical object, or in other words, the sub-tag set of each historical object constitutes the historical tag set.

[0115] In an embodiment of the present application, the preset historical period may be pre-set, for example, the preset historical period may be the 10 days, the week, or the month before the current moment.

[0116] In an optional embodiment, the historical object set may include all objects that are tagged in a preset historical period, for example, the historical object set includes 10,000 historical videos. Optionally, multiple tag identification sources included in the tag system may be used to perform multi-dimensional tag identification on each historical object to obtain a sub-tag set for each historical object, and then the 10,000 sub-tag sets corresponding to the 10,000 historical videos may form a historical tag set.

[0117] In another optional embodiment, the historical object set may include some objects among all objects that are subject to tag identification within a preset historical period. Optionally, a preset object set for a preset historical period may be obtained, and the preset object set may be sampled based on at least one of the account information, type information, and tag quantity information corresponding to each object in the preset object set to obtain the historical object set. The preset object set is a collection of all objects that are subject to tag identification within a preset historical period.

[0118] Optionally, a random number of objects may be sampled from the historical objects in the preset object set, and the sampled historical objects may be used to form a historical object set. Optionally, preset information may be sampled from the historical objects in the preset object set, and the sampled objects may be used to form a historical object set.

[0119] In an optional embodiment, the preset object set may be sampled based on the account information corresponding to each object in the preset object set to obtain a historical object set. For example, historical objects corresponding to account information that meets the account condition from the account information to which the historical objects in the preset object set belong may be sampled to form a historical object set.

[0120] In the embodiment of the present application, satisfying the account condition may mean that the number of people following the account information exceeds a preset number, or satisfying the account condition may mean that the amount of information posted exceeds a preset amount, or satisfying the account condition may mean that the average number of comments exceeds a preset number of comments, and so on.

[0121] In another optional embodiment, the preset object set can be sampled based on the type information corresponding to each historical object in the preset object set to obtain a historical object set. Taking the video object as an example, some video objects can be slices of TV series, movies, and variety shows. Therefore, the preset object set can include historical objects whose type information is TV series, movies, and variety shows. On this basis, the historical objects whose type information satisfies TV series, movies, and variety shows can be extracted from the preset object set to form a historical object set.

[0122] In another optional embodiment, the preset object set can be sampled based on the label quantity information corresponding to each historical object in the preset object set to obtain the historical object set. Generally speaking, each historical object will obtain a certain number of labels after being identified by the label system. If there are historical objects in the preset object set whose label quantity is not within the preset quantity range, these historical objects can be sampled to form the historical object set.

[0123] In another optional embodiment, the sub-tag set of historical objects in the preset object set should have a certain tag, but the object that actually does not have the tag (such as the title of a play) can be extracted to form a historical object set. Because some tags that should have appeared but did not appear are due to errors in the tag recognition source in the tag system, the historical objects can be sampled for manual labeling and review.

[0124] Continuing with the above example, the preset object set includes 10,000 historical videos. Optionally, the multiple label recognition sources included in the label system can be used to perform multi-dimensional label recognition on each object in the preset object set to obtain a sub-label set for each historical object, that is, 10,000 sub-label sets corresponding to 10,000 historical videos. Then, sampling is performed based on the above content to obtain a historical label set consisting of 100 sub-label sets corresponding to 100 historical videos.

[0125] The embodiment of the present application forms a historical object set by sampling a preset object set. Since the historical objects in the historical object set are relatively small, only a small amount of annotation cost needs to be added when subsequently determining the tag addition and deletion reference information, which can greatly save workload and improve work efficiency.

[0126] In the embodiment of the present application, the tag addition and deletion reference information can be determined by using the recognition accuracy of the tags in the historical tag set.

[0127] Optionally, if the historical tag set includes a first tag whose recognition accuracy is less than or equal to a first preset threshold, the first tag is added to the first deletion reference information in the tag addition and deletion reference information. In this case, when further processing the tags in the tag set to be processed, the first deletion reference information can be obtained. If the tag set to be processed contains the first tag, the first tag is deleted from the tag set to be processed according to the first deletion reference information to obtain the target tag set.

[0128] See also Figure 3 , which is a flow chart of a first method for determining reference information for deletion provided by an embodiment of the present invention, comprising:

[0129] S301, grouping the first tags in the historical tag set according to the tag identification source information to obtain at least one first tag group; the first tags in each first tag group correspond to the same tag identification source information; each tag in the sub-tag set of each historical object carries the tag identification source information.

[0130] The following is an explanation of the historical label set containing 100 sub-label sets corresponding to 100 sampled historical videos, and the label system including 3 label recognition sources.

[0131] Assume that after each of the 100 historical videos is identified by the tag system, 10 tags can be obtained, that is, the sub-tag set of each historical video includes 10 tags, so that 100 historical videos can obtain 1000 tags. In some optional embodiments, there are no repeated tags in the 1000 tags, and the first tags in the historical tag set are 1000. In other optional embodiments, there are repeated tags in the 1000 tags, and the first tags in the historical tag set are less than 1000.

[0132] Since each tag in the subtag set of each historical object carries tag identification source information, the first tags in the historical tag set can be grouped according to the tag identification source information to obtain at least one first tag group, and the first tag in each first tag group corresponds to the same tag identification source information.

[0133] For example, 1000 tags contain 200 first tags "Sun Wukong". Among them, the first first tag group contains 180 "Sun Wukong", and the tag identification source information corresponding to the first tag in the first first tag group is tag identification source A. The second first tag group contains 20 "Sun Wukong", and the tag identification source information corresponding to the first tag in the second first tag group is tag identification source B.

[0134] S303: Determine the recognition accuracy of each first tag group.

[0135] From the above content, we can conclude that the labels in the historical label set corresponding to the sampled historical object set containing 100 historical objects are all obtained by label recognition by multiple label recognition sources in the label system. However, the recognition of the label system is not 100% accurate. Therefore, the labels in the 100 sub-label sets corresponding to the 100 historical objects can be manually reviewed to obtain 100 manual sub-label sets corresponding to the 100 historical objects.

[0136] Among them, in the 100 manual sub-label sets, the first first label group contains 170 "Sun Wukongs", and the label identification source information corresponding to the first label in the first first label group is label identification source A, that is, 170 of the 180 "Sun Wukongs" identified by the label system are accurate after manual review. The second first label group contains 15 "Sun Wukongs", and the label identification source information corresponding to the first label in the second first label group is label identification source B, that is, 15 of the 20 "Sun Wukongs" identified by the label system are accurate after manual review.

[0137] Thus, it can be obtained that the recognition accuracy of the first first tag group is 0.944, and the recognition accuracy of the second first tag group is 0.75.

[0138] S305, obtaining tag identification source information to be marked according to the identification accuracy of each first tag group; the tag identification source information to be marked is the tag identification source information of the first tag group whose identification accuracy is less than or equal to a first preset threshold.

[0139] The embodiment of the present application does not limit the first preset threshold. Optionally, it can be set according to the actual application situation. For example, the first preset threshold is 0.94. As can be seen from the above, if the recognition accuracy of the second first tag group is less than the first preset threshold, the tag recognition source information to be marked is tag recognition source B. Since the recognition accuracy of the first first tag group is 0.944, which is greater than the first preset threshold, tag recognition source A is not counted as the tag recognition source information to be marked.

[0140] S307: Add the first tag, the tag identification source information to be marked, and the relationship information between the first tag and the tag identification source information to be marked to the first deletion reference information in the tag addition and deletion reference information.

[0141] Optionally, the first label "Sun Wukong", the label identification source information to be marked "label identification source B", and the relationship information between the first label and the label identification source information to be marked can be added to the first deletion reference information in the label addition and deletion reference information.

[0142] Thus, it can be seen in the first deletion reference information that: tag identification source B: [Sun Wukong, other tags 1, other tags 2, ...]. In this way, tags with lower identification accuracy from tag identification source B are added to the first deletion reference information, so that tags identified by the tag system can be processed later according to the first deletion reference information to improve the accuracy of the output tags.

[0143] On the basis of determining the first deletion reference information, if the first tag "Sun Wukong" is included in the tag set to be processed, and the tag identification source information "tag identification source B" carried by the first tag in the tag set to be processed is the same as the tag identification source information carried by the first tag in the first deletion reference information, the first tag is deleted from the tag set to be processed to obtain the target tag set.

[0144] In some possible embodiments, assume that 100 sub-tag sets corresponding to 100 historical objects contain 1000 tags, and these 1000 tags contain 200 first tags "Sun Wukong". After manual review, it is determined that only 185 of the 1000 tags actually contain the first tag "Sun Wukong", and the recognition accuracy of the first tag "Sun Wukong" can be obtained to be 0.925, which is less than the first preset threshold value of 0.94. In this way, the first tag "Sun Wukong" can be added to the first deletion reference information in the signature addition and deletion reference information. After obtaining the pending tag set of the current object, if the pending tag set contains the first tag "Sun Wukong", the first tag can be deleted from the pending tag set according to the first deletion reference information to obtain the target tag set.

[0145] The tag included in the first deletion reference information in the embodiment of the present application indicates that the tag system or the tag identification source in the tag system does not meet the recognition accuracy requirement for the tag. Therefore, if there is a tag included in the first deletion reference information in the tag set to be processed, it needs to be deleted to improve the accuracy of the final output tag.

[0146] In an embodiment of the present application, the positive correlation mapping probability value of the tag pair in the historical tag set can be used to determine the tag addition and deletion reference information. Optionally, the positive correlation mapping probability value of the tag pair in the historical tag set can be the positive correlation mapping probability value of the tag pair in the historical tag set after the tag correction processing. Among them, the positive correlation mapping probability value represents the probability value of the second tag existing in the sub-tag set of the first tag.

[0147] See also Figure 4 , which is a flow chart of a first method for determining added reference information provided by an embodiment of the present invention, comprising:

[0148] S401, performing label correction processing on each sub-label set in the historical label set to obtain a processed historical label set.

[0149] Based on the above example, the label correction process in the embodiment of the present application can be manual review, that is, each sub-label set in the historical label set can be manually reviewed to obtain a processed historical label set.

[0150] S403, determining multiple label pairs according to the labels in each sub-label set in the processed historical label set; wherein each label pair includes two labels, and the two labels included in each label pair are labels included in the same sub-label set; and each label pair carries a positive correlation mapping relationship between the two labels.

[0151] Optionally, multiple tag pairs can be determined based on the tags in each sub-tag set of the processed historical tag set. For example, assuming that a sub-tag set in the processed historical tag set contains four tags, tag 1, tag 2, tag 3, and tag 4, then 12 tag pairs can be determined, namely (tag 1, tag 2), (tag 1, tag 3), (tag 1, tag 4), (tag 2, tag 1), (tag 2, tag 3), (tag 2, tag 4), (tag 3, tag 1), (tag 3, tag 2), (tag 3, tag 4), (tag 4, tag 1), (tag 4, tag 2), and (tag 4, tag 3). Similarly, tag pairs can be determined for other sub-tag sets in the processed historical tag set.

[0152] If there are different sub-tag sets that determine the same tag pair, only one can be retained, that is, the multiple tag pairs determined based on the tags in each sub-tag set in the processed historical tag set are all non-repeating tag pairs. And (tag 1, tag 2) and (tag 2, tag 1) are two different tag pairs, the positive correlation mapping relationship between (tag 1, tag 2) is the relationship between tag 1 and tag 2, and the positive correlation mapping relationship between (tag 2, tag 1) is the relationship between tag 2 and tag 1.

[0153] S405, determining a positive correlation mapping probability value of each tag pair in the plurality of tag pairs; the positive correlation mapping probability value represents a probability value of the second tag existing in the subtag set of the first tag; the first tag and the second tag are tags in the same tag pair.

[0154] See also Figure 5 , which is a flow chart of a method for determining a positive correlation mapping probability value provided by an embodiment of the present invention, and for each tag pair in a plurality of tag pairs, the following is performed:

[0155] S501: Determine the tag pair currently being executed as the current tag pair.

[0156] Based on the 100 sub-label sets obtained after the label correction processing in the above text, assuming that the number of label pairs determined based on the labels in each sub-label set in the processed historical label set is 1500, the same calculation can be performed on each of the 1500 label pairs.

[0157] S503, based on the first tag in the current tag pair, traverse the tags in each subtag set in the processed historical tag set, and determine the first historical tag set from the processed historical tag set; the first historical tag set is a set of subtag sets including the first tag.

[0158] Assuming that the current tag pair is (Sun Wukong, Zhu Bajie), we can traverse the tags in each sub-tag set in the processed historical tag set based on the first tag "Sun Wukong" in the tag pair (Sun Wukong, Zhu Bajie). If the tag "Sun Wukong" appears in 50 of the 100 sub-tag sets, then "Sun Wukong" exists in 50 of the 100 historical objects. In this way, these 50 sub-tag sets constitute the first historical tag set.

[0159] S505, based on the second tag in the current tag pair, traverse the tags in each subtag set in the first historical tag set, and determine the second historical tag set from the first historical tag set; the second historical tag set is a collection of subtag sets containing the second tag.

[0160] Next, based on the second tag "Zhu Bajie" in the tag pair (Sun Wukong, Zhu Bajie), the tags in each subtag set of the first historical tag set can be traversed. If the tag "Zhu Bajie" appears in 45 of the 50 subtag sets, then the 50 historical objects with "Sun Wukong" include 45 historical objects with "Zhu Bajie". In this way, these 45 subtag sets constitute the second historical tag set.

[0161] S507 : Determine a positive correlation mapping probability value according to the number of sub-tag sets included in the second historical tag set and the number of sub-tag sets included in the first historical tag set.

[0162] Optionally, the positive correlation mapping probability value may be determined according to the quotient of the number of sub-tag sets included in the second historical tag set and the number of sub-tag sets included in the first historical tag set. In the above example, the positive correlation mapping probability value is 0.9.

[0163] S407: Add the tag pairs whose positive correlation mapping probability values ​​are greater than or equal to the second preset threshold into the first added reference information in the tag addition and deletion reference information.

[0164] The embodiment of the present application does not limit the second preset threshold. Optionally, it can be set according to the actual application situation, for example, the second preset threshold is 0.9. As can be seen from the above, the positive correlation mapping probability value of (Sun Wukong, Zhu Bajie) is 0.9, which is equal to the second preset threshold. Therefore, the (Sun Wukong, Zhu Bajie) label pair can be added to the first added reference information in the label addition and deletion reference information.

[0165] On the basis of determining the first added reference information, each tag pair in the first added reference information is traversed. If in the first added reference information, the first tag in the first tag pair is the second tag included in the tag set to be processed, and the tag set to be processed does not include the second tag in the first tag pair, the second tag in the first tag pair is added to the tag set to be processed to obtain the target tag set.

[0166] Assuming that the first tag pair is the above (Sun Wukong, Zhu Bajie), and the tag set to be processed contains the second tag "Sun Wukong", and the tag set to be processed does not contain the second tag "Zhu Bajie" in the first tag pair, then the second tag "Zhu Bajie" in the first tag pair can be added to the tag set to be processed to obtain the target tag set. In this way, the number of tags in the tag set to be processed increases.

[0167] The tag pair included in the first added reference information in the embodiment of the present application indicates that if the first tag appears in the historical object, the second tag is likely to appear. Therefore, when the first tag in a tag pair exists in the tag set to be processed but the second tag is not, it is very likely that the second tag is missed. Therefore, the present application aims to fill in the gaps and improve the tags in the tag set to be processed by adding the missed second tag to the tag set to be processed.

[0168] In the embodiment of the present application, the mapping accuracy of the tag pairs in the historical tag set can be used to determine the tag addition and deletion reference information.

[0169] See also Figure 6, which is a flow chart of a second method for determining reference information for deletion provided by an embodiment of the present invention, comprising:

[0170] S601, determining multiple original label pairs according to the labels in each sub-label set in the historical label set; wherein each original label pair contains two labels, and the two labels contained in each original label pair are labels contained in the same sub-label set; each original label pair carries a positive correlation mapping relationship between the two labels.

[0171] In the embodiment of the present application, the implementation steps involved in S601 can refer to the implementation steps involved in step S403 above.

[0172] S603, performing label correction processing on each sub-label set in the historical label set to obtain a processed historical label set.

[0173] Based on the above example, the label correction process in the embodiment of the present application can be manual review, that is, each sub-label set in the historical label set can be manually reviewed to obtain a processed historical label set.

[0174] S605, determining multiple corrected label pairs according to the labels in each sub-label set in the processed historical label set; wherein each corrected label pair includes two labels, and the two labels included in each corrected label pair are labels included in the same sub-label set in the processed historical label set; each corrected label pair carries a positive correlation mapping relationship between the two labels.

[0175] In the embodiment of the present application, the implementation steps involved in S605 can refer to the implementation steps involved in step S403 above. Since the multiple original label pairs obtained in S601 are determined before manual review, and the multiple corrected label pairs obtained in S605 are determined after manual review. Therefore, the label pairs may appear as follows: in the first case, some of the original label pairs obtained in S601 and some of the corrected label pairs obtained in S605 are the same; in the second case, some corrected label pairs exist in S601 but do not exist in S605; in the third case, some corrected label pairs exist in S605 but do not exist in S601. The second and third cases occur because the manual review corrects the labels in the sub-label set in the historical label set (such as deleting and adding some labels).

[0176] S607 , determining a first positive correlation mapping probability value of each original tag pair based on the historical tag set.

[0177] In the embodiment of the present application, according to step S405 or Figure 5 The illustrated embodiment determines a first positive correlation mapping probability value for each original label pair based on the historical label set.

[0178] S609 , determining a second positive correlation mapping probability value for each corrected label pair based on the processed historical label set.

[0179] In the embodiment of the present application, according to step S405 or Figure 5 The illustrated embodiment determines a second positive correlation mapping probability value for each corrected label pair based on the processed historical label set.

[0180] S611, determining the mapping accuracy of each preset label pair based on the second positive correlation mapping probability value and the first positive correlation mapping probability value of each preset label pair; the preset label pair is a label pair existing in both the original label pair and the corrected label pair.

[0181] Optionally, a preset label pair may be determined from the original label pair and the corrected label pair. The preset label pair is a label pair that appears in both the original label pair and the corrected label pair, such as (Holmes, Watson).

[0182] Assuming that the first positive correlation mapping probability value of the preset label pair (Holmes, Watson) is 0.95, and the second positive correlation mapping probability value of the preset label pair (Holmes, Watson) is 0.095, the mapping accuracy of the preset label pair can be determined as 0.1 based on the quotient of the second positive correlation mapping probability value 0.095 and the first positive correlation mapping probability value 0.95 of the preset label pair.

[0183] S613: Add the preset tag pairs whose mapping accuracy is less than or equal to the third preset threshold into the second deletion reference information in the tag addition and deletion reference information.

[0184] The embodiment of the present application does not limit the third preset threshold. Optionally, it can be set according to the actual application situation, for example, the third preset threshold is 0.5. As can be seen from the above, the mapping accuracy of the preset label pair (Holmes, Watson) 0.1 is less than the third preset threshold. Therefore, the preset label pair (Holmes, Watson) can be added to the second deletion reference information in the label addition and deletion reference information.

[0185] In another optional embodiment, when the first positive correlation mapping probability value of each preset label pair is greater than or equal to a fourth preset threshold, the mapping accuracy of each preset label pair is determined based on the second positive correlation mapping probability value and the first positive correlation mapping probability value of the preset label pair, and the preset label pairs with a mapping accuracy less than or equal to the third preset threshold are added to the second deletion reference information in the label addition and deletion reference information.

[0186] The fourth preset threshold is not limited in the embodiment of the present application. Optionally, it can be set according to the actual application situation, for example, the fourth preset threshold is 0.9. That is, only when the first label appears in the preset label pair before manual review and the second label is very likely to appear, the second positive correlation mapping probability value of the preset label pair after manual review will be determined.

[0187] In an embodiment of the present application, after determining the second deletion reference information in the tag addition and deletion reference information, each tag pair in the second deletion reference information can be traversed. If in the second deletion reference information, the first tag in the second tag pair is the third tag included in the tag set to be processed, and the tag set to be processed contains the fourth tag corresponding to the second tag in the second tag pair, the fourth tag is deleted from the tag set to be processed to obtain the target tag set.

[0188] Assuming that the second label pair is the above-mentioned (Holmes, Watson), and the label set to be processed contains the third label "Holmes" in the second label pair, and the label set to be processed contains the fourth label "Watson" in the second label pair, then the fourth label "Watson" in the second label pair can be deleted from the label set to be processed to obtain the target label set.

[0189] The tag pair included in the second deletion reference information in the embodiment of the present application indicates that if the first tag appears in the historical object, the tag system is very likely to identify the second tag, but the second tag identified by the tag system may have a high probability of being recognized incorrectly. Therefore, when the first tag and the second tag in a preset tag pair exist in the tag set to be processed, the second tag can be deleted from the tag set to be processed.

[0190] To summarize, the present application can, after determining the first deletion reference information in the tag addition and deletion reference information, perform tag deletion processing on the pending tag set of the current object based on the first deletion reference information to obtain a first target tag set; after determining the first addition reference information in the tag addition and deletion reference information, perform tag addition processing on the first target tag set based on the first addition reference information to obtain a second target tag set; after determining the second deletion reference information in the tag addition and deletion reference information, perform tag deletion processing on the second target tag set based on the second deletion reference information to obtain a final target tag set.

[0191] Alternatively, after determining the first deletion reference information, the first addition reference information and the second deletion reference information in the tag addition and deletion reference information, the present application can perform tag addition and deletion processing on the tag set to be processed of the current object according to the tag addition and deletion reference information to obtain the final target tag set.

[0192] See also Figure 7, which is a schematic diagram of an object label processing process provided by an embodiment of the present invention. Figure 7 First, the current object, i.e., the first video, can be obtained. Then, the label multi-channel recall system can be used to perform multi-dimensional label recognition on the first video to obtain a label set to be processed for the current object. The label multi-channel recall system can include the following: Figure 7 The face tag recognition source, object retrieval tag recognition source, classification tag recognition source, drama title tag recognition source, song tag recognition source, movie tag recognition source, video retrieval tag recognition source, etc. are shown. Then, the tag addition and deletion reference information can be used to filter the tags of the tag set to be processed, including tag deletion and / or tag addition, so as to obtain the final target tag set including tag 1, tag 2, tag 3...tag N.

[0193] S207: Perform business processing corresponding to the preset business on the current object based on the target tag set.

[0194] In an embodiment of the present application, the current object may be processed for a preset service based on the target tag set of the current object, wherein the preset service may include a preset distribution service, a preset recommendation service, or a preset search service, etc.

[0195] See also Figure 8 , which is a schematic diagram of an object label processing process provided by an embodiment of the present invention. First, multiple objects can be obtained, and then, a label multi-channel recall system can be used to perform multi-dimensional label recognition on each object to obtain a label set to be processed for each object. Among them, the label multi-channel recall system can include the following: Figure 7 The face tag recognition source, object retrieval tag recognition source, classification tag recognition source, drama title tag recognition source, song tag recognition source, movie tag recognition source, video retrieval tag recognition source, etc. are shown. Then, the tag addition and deletion reference information can be used to filter the tags of the tag set to be processed for each object, including tag deletion and / or tag addition, to obtain the target tag set of each object, so that the object can be processed according to the preset business based on the target tag set of each object.

[0196] Optionally, before performing business processing corresponding to the preset business on the current object based on the target label set, manual review can be performed on the target label sets of some sampled objects. This manual review can not only improve the accuracy of the labels in the target label set of the current object, but also continuously update the label addition and deletion reference information to prepare for the label processing of future objects.

[0197] How to obtain label addition and deletion reference information in the embodiment of the present application includes: obtaining the system output label of each historical object in a preset historical period, and extracting the system output label of the historical object that needs manual review from the system output label of each historical object in the preset historical period. The specific extraction method can refer to the above implementation method and will not be repeated here. Then, the first deletion reference information, the first addition reference information, and the second deletion reference information can be determined based on the recognition accuracy of the system in the system output label of the manually reviewed historical object, so as to construct the label addition and deletion reference information, so that the label addition and deletion reference information can be reprocessed for the current object label recognition to increase the accuracy of the final label.

[0198] In summary, a small amount of manually reviewed data from historical periods can be used to obtain continuously updated reference information for label additions and deletions, to filter and screen the labels of the current object, such as a video, to improve the accuracy and recall of the labels, without the need for high-frequency training of the label system through a large amount of annotation.

[0199] See also Fig. 9 , which is a schematic diagram of the structure of an object tag processing device provided by an embodiment of the present invention, the device has the function of implementing the object tag processing method in the above method embodiment, and the function can be implemented by hardware or by hardware executing corresponding software. Fig. 9 As shown, the object tag processing device 900 may include:

[0200] The object acquisition module 901 is used to acquire the current object;

[0201] The tag recognition module 902 is used to perform multi-dimensional tag recognition on the current object to obtain a tag set to be processed for the current object;

[0202] The tag processing module 903 is used to perform tag addition and deletion processing on tags in the tag set to be processed according to the tag addition and deletion reference information to obtain a target tag set; the tag addition and deletion reference information is determined based on the recognition accuracy of tags of historical objects in the historical object set within a preset historical period; the reliability of the tag addition and deletion reference information meets the preset reliability;

[0203] The tag application module 904 is used to perform business processing corresponding to the preset business on the current object based on the target tag set.

[0204] In some possible embodiments, the device further includes a tag addition and deletion reference information determination module, which is used to:

[0205] Obtain a historical object set for a preset historical period and a historical label set corresponding to the historical object set; the historical label set includes a sub-label set for each historical object;

[0206] If the historical tag set includes a first tag whose recognition accuracy is less than or equal to a first preset threshold, adding the first tag to the first deletion reference information in the tag addition and deletion reference information;

[0207] Tag processing module for:

[0208] Obtaining first deletion reference information;

[0209] If the to-be-processed tag set includes the first tag, the first tag is deleted from the to-be-processed tag set according to the first deletion reference information to obtain the target tag set.

[0210] In some possible embodiments, each tag in the sub-tag set of each historical object carries tag identification source information;

[0211] The tag addition and deletion reference information determination module is used to:

[0212] The first tags in the historical tag set are grouped according to the tag identification source information to obtain at least one first tag group; the first tags in each first tag group correspond to the same tag identification source information;

[0213] determining the recognition accuracy of each first tag group;

[0214] Acquire the tag identification source information to be marked according to the identification accuracy of each first tag group; the tag identification source information to be marked is the tag identification source information of the first tag group whose identification accuracy is less than or equal to the first preset threshold;

[0215] The first label, the label identification source information to be labeled, and the relationship information between the first label and the label identification source information to be labeled are added to the first deletion reference information in the label addition and deletion reference information.

[0216] In some possible embodiments, each tag in the to-be-processed tag set of the current object carries tag identification source information;

[0217] Tag processing module for:

[0218] If the tag set to be processed includes the first tag, and the tag identification source information carried by the first tag in the tag set to be processed is the same as the tag identification source information carried by the first tag in the first deletion reference information, the first tag is deleted from the tag set to be processed to obtain the target tag set.

[0219] In some possible embodiments, the device further includes a tag addition and deletion reference information determination module, which is used to:

[0220] Perform label correction processing on each sub-label set in the historical label set to obtain a processed historical label set;

[0221] Determine multiple label pairs according to the labels in each sub-label set in the processed historical label set; wherein each label pair includes two labels, and the two labels included in each label pair are labels included in the same sub-label set; each label pair carries a positive correlation mapping relationship between the two labels;

[0222] Determine a positive correlation mapping probability value for each label pair in a plurality of label pairs; the positive correlation mapping probability value represents a probability value of the second label existing in a sub-label set of the first label; the first label and the second label are labels in the same label pair;

[0223] The tag pairs whose positive correlation mapping probability values ​​are greater than or equal to the second preset threshold are added into the first added reference information in the tag addition and deletion reference information.

[0224] In some possible embodiments, the tag addition and deletion reference information determination module is used to:

[0225] For each of multiple label pairs, execute:

[0226] Determine the tag pair currently being executed as the current tag pair;

[0227] Based on the first tag in the current tag pair, the tags in each sub-tag set in the processed historical tag set are traversed to determine the first historical tag set from the processed historical tag set; the first historical tag set is a set of sub-tag sets including the first tag;

[0228] Based on the second tag in the current tag pair, the tags in each sub-tag set in the first historical tag set are traversed to determine the second historical tag set from the first historical tag set; the second historical tag set is a set of sub-tag sets containing the second tag;

[0229] The positive correlation mapping probability value is determined according to the number of sub-tag sets included in the second historical tag set and the number of sub-tag sets included in the first historical tag set.

[0230] In some possible embodiments, the tag processing module is used to:

[0231] Traverse each tag pair in the first added reference information;

[0232] If in the first added reference information, the first tag in the first tag pair is the second tag included in the tag set to be processed, and the tag set to be processed does not include the second tag in the first tag pair, the second tag in the first tag pair is added to the tag set to be processed to obtain the target tag set.

[0233] In some possible embodiments, the device further includes a tag addition and deletion reference information determination module, which is used to:

[0234] Determine multiple original label pairs according to the labels in each sub-label set in the historical label set; wherein each original label pair contains two labels, and the two labels contained in each original label pair are labels contained in the same sub-label set; each original label pair carries a positive correlation mapping relationship between the two labels;

[0235] Perform label correction processing on each sub-label set in the historical label set to obtain a processed historical label set;

[0236] Determine multiple corrected label pairs according to the labels in each sub-label set in the processed historical label set; wherein each corrected label pair includes two labels, and the two labels included in each corrected label pair are labels included in the same sub-label set in the processed historical label set; each corrected label pair carries a positive correlation mapping relationship between the two labels;

[0237] Determine a first positive correlation mapping probability value for each original label pair based on the historical label set;

[0238] Determine a second positive correlation mapping probability value for each corrected label pair based on the processed historical label set;

[0239] Determining the mapping accuracy of each preset label pair based on the second positive correlation mapping probability value and the first positive correlation mapping probability value of each preset label pair; the preset label pair is a label pair that exists in both the original label pair and the corrected label pair;

[0240] The preset tag pairs whose mapping accuracy is less than or equal to the third preset threshold are added into the second deletion reference information in the tag addition and deletion reference information.

[0241] In some possible embodiments, the tag processing module is used to:

[0242] Traversing each label pair in the second deletion reference information;

[0243] If in the second deletion reference information, the first tag in the second tag pair is the third tag included in the tag set to be processed, and the tag set to be processed includes the fourth tag corresponding to the second tag in the second tag pair, the fourth tag is deleted from the tag set to be processed to obtain the target tag set.

[0244] In some possible embodiments, the tag addition and deletion reference information determination module is used to:

[0245] Get the preset object set for the preset historical period;

[0246] The preset object set is sampled based on at least one of the account information, type information, and tag quantity information corresponding to each object in the preset object set to obtain a historical object set.

[0247] In some possible embodiments, the tag identification module is used to:

[0248] Using multiple label recognition sources to perform multi-dimensional label recognition on the current object, and obtaining a label set to be processed for the current object;

[0249] Each tag in the tag set to be processed carries tag identification source information.

[0250] It should be noted that the device provided in the above embodiment, when implementing its functions, is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0251] In the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program with a predetermined function, and works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as processing circuits or memories) or a combination thereof. Similarly, a processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be part of an overall module or unit that includes the function of the module or unit.

[0252] An embodiment of the present invention provides an electronic device, which includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the object label processing method provided in the above method embodiment.

[0253] The memory can be used to store software programs and modules, and the processor executes various functional applications and clustering results by running the software programs and modules stored in the memory. The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, application programs required for functions, etc.; the data storage area may store data created according to the use of the device, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage device. Accordingly, the memory may also include a memory controller to provide the processor with access to the memory.

[0254] The method embodiments provided in the embodiments of the present invention can be executed in a computer terminal, a server or a similar computing device. Taking running on a server as an example, Fig.10 is a hardware structure block diagram of a server running an object tag processing method provided by an embodiment of the present invention, such as Fig.10 As shown, the server 2000 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 2010 (the processor 2010 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 2030 for storing data, and one or more storage media 2020 (such as one or more mass storage devices) for storing application programs 2023 or data 2022. Among them, the memory 2030 and the storage medium 2020 can be short-term storage or permanent storage. The program stored in the storage medium 2020 may include one or more modules, each of which may include a series of instruction operations on the server. Furthermore, the central processing unit 2010 can be configured to communicate with the storage medium 2020 and execute a series of instruction operations in the storage medium 220 on the server 2000. The server 2000 may also include one or more power supplies 2060, one or more wired or wireless network interfaces 2050, one or more input and output interfaces 2040, and / or one or more operating systems 2021, such as Windows ServerTM, Mac OSXTM, UnixTM, LinuxTM, FreeBSDTM, etc.

[0255] The input / output interface 2040 may be used to receive or send data via a network. The specific example of the network may include a wireless network provided by a communication provider of the server 2000. In one example, the input / output interface 2040 includes a network adapter (Network Interface Controller, NIC), which may be connected to other network devices via a base station so as to communicate with the Internet. In one example, the input / output interface 2040 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0256] It can be understood by those skilled in the art that Fig.10 The structure shown is only for illustration and does not limit the structure of the above electronic device. Fig.10 More or fewer components as shown, or with Fig.10 Different configurations are shown.

[0257] An embodiment of the present invention also provides a computer-readable storage medium, which can be set in an electronic device to store at least one instruction or at least one program related to implementing an object label processing method. The at least one instruction or the at least one program is loaded and executed by the processor to implement the object label processing method provided by the above method embodiment.

[0258] Optionally, in this embodiment, the above-mentioned storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store program codes.

[0259] It should be noted that the sequence of the embodiments of the present invention described above is for description only and does not represent the advantages and disadvantages of the embodiments. The above describes specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0260] An embodiment of the present invention further provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the above-mentioned object tag processing method.

[0261] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0262] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0263] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for object label processing, It is characterized in that The method comprises: Get the current object; Perform multi-dimensional label recognition on the current object to obtain a label set to be processed for the current object; Add or delete tags in the tag set to be processed according to the tag addition or deletion reference information to obtain a target tag set; the tag addition or deletion reference information is determined based on the recognition accuracy of tags of historical objects in the historical object set within a preset historical period; the reliability of the tag addition or deletion reference information meets the preset reliability; Based on the target tag set, business processing corresponding to the preset business is performed on the current object.

2. The object label processing method according to claim 1, It is characterized in that The method further comprises: Acquire a historical object set of the preset historical period and a historical tag set corresponding to the historical object set; the historical tag set includes a sub-tag set of each historical object; If the historical tag set includes a first tag whose recognition accuracy is less than or equal to a first preset threshold, adding the first tag to the first deletion reference information in the tag addition and deletion reference information; The adding and deleting of tags in the tag set to be processed is performed according to the tag adding and deleting reference information to obtain a target tag set, including: Acquire the first deletion reference information; If the to-be-processed tag set includes the first tag, the first tag is deleted from the to-be-processed tag set according to the first deletion reference information to obtain the target tag set.

3. The object label processing method according to claim 2, It is characterized in that Each tag in the sub-tag set of each historical object carries tag identification source information; If the historical tag set includes a first tag whose recognition accuracy is less than or equal to a first preset threshold, adding the first tag to the first deletion reference information in the tag addition and deletion reference information includes: The first tags in the historical tag set are grouped according to the tag identification source information to obtain at least one first tag group; the first tags in each first tag group correspond to the same tag identification source information; determining the recognition accuracy of each first tag group; Acquire the tag identification source information to be marked according to the identification accuracy of each first tag group; the tag identification source information to be marked is the tag identification source information of the first tag group whose identification accuracy is less than or equal to the first preset threshold; The first tag, the tag identification source information to be marked, and the relationship information between the first tag and the tag identification source information to be marked are added to the first deletion reference information in the tag addition and deletion reference information.

4. The object label processing method according to claim 3, It is characterized in that Each tag in the to-be-processed tag set of the current object carries the tag identification source information; If the to-be-processed tag set includes the first tag, deleting the first tag from the to-be-processed tag set to obtain the target tag set includes: If the first tag is included in the tag set to be processed, and the tag identification source information carried by the first tag in the tag set to be processed is the same as the tag identification source information carried by the first tag in the first deletion reference information, the first tag is deleted from the tag set to be processed to obtain the target tag set.

5. The object label processing method according to claim 2, It is characterized in that After obtaining the historical object set of the preset historical period and the historical tag set corresponding to the historical object set, the method further includes: Perform label correction processing on each sub-label set in the historical label set to obtain a processed historical label set; Determine a plurality of tag pairs according to the tags in each sub-tag set in the processed historical tag set; wherein each tag pair includes two tags, and the two tags included in each tag pair are tags included in the same sub-tag set; and each tag pair carries a positive correlation mapping relationship between the two tags; Determine a positive correlation mapping probability value for each tag pair in the plurality of tag pairs; the positive correlation mapping probability value represents a probability value of the second tag existing in the subtag set of the first tag; the first tag and the second tag are tags in the same tag pair; The tag pairs whose positive correlation mapping probability values ​​are greater than or equal to a second preset threshold are added into the first added reference information in the tag addition and deletion reference information.

6. The object label processing method according to claim 5, It is characterized in that The determining of a positive correlation mapping probability value of each tag pair in the plurality of tag pairs comprises: For each tag pair in the plurality of tag pairs, execute: Determine the tag pair currently being executed as the current tag pair; Based on the first tag in the current tag pair, traverse the tags in each sub-tag set in the processed historical tag set, and determine a first historical tag set from the processed historical tag set; the first historical tag set is a set of sub-tag sets including the first tag; Based on the second tag in the current tag pair, traverse the tags in each subtag set in the first historical tag set, and determine a second historical tag set from the first historical tag set; the second historical tag set is a set of subtag sets including the second tag; The positive correlation mapping probability value is determined according to the number of sub-tag sets included in the second historical tag set and the number of sub-tag sets included in the first historical tag set.

7. The object label processing method according to claim 5 or 6, It is characterized in that The adding and deleting of tags in the tag set to be processed is performed according to the tag adding and deleting reference information to obtain a target tag set, including: Traversing each tag pair in the first added reference information; If in the first added reference information, the first tag in the first tag pair is the second tag included in the tag set to be processed, and the tag set to be processed does not include the second tag in the first tag pair, the second tag in the first tag pair is added to the tag set to be processed to obtain the target tag set.

8. The object label processing method according to claim 2, It is characterized in that After obtaining the historical object set of the preset historical period and the historical tag set corresponding to the historical object set, the method further includes: Determine multiple original label pairs according to the labels in each sub-label set in the historical label set; wherein each original label pair includes two labels, and the two labels included in each original label pair are labels included in the same sub-label set; each original label pair carries a positive correlation mapping relationship between the two labels; Perform label correction processing on each sub-label set in the historical label set to obtain a processed historical label set; Determine a plurality of corrected label pairs according to the labels in each sub-label set in the processed historical label set; wherein each corrected label pair includes two labels, and the two labels included in each corrected label pair are labels included in the same sub-label set in the processed historical label set; each corrected label pair carries a positive correlation mapping relationship between the two labels; Determine a first positive correlation mapping probability value of each original label pair based on the historical label set; Determine a second positive correlation mapping probability value of each corrected label pair based on the processed historical label set; Determine the mapping accuracy of each preset label pair based on the second positive correlation mapping probability value and the first positive correlation mapping probability value of each preset label pair; the preset label pair is a label pair existing in both the original label pair and the corrected label pair; The preset label pairs whose mapping accuracy is less than or equal to the third preset threshold are added into the second deletion reference information in the label addition and deletion reference information.

9. The object label processing method according to claim 8, It is characterized in that The adding and deleting of tags in the tag set to be processed is performed according to the tag adding and deleting reference information to obtain a target tag set, including: Traversing each label pair in the second deletion reference information; If in the second deletion reference information, the first label in the second label pair is the third label included in the label set to be processed, and the label set to be processed contains the fourth label corresponding to the second label in the second label pair, delete the fourth label from the label set to be processed to obtain the target label set.

10. The object label processing method according to any one of claims 2-6, 8-9, It is characterized in that The step of obtaining the historical object set of the preset historical period includes: Obtaining a preset object set for the preset historical period; The preset object set is sampled based on at least one of the account information, type information, and tag quantity information corresponding to each object in the preset object set to obtain the historical object set.

11. The object label processing method according to any one of claims 1-6, 8-9, It is characterized in that The performing multi-dimensional label recognition on the current object to obtain a label set to be processed for the current object includes: Using multiple label recognition sources to perform multi-dimensional label recognition on the current object to obtain a label set to be processed for the current object; Each tag in the tag set to be processed carries tag identification source information.

12. An object label processing device, It is characterized in that The device comprises: Object acquisition module, used to obtain the current object; A tag recognition module, used to perform multi-dimensional tag recognition on the current object to obtain a tag set to be processed for the current object; A tag processing module, configured to perform tag addition and deletion processing on tags in the tag set to be processed according to tag addition and deletion reference information to obtain a target tag set; the tag addition and deletion reference information is determined based on the recognition accuracy of tags of historical objects in the historical object set within a preset historical period; the reliability of the tag addition and deletion reference information meets a preset reliability; The tag application module is used to perform business processing corresponding to the preset business on the current object based on the target tag set.

13. An electronic device, It is characterized in that It comprises a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the object label processing method according to any one of claims 1 to 11.

14. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the object tag processing method according to any one of claims 1 to 11.

Images