Document label management method and device

Abstract

The invention provides a document label management method and device. The method comprises the following steps: determining a target node from a document label tree; each node in the document label tree corresponds to one label; the preset document database stores documents corresponding to different tags; the number of documents corresponding to the target node in a preset document database meets a preset condition, and the document database is used for providing reference information for reasoning of the target language model; and adjusting the document label of the document corresponding to the target node.

Description

Technical Field

[0001] This application relates to document tag management technology, and more particularly to a document tag management method and apparatus. Background Technology

[0002] In the knowledge management system of an Artificial Intelligence Personal Computer (AIPC), the document tagging system serves as a core organizational structure, and document tag management is a crucial means of achieving efficient information organization and retrieval. By categorizing documents with tags, users can improve the efficiency and depth of their information retrieval, which is particularly significant in large-scale data scenarios. Existing document tagging systems are typically based on a pre-defined, fixed tagging framework, using machine learning or manual rules to classify and label documents for structured storage and management.

[0003] In existing technologies, common methods include flat multi-label classification and hierarchical text classification (HTC), the core of which lies in pre-defining a fixed set of label granularity. These labeling systems often pursue globally optimal performance during training, ignoring the uneven distribution of different document types in real-world applications. This leads to a severe bias in the model towards document types with larger sample sizes, resulting in significantly reduced labeling performance for documents with smaller sample sizes. This detaches from the actual data situation, leading to low accuracy in document label management. Summary of the Invention

[0004] This application provides a document tag management method and apparatus.

[0005] The technical solution of this application embodiment is implemented as follows: This application provides a document tag management method, the method including: The target node is determined from the document tag tree; each node in the document tag tree corresponds to a tag; a preset document database stores documents corresponding to different tags; the number of documents corresponding to the target node in the preset document database meets a preset condition, and the document database is used to provide reference information for the inference of the target language model; Adjust the document tags of the document corresponding to the target node.

[0006] In the above document tag management method, adjusting the document tags of the document corresponding to the target node includes: in response to the number of documents corresponding to the target node being greater than the first target number, obtaining multiple sub-tags corresponding to the target node; and adjusting the tags of the document corresponding to the target node based on the multiple sub-tags.

[0007] In the above document tag management method, obtaining multiple sub-tags corresponding to the target node includes at least one of the following: determining multiple sub-tags corresponding to the target node from a preset tag tree, wherein the preset tag tree has tags corresponding to each node in the document tag tree, and the number of tags in the preset tag tree is greater than that in the document tag tree; determining multiple sub-tags corresponding to the target node from the leaf nodes of the target node in the document tag tree; generating multiple sub-tags corresponding to the target node through the target model.

[0008] In the above document tag management method, after obtaining multiple sub-tags corresponding to the target node, the method also includes: adding leaf nodes to the target node based on the multiple sub-tags.

[0009] In the above document tag management method, adjusting the tags of the document corresponding to the target node based on multiple sub-tags includes: classifying the documents corresponding to the target node based on multiple sub-tags to obtain the document corresponding to each sub-tag; and adjusting the tag of the document corresponding to each sub-tag to the corresponding sub-tag.

[0010] In the above document tag management method, adjusting the document tags of the document corresponding to the target node includes: in response to the number of documents corresponding to the target node being less than the second target number, adjusting the tags of the document corresponding to the target node to the tags of the parent node corresponding to the target node.

[0011] In the above document tag management method, adjusting the tags of the document corresponding to the target node includes: in response to the total number of documents corresponding to all child nodes under the parent node corresponding to the target node being less than the third target number, adjusting the tags of all documents under the parent node corresponding to the target node to the tags of the parent node corresponding to the target node.

[0012] In the above document tag management method, determining the target node from the document tag tree includes: obtaining document quantity information corresponding to multiple nodes in the document tag tree according to a preset time period; and determining the target node from the multiple nodes based on the document quantity information.

[0013] The above document tag management method also includes: determining the first target quantity based on the total number of documents corresponding to the document tag tree.

[0014] This application provides a document tag management device, including: The determination module is used to identify the target node from the document tag tree; each node in the document tag tree corresponds to a tag; the preset document database stores documents corresponding to different tags; the number of documents corresponding to the target node in the preset document database meets the preset conditions; the document database is used to provide reference information for the inference of the target language model. The adjustment module is used to adjust the document tags of the document corresponding to the target node.

[0015] This application provides a document tag management device, *including: Memory is used to store executable instructions or computer programs. A processor, when executing computer-executable instructions or computer programs stored in memory, implements the methods provided in the embodiments of this application.

[0016] This application provides a computer-readable storage medium storing a computer program or computer-executable instructions, which, when executed by a processor, implements the document tag management method provided in this application.

[0017] This application provides a computer program product, including a computer program or computer executable instructions. When the computer program or computer executable instructions are executed by a processor, they implement the document tag management method provided in this application. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of a document tag management process provided in an embodiment of this application; Figure 2 This is an exemplary flowchart of adjusting document tags provided in an embodiment of this application. Figure 1 ; Figure 3 This is an exemplary flowchart of adjusting document tags provided in an embodiment of this application. Figure 2 ; Figure 4 This is a schematic diagram illustrating an exemplary process for determining a target node, provided in an embodiment of this application. Figure 5 This is a schematic flowchart of an exemplary label management method provided in an embodiment of this application; Figure 6 A schematic diagram of the structure of a document tag management device provided in an embodiment of this application; Figure 7 This is a schematic diagram of the structure of a document tag management device provided in an embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] This application provides a document tag management method, implemented by a document tag management device, such as... Figure 1As shown, the process includes the following steps S101 and S102: Step S101: Determine the target node from the document tag tree; each node in the document tag tree corresponds to a tag; the preset document database stores documents corresponding to different tags; the number of documents corresponding to the target node in the preset document database meets the preset conditions, and the document database is used to provide reference information for the inference of the target language model.

[0021] In the embodiments of this application, the document tag management device is a device with document tag management function. Exemplarily, the document tag management device can be a tablet computer, a laptop computer, a handheld computer, a personal digital assistant (PDA), a desktop computer, etc. The exemplary document tag management device is not limited here.

[0022] In the embodiments of this application, the document tag tree is a hierarchical tag organization structure. Each node in the document tag tree represents a document tag, and the tag corresponding to the parent node contains the tags corresponding to the child nodes. The document tag tree is used to build a multi-level management system for documents, helping users to retrieve and manage documents from different dimensions of tags.

[0023] In the embodiments of this application, the preset document database is a database that stores documents, each of which has a corresponding document tag. This document tag is the tag corresponding to a certain node in the preset document tree. The preset document database supports the target language model in quickly finding relevant document information during inference, serving as an auxiliary decision-making basis. The data in the preset document database is dynamically updated, changing as the tag structure of the document tag tree is adjusted. The target language model is one of the following: a Large Language Model (LLM), a multimodal large model, a model capable of generating text, a model with more than 100 million adjustable parameters, a model that can output corresponding results based on input natural language instructions, a model that can convert input natural language instructions into feature sequences and infer results based on the feature sequences, or a model that can convert input information into feature sequences and infer subsequent features of the feature sequences.

[0024] In the embodiments of this application, the target node is a node identified in the document tag tree as needing adjustment. The determination of the target node is based on whether the number of documents corresponding to the target node in the preset document database meets a preset condition. For example, the document tag management device periodically scans the leaf nodes in the document tag tree and counts the number of documents associated with each leaf node's tag in the preset document database. If the number of documents under a certain leaf node meets the preset condition, it is considered that the tags of the documents under that tag need to be adjusted, and the tag adjustment operation of the documents in the preset document database is triggered by identifying the target node.

[0025] In the embodiments of this application, the document quantity meeting the preset condition can be achieved by setting an upper and lower threshold (e.g., the upper threshold is set to 1000 documents, and the lower threshold is set to 200 documents), or it can be an upper or lower value set based on the number of documents corresponding to different tags in a preset tag database (e.g., the upper value is more than 100% of the average number of documents corresponding to different tags, and the lower value is less than 30% of the average number of documents corresponding to different tags). For example, if the number of documents under a certain leaf node exceeds the set upper threshold (upper value), it is considered that the documents corresponding to that tag are too abundant, and further classification can be performed to adjust the document tags; conversely, if the number of documents is lower than the lower threshold (lower value), it is considered that the documents under that tag are too few, the classification is too fragmented, and they need to be merged into the tags corresponding to its parent node.

[0026] Step S102: Adjust the document tags of the document corresponding to the target node.

[0027] In the embodiments of this application, the document tag management device will perform corresponding tag adjustment operations based on whether the number of documents corresponding to the target node meets preset conditions. Document tag adjustment refers to the process of dynamically modifying the document tags according to changes in the number of documents corresponding to the target node.

[0028] For example, when the number of documents is too large, the document tags are adjusted to match the tags of the child nodes of the target node; when the number of documents is too small, the document tags are adjusted to match the tags of the parent node of the target node. By applying the above document tag adjustment mechanism, adaptive optimization of the tag structure is achieved, thereby improving the efficiency of document management and the accuracy of retrieval.

[0029] In the embodiments of this application, if the number of documents in the target node exceeds the upper threshold (upper value), the document tag management device will perform a tag refinement operation. Tag refinement involves adding a new set of tags for child nodes under the target node and reclassifying the original target node's corresponding documents according to content characteristics into documents corresponding to each child node in the newly added set of child nodes. Tag refinement helps improve classification accuracy and retrieval efficiency, and is especially suitable for scenarios with a large number of documents and complex content. If the number of documents in the target node is below the lower threshold (lower value), the document tag management device will perform a tag merging operation. The document tag management device will merge the leaf nodes with fewer documents under the target node into the target node, canceling the independent document management status of the leaf nodes. The merging operation helps reduce tag fragmentation, avoid resource waste, and maintain the simplicity and consistency of the overall classification structure.

[0030] In the embodiments of this application, the document tag management device updates the structure of the document tag tree and ensures its integrity and consistency, preparing for the next round of monitoring and adjustments. The document tag device operates without manual intervention throughout the entire adjustment process, achieving automated and intelligent operation and maintenance of the tags.

[0031] In this way, by combining a tree-like tag structure with a pre-set document database, the tags of documents in the pre-set document database are adjusted in real time according to the number of documents corresponding to the target node, thereby achieving adaptive optimization of the tag system, improving document classification efficiency and retrieval accuracy, and reducing manual maintenance costs.

[0032] In some embodiments, when the document tag management device performs step S102 as described above, Figure 2 As shown, the following steps S201 and S202 can be performed: Step S201: In response to the number of documents corresponding to the target node being greater than the number of the first target, obtain multiple sub-tags corresponding to the target node.

[0033] In the embodiments of this application, the design of the first target number aims to avoid overgeneralization and ensure that the document tags are always in an optimal state. The first target number can be the upper limit number discussed above, which can be 120%, 150%, or other percentages exceeding 1, based on the average number of documents corresponding to different tags. It can be set based on actual needs and application scenarios, and this application does not limit it.

[0034] In the embodiments of this application, when the number of documents corresponding to a certain target node (i.e., a node in a tree structure) exceeds the first target number, it indicates that the number of documents corresponding to the target node is too large, and the tags corresponding to the target node are insufficient to effectively distinguish the content of the documents corresponding to the target node, thereby affecting retrieval efficiency and classification accuracy. At this time, the document tag management device will automatically trigger a mechanism to refine the tags of the documents corresponding to the target node, and obtain multiple sub-tags corresponding to the target node for further subdividing the documents corresponding to the target node.

[0035] In the embodiments of this application, the function of the tag is to serve as a container or identifier for the document management corresponding to the node. If the number of documents corresponding to the target node is too large, the target node will become a parent node in the document tag tree. The parent node corresponds to multiple child nodes, and each child node corresponds to a child tag. That is, the document tag management device can obtain multiple child tags corresponding to the target node.

[0036] For example, artificial intelligence can be a label corresponding to a target node, while natural language processing, computer vision, machine learning, etc., can be multiple sub-labels corresponding to the target node.

[0037] Step S202: Adjust the tags of the document corresponding to the target node based on multiple sub-tags.

[0038] In the embodiments of this application, after obtaining multiple sub-tags corresponding to the target node, the document tag management device can reclassify all documents corresponding to the target node based on these sub-tags. That is, the documents originally belonging to the target node are matched to the corresponding sub-tags according to the content of the documents corresponding to the target node. The process of reclassifying all documents corresponding to the target node based on sub-tags not only improves the accuracy of document classification for the target node, but also makes the entire document tagging system more structured and easier to maintain.

[0039] In the embodiments of this application, adjusting the labels of documents corresponding to target nodes can be achieved using text-based automatic classification algorithms, such as keyword matching, vector models, or deep learning models, to determine the most suitable sub-label for the document corresponding to the target node. Furthermore, user behavior data or feedback information can be combined to further optimize the label assignment results for documents corresponding to target nodes. For example, if a document corresponding to a target node is repeatedly assigned to a natural language processing sub-label, the adaptability of the natural language processing sub-label to the document corresponding to the target node can be enhanced.

[0040] In the embodiments of this application, the document tag management device can continuously evolve and optimize by dynamically responding to changes in the number and content of documents during operation, enabling it to have adaptive adjustment capabilities, reducing the need for manual intervention, and significantly improving the level of intelligence and long-term availability.

[0041] In this way, by responding to changes in the number of documents under the target node, multiple sub-tags corresponding to the target category node are dynamically acquired and applied to adjust the tags of the documents corresponding to the target category node. This effectively avoids the problem of a rigid document tag system, thereby improving the accuracy of document classification and retrieval efficiency corresponding to the target category node, and ultimately realizing the automated evolution and continuous optimization of the knowledge management system.

[0042] In some embodiments, when performing step S201 above, the document tag management device may also perform the following steps: determine multiple sub-tags corresponding to the target node from a preset tag tree, wherein the preset tag tree has tags corresponding to each node in the document tag tree, and the number of tags in the preset tag tree is greater than that in the document tag tree.

[0043] In the embodiments of this application, the preset tag tree refers to a tag architecture pre-built by the document tag management device that is more scalable and has a wider coverage. Compared with the document tag tree, the preset tag tree not only includes all the tags actually used in the current document, but also introduces potentially available but unused tags as possible expansion options. For example, under the parent node "artificial intelligence," the preset tag tree may include multiple sub-tags such as natural language processing, computer vision, machine learning, and deep learning, while the document tag tree may include only the tags "artificial intelligence" and "machine learning." Alternatively, under the node corresponding to the tag "deep learning," there may be sub-nodes, and the tags corresponding to the sub-nodes may include convolutional neural networks, etc. The design of the preset tag tree makes the tag management system more flexible and scalable.

[0044] In the embodiments of this application, after the document tag management device determines the target node, it can directly determine the child node corresponding to the target node from the preset tag tree. Each child node corresponds to a tag, resulting in multiple sub-tags.

[0045] In the embodiments of this application, a rich pool of candidate labels is provided through a preset label tree, which effectively avoids the problem of decreased classification accuracy due to insufficient available labels. Furthermore, suitable sub-labels can be dynamically selected from the preset label tree according to actual needs, thereby improving the adaptability of the label system and the accuracy of the classification results.

[0046] In some embodiments, when performing step S201 above, the document tag management device may also perform the following step: determine multiple sub-tags corresponding to the target node from the leaf nodes of the target node in the document tag tree.

[0047] In the embodiments of this application, the document tag tree is a hierarchical structure of document tags, reflecting the actual classification of documents. The document tag tree refers to a tag tree constructed by the document tag management device based on documents in a preset document database. For example, in response to storing the first document, based on a document classification algorithm, the tag to which the first document belongs is determined, the tag of the first document is set as its belonging tag, and it is stored in the preset document database. Furthermore, a child node is created at the root node of the document tag tree, and its corresponding tag is set to the tag to which the first document belongs. Then, the second document is stored. If the second document has the same tag as the first document, the corresponding tag is set and stored in the preset document database; otherwise, the second document is stored in the same manner as the first document, and the document tag tree is updated. As the number of stored documents increases, if a certain node (target node) in the document tag tree corresponds to too many documents in the preset document database, a document classification algorithm can be used to determine the child tags of the leaf nodes of the corresponding target node based on the documents under that node.

[0048] For example, if the target node is artificial intelligence and the corresponding document belongs to the tag of deep learning, machine learning, etc., then the multiple sub-tags corresponding to the determined target node are deep learning, machine learning, etc. Here, the sub-tags are determined based on the tag of the document corresponding to the target node. Therefore, the multiple sub-tags obtained here can be some or all of the sub-tags under the target node in the preset number of tags discussed above.

[0049] In some embodiments, when performing step S201 above, the document tag management device may also perform the following step: generating multiple sub-tags corresponding to the target node through the target model.

[0050] In the embodiments of this application, the target model refers to an intelligent model used to automatically generate sub-tags, which can be based on keyword extraction, vector representation (such as Word2Vec), deep learning models (such as BERT), or other text classification models. The target model automatically generates sub-tags that conform to semantic logic and contextual relationships by analyzing the content of the documents associated with the target node. For example, if the target node is artificial intelligence, the target model may generate sub-tags such as image recognition and speech synthesis based on the content of the documents associated with the target node.

[0051] In the embodiments of this application, sub-tags are automatically generated through a target model. The target model can effectively cope with rapid changes in the tag system and users' personalized needs, reducing manual intervention and achieving intelligent and automated tag management. Tags generated by the target model have stronger semantic coherence, contributing to improved overall classification quality and user experience.

[0052] In the embodiments of this application, the method of obtaining multiple sub-tags corresponding to the target node can be a combination of any of the two methods mentioned above, or a combination of the three methods mentioned above. For example, firstly, a basic tag candidate set is provided through a preset tag tree, then the leaf nodes in the document tag tree are combined to ensure that the tags are close to the actual document content, and finally the tags are intelligently generated through the target model. The three work together to realize the dynamic adjustment and continuous optimization of the tag system, thereby improving the classification efficiency and accuracy of the system.

[0053] In this way, by obtaining multiple sub-tags corresponding to the target node from at least the leaf nodes of the preset tag tree, the document tag tree, and one or more of the target model, the diversity of the tag system can be enriched, thereby improving the coverage and applicability of tags, and thus enabling dynamic optimization and precise management of document classification.

[0054] In some embodiments, after performing the above step S201, the document tag management device may also perform the following step: adding leaf nodes to the target node based on multiple sub-tags.

[0055] In the embodiments of this application, adding a leaf node refers to creating a new classification node under the target node. A leaf node represents a digital document that can be categorized into the corresponding leaf node. For example, if the label corresponding to the target node is "artificial intelligence," and the resulting multiple sub-labels are "natural language processing," "computer vision," and "machine learning," then for each sub-label, a leaf node is created under the target node, such as the leaf node corresponding to each sub-label in "natural language processing," "computer vision," and "machine learning."

[0056] In the embodiments of this application, a parent-child relationship exists between the leaf nodes and the target nodes, with the target node acting as the parent node and the leaf nodes as direct child nodes. By establishing a parent-child relationship between the leaf nodes and the target nodes, and making the target node the parent node and the leaf nodes the direct child nodes, the file tags can be further refined based on the original tag structure, thereby improving the accuracy of digital document organization and retrieval efficiency.

[0057] In the embodiments of this application, when the number of digital documents under a target node exceeds a first target number, the document tag management device automatically creates several leaf nodes for the target node and redistributes the original digital documents to the newly created leaf nodes according to their content. This operation enables dynamic expansion of the tag structure, allowing for flexible adjustment of the classification structure based on the actual distribution of digital documents.

[0058] In this way, by introducing leaf nodes, more detailed classification nodes can be provided while maintaining the consistency of the overall classification logic, improving the efficiency of users in finding specific types of digital documents. By adding leaf nodes to the target node based on multiple sub-tags, the dynamic expansion of the tag structure is realized, thereby improving the precision and accuracy of digital document classification, and further enhancing the level of intelligence and user experience.

[0059] In some embodiments, when the document tag management device performs step S202 above, such as Figure 3 As shown, the following steps S301 and S302 can be performed: Step S301: Based on multiple sub-tags, classify the documents corresponding to the target node to obtain the documents corresponding to each sub-tag.

[0060] In the embodiments of this application, classification refers to categorizing the documents corresponding to target nodes into different sub-tags according to semantic or topical attributes to achieve more refined document management. Classification operations typically rely on methods such as text feature extraction, keyword matching, and semantic models. For example, for a document containing a large number of AI-related target nodes, it can be subdivided into sub-tags such as Natural Language Processing, Computer Vision, and Machine Learning based on the content of the documents corresponding to the target nodes, and then categorized into their respective sub-tags.

[0061] In the embodiments of this application, sub-tags are relative to parent tags, and are used to represent categories that are more specific or subdivided than the parent tag. For example, the parent tag "artificial intelligence" can be broken down into sub-tags "natural language processing" and "computer vision." This hierarchical structure of parent and child tags enables the tagging system to possess multi-dimensional expressive capabilities, allowing it to better adapt to the document classification needs of target nodes in different scenarios.

[0062] In the embodiments of this application, the documents stored in the preset document database can be in various forms such as text files, images, audio, and video. Of course, they can also be just text documents such as research reports, papers, and technical documents, which are not limited here.

[0063] In the embodiments of this application, different classification algorithms can be used for different document types to classify the documents corresponding to the target node, so as to obtain a sub-tag for each document. This sub-tag is one of multiple sub-tags. For example, images can be classified using image recognition algorithms or image classification algorithms, and audio can be classified using audio classification algorithms, etc.

[0064] Step S302: Adjust the document tags corresponding to each sub-tag to the corresponding sub-tags.

[0065] In the embodiments of this application, adjusting the label refers to relabeling the document corresponding to the target node that originally belonged to the parent label as its own sub-label. This process typically involves data updates, index reconstruction, and modification of related metadata. For example, when a document corresponding to a target node that was originally labeled as "artificial intelligence" is classified into the "natural language processing" sub-label, the document's label management device will change the label of the document corresponding to the target node from "artificial intelligence" to "natural language processing".

[0066] In the embodiments of this application, the tag of the document corresponding to the target node refers to the metadata field used to identify the category or topic to which the document corresponding to the target node belongs. The tag is not only used for classification, but also for application scenarios such as search, filtering, and recommendation. The adjustment of the tag of the document corresponding to the target node is one of the key links to realize the dynamic balance mechanism. The adjustment of the tag of the document corresponding to the target node enables the document corresponding to the target node to maintain the latest classification status in the constantly changing tag system.

[0067] In this way, by adjusting the tags of the documents corresponding to the target nodes to the corresponding sub-tags, the structure of the tag system can be continuously optimized. This optimization makes the structure of the tag system more in line with the actual distribution of documents corresponding to the target nodes, thereby improving the intelligence and responsiveness of the overall system. By building a dynamic adjustment mechanism, the tag system can be made to adapt and evolve, thereby improving the level of intelligence in managing the documents corresponding to the target nodes. It can effectively cope with the large-scale and continuously growing document data environment corresponding to target nodes, optimize the document organization structure, and improve the efficiency of knowledge management and user experience.

[0068] In some embodiments, during the execution of step S102, the document tag management device may also perform the following steps: in response to the number of documents corresponding to the target node being less than the second target number, the tags of the documents corresponding to the target node are adjusted to the tags of the parent node corresponding to the target node.

[0069] In the embodiments of this application, the second target quantity (corresponding to the lower limit value discussed above) can be 20%, 50%, or other percentages less than 1, representing the average number of documents corresponding to different tags. For example, when the number of documents contained in a target node (i.e., a leaf node in a tree-like tag structure) is lower than the preset second target quantity, it indicates that the number of documents corresponding to that target node is too small, the classification is too fine, which easily leads to information fragmentation and affects retrieval efficiency. In this case, the document tagging device can automatically adjust the tags of the documents corresponding to that target node to the tags corresponding to the parent node of that target node, thereby achieving the merging operation of these documents, reducing classification levels, and improving the stability and usability of the entire tagging system.

[0070] In the embodiments of this application, a target node refers to a specific node in a document multi-category labeling system, typically a leaf node, used to represent a specific content category or topic. For example, under the parent node "Artificial Intelligence," there may be child nodes such as "Natural Language Processing" and "Computer Vision." The number of documents corresponding to the target node refers to the number of documents corresponding to the current target node in a preset document database. The second target number is used to determine whether the labels of the documents corresponding to the target node need to be adjusted.

[0071] In the embodiments of this application, after adjusting the label of the document corresponding to the target node to the label of the parent node corresponding to the target node, operations such as deletion or disabling can be performed on the target node in the document label tree.

[0072] In this way, by responding to changes in the number of documents corresponding to a target node and dynamically adjusting the tag structure associated with that target node, the problem of over-segmentation of the tag system caused by the target node can be effectively prevented. This maintains the rationality and stability of the tag distribution, ultimately improving document management efficiency and enhancing the user's search experience.

[0073] In some embodiments, during the execution of step S102, the document tag management device may also perform the following steps: in response to the total number of documents corresponding to all child nodes under the parent node corresponding to the target node being less than the third target number, the tags of all documents under the parent node corresponding to the target node are adjusted to the tags of the parent node corresponding to the target node.

[0074] In the embodiments of this application, when the total number of documents corresponding to target nodes contained in all child nodes under a parent node is less than a set third target number, the document tag management device will uniformly mark all documents corresponding to target nodes under these child nodes as the tags of the parent node's documents. This mechanism helps prevent fragmentation problems caused by overly detailed document tag structures, thereby maintaining the system's document classification efficiency and document retrieval performance.

[0075] In the embodiments of this application, the parent node corresponding to the target node may have multiple child nodes. For example, in a document tag tree, Artificial Intelligence can be the parent node, with child nodes such as Natural Language Processing and Computer Vision under Artificial Intelligence. The node corresponding to Natural Language Processing is the target node. The total number of documents refers to the sum of the number of documents corresponding to the target node owned by all child nodes (natural language processing, computer vision, etc.) under a certain parent node (Artificial Intelligence). The total number of documents is used to measure whether the distribution of documents corresponding to the target nodes under the parent nodes corresponding to the total number of documents is reasonable.

[0076] In the embodiments of this application, the third target number is a preset value of the document tag management device, used to determine whether document tag adjustment is needed (i.e., adjusting the tags of all documents under the target node corresponding to the parent node to the tags of the parent node corresponding to the target node). For example, when the total number of documents corresponding to the target node of all child nodes under a certain parent node is lower than the third target number, it is determined that the total number of documents corresponding to the target node of all child nodes under a certain parent node is lower than the third target number, indicating that there is a problem of too few documents corresponding to the parent node. At this time, the tags of the documents corresponding to the target nodes of all child nodes under a certain parent node are unified to the tags of the parent node's documents to avoid over-segmentation of the document tag system, which would affect document retrieval efficiency and the user experience of document management. The third target number can be 30% of the average number of documents corresponding to the parent nodes of child nodes in the document tag tree, or it can be a directly set value, such as 300 documents, or other values.

[0077] In the embodiments of this application, after the tags of all documents under the target node corresponding to the parent node are adjusted to the tags of the target node corresponding to the parent node, operations such as deletion or disabling can be performed on all child nodes under the target node corresponding to the parent node in the document tag tree.

[0078] Thus, when the number of documents corresponding to the parent node of the target node is insufficient, automatically merging document tags can reduce unnecessary subcategories, optimize the rationality of the document tag structure, and further improve the overall efficiency of document management corresponding to the target node and the user experience of document retrieval.

[0079] In some embodiments, during the execution of step S101 above, the document tag management device, such as Figure 4 As shown, the following steps S401 and S402 can also be performed: Step S401: Obtain the document quantity information corresponding to multiple nodes in the document tag tree according to the preset time period.

[0080] In the embodiments of this application, the preset time period refers to a set time interval for periodically executing tasks, such as once a day, once a week, or once an hour. The preset time period can be flexibly configured according to actual application needs to balance system load and response speed. For example, the document monitoring module scans the entire document tag tree according to the preset time period, counts the number of documents under each leaf node, and records the trend of change.

[0081] In the embodiments of this application, document quantity information refers to the data set obtained by statistically analyzing the number of documents contained in a preset document database for each tag corresponding to a node in the document tag tree at a specific point in time. Document quantity information reflects the usage frequency and activity level of each tag under the current classification system and is a key indicator for measuring the rationality of tags. By statistically analyzing the number of documents, it is possible to identify which tags are too dense (which may need to be subdivided) and which tags are too sparse (which may need to be merged).

[0082] In the embodiments of this application, if the document tag management device disables the target node or all child nodes corresponding to the parent node of the target node before step S401, then the document quantity information of the disabled node will not be obtained in this step, because it does not correspond to a document. The reason for not deleting it is that if the document quantity of the parent node of the target node is higher than the first target quantity again, the target node or all child nodes corresponding to the parent node of the target node will be directly started, without having to perform the step of adding leaf nodes to the parent node.

[0083] In the embodiments of this application, by setting a preset time period and having the data acquisition module obtain document quantity information, continuous monitoring of the tag system can be achieved, ensuring that the tag structure always adapts to changes in document distribution, thereby avoiding the problem of decreased classification efficiency caused by using a static tag system.

[0084] Step S402: Determine the target node from multiple nodes based on the document quantity information.

[0085] In the embodiments of this application, the target node is a node selected in the document tag tree for further processing. The selection of the target node is based on the document quantity information. For example, when the number of documents contained in the target node exceeds the first target quantity, it indicates that the number of documents contained in the node is too large and the classification is too coarse, and the node should be selected as the target node for refinement. Conversely, if the number of documents contained in the target node is less than the second target quantity or the total number of documents corresponding to all child nodes under the target node's parent node is less than the third target quantity, it indicates that the number of documents contained in the node is too small and the classification is too fragmented, and the node should also be selected as the target node for merging.

[0086] In the embodiments of this application, the document quantity information provides a quantitative basis for evaluating the rationality of the tags, and the selection of target nodes is a decision made based on this information. By using the mechanism of determining target nodes based on document quantity information, it is possible to automatically identify tag nodes that need to be optimized and take corresponding structural adjustment measures to maintain the dynamic balance of the tag system.

[0087] In this way, by setting a preset time period to obtain document quantity information and determining the target node based on the obtained document quantity information, dynamic monitoring and adjustment of the tag system can be achieved, thereby maintaining the rationality and adaptability of the tag system and further improving the overall efficiency of document management.

[0088] In the embodiments of this application, the logical relationship between the preset time period and the document quantity information is that the preset time period determines the frequency and accuracy of document quantity information collection. A longer preset time period may lead to system response lag; a shorter preset time period may increase system resource overhead. Therefore, in actual deployment, the document quantity corresponding to each node can be recorded in real time during file storage or deletion. Once the preset conditions are met, the document tag of the target node's corresponding document can be directly adjusted, thus achieving optimal system performance.

[0089] In some embodiments, the document tag management device may also perform the following steps: determining a first target number based on the total number of documents corresponding to the document tag tree.

[0090] In the embodiments of this application, a document tag tree refers to a tree-like structure used to organize and classify digital documents. A parent node in a document tag tree can contain multiple child nodes, thus forming a multi-level tag system. For example, artificial intelligence can serve as the parent node in a document tag tree, with child nodes for natural language processing, computer vision, and other document tag trees. This document tag tree structure not only facilitates users' understanding of the hierarchical relationships within digital document content but also allows document management systems to dynamically adjust and manage the document tag tree.

[0091] In the embodiments of this application, the total number of digital documents corresponding to the document tag tree refers to the total number of all digital documents in the current tagging system, including the digital documents under each node of the document tag tree. The total number of digital documents corresponding to the document tag tree reflects the overall data scale of the document management system and serves as the basis for calculating the first target quantity. By statistically analyzing the total number of digital documents under the entire document tag tree, the document management system can more scientifically set a reasonable threshold range, thereby achieving adaptive adjustment of digital document classification.

[0092] In the embodiments of this application, the first target number is a dynamically generated value based on the total number of digital documents. It is used to determine whether there are too many digital documents under a certain tag, and thus decide whether it is necessary to split the tags under the leaf nodes in a document tag tree. For example, if the total number of digital documents is 10,000, the first target number may be set to 100. When the number of digital documents under a leaf node in a document tag tree exceeds the first target number, a tag refinement operation will be triggered, creating new sub-tags and reallocating the digital documents.

[0093] For example, the first target number can also be based on the average number of documents corresponding to each node. For instance, if the average number of documents corresponding to each node is 100, then the first target number can be set to 120% of that, i.e., 120. Alternatively, it can be based on the mode of the number of documents corresponding to each node, or other values. By dynamically setting the first target number according to the total number of digital documents, rather than using a fixed value, it can better adapt to datasets of different sizes, making the tagging system more flexible and scalable. This avoids the problem of inappropriate tag distribution caused by changes in data volume, thereby maintaining the rationality and practicality of the tag structure and further improving the overall efficiency and user experience of digital document management.

[0094] For example, the core of the embodiments of this application lies in constructing a multi-level document tag system (corresponding to a document tag tree), wherein the multi-level document tag system is a tree structure, with parent nodes being coarse-grained tags and child nodes being fine-grained tags. By monitoring the document quantity distribution of leaf nodes and adaptively adjusting the document tag granularity based on the monitoring results, tags with a large number of documents are split into finer-grained sub-tags, and documents under these tags are further classified according to the sub-tags; tags with a small number of documents are merged into the coarse-grained tags of the parent node. This approach differs from traditional static and fixed tag classification methods by introducing a dynamic feedback adjustment mechanism, which personalizes the classification structure according to the actual situation of document accumulation (document quantity distribution), enabling the tag system to grow and evolve. Secondly, a dynamic balancing mechanism is introduced, which automatically triggers the tag granularity adjustment process by monitoring changes in document type distribution in real time, always maintaining a tag distribution state with appropriate granularity. Finally, adaptive granularity control is implemented. When the number of documents in a certain category exceeds a preset threshold, tag subdivision is automatically performed; and when the number of documents is too small, tags are automatically merged to prevent fragmentation.

[0095] In practical applications, it has the following advantages: adaptive adjustment ensures that the tag system always fits the actual data situation, allowing users to locate target documents more quickly and significantly improving information retrieval speed and accuracy; the document tag management method provided in this application embodiment reduces the large amount of work that relies on manual periodic review and reclassification of tags, realizes the automated and intelligent operation and maintenance of the tag system, and significantly reduces the manpower and time costs of long-term maintenance; the personalized document tag system avoids the retrieval difficulties caused by rigid classification, provides users with a continuously optimized information organization experience, and is particularly suitable for application scenarios with a large number of documents and rapid growth.

[0096] like Figure 5 As shown, the implementation of an exemplary document tag management method is as follows: Step S501, Initial State and Monitoring.

[0097] Here, the process starts from a predefined multi-level tag tree (corresponding to the document tag tree mentioned above). The document monitoring module will periodically traverse all leaf nodes (i.e., the lowest level tag in the document tag tree, or all nodes) and count the number of documents under the leaf nodes.

[0098] Step S502: Threshold Judgment and Decision.

[0099] Here, the number of documents in each leaf node is compared with preset upper threshold (corresponding to the first target number mentioned above) and lower threshold (corresponding to the second document number mentioned above, or the second document number mentioned above), and different operations are triggered based on the comparison results: If the number of documents is greater than the upper threshold: it means there are too many documents under this tag, indicating that the current classification is too coarse and cannot effectively distinguish different content. Therefore, this tag needs to be further split to improve classification accuracy and management efficiency. If the number of documents is less than the lower threshold: it means there are too few documents under this tag, and the classification is too fragmented. The documents in the current tag need to be adjusted to the tags corresponding to the parent node.

[0100] Step S503: Label splitting.

[0101] Here, the document tagging device creates a new set of more granular sub-tags for the leaf nodes. For example, a tag named "Artificial Intelligence" is split into sub-tags such as "Natural Language Processing," "Computer Vision," and "Machine Learning." Subsequently, using text classification algorithms (such as those based on keywords, vector models, or machine learning models), all documents originally tagged with "Artificial Intelligence" are automatically categorized into the newly created sub-tags according to their content; that is, the tags of all documents originally tagged with "Artificial Intelligence" are modified to the newly created sub-tags.

[0102] Step S504: Tag merging.

[0103] Here, the document tagging device uniformly relabels all documents under a leaf node with the tag corresponding to the parent node. For example, if the number of documents corresponding to the leaf node of the convolutional neural network is too small, the tags of all documents corresponding to that leaf node are updated to the tag corresponding to the parent node: deep learning. Subsequently, the leaf node can be disabled or retained, but the documents under that leaf node will no longer belong to that leaf node.

[0104] Step S505: Update the tag tree.

[0105] Here, whether you execute step S503 or step S504, you need to update the entire tag tree structure after the operation is completed to ensure the integrity and consistency of the tag tree structure and prepare for the next round of monitoring and adjustment.

[0106] For example, to achieve the above functions, the document tag management device may include the following main functional modules: a tag tree management module, a document monitoring module, a threshold judgment logic module, a tag refinement module, and a tag merging module; wherein, the tag tree management module is responsible for maintaining the entire tree-like tag structure and recording the parent-child relationships; the document monitoring module, as the core monitoring unit of the system, is responsible for periodically scanning and counting the number of documents under each leaf tag; the threshold judgment logic module is used for a preset decision mechanism to determine whether the number of documents is too many or too few based on the threshold; the tag refinement module performs tag splitting and document refinement reclassification; the tag merging module performs tag merging operations and updates the tag affiliation of documents.

[0107] In summary, by constructing a dynamically feedback-adjustable document tag management system, the self-evolution and intelligent adjustment of the tag system are realized, thereby effectively solving the problem of insufficient adaptability of the traditional fixed tag system when facing changes in the actual document distribution, and further improving the accuracy and efficiency of document classification and retrieval.

[0108] This application provides a document tag management method, which includes: determining a target node from a document tag tree; each node in the document tag tree corresponding to a tag; a preset document database storing documents corresponding to different tags; the number of documents corresponding to the target node in the preset document database meeting a preset condition; the document database being used to provide reference information for inference in a target language model; and adjusting the document tags of the documents corresponding to the target node. The document tag management method provided in this application combines a tree-like tag structure with a preset document database, and adjusts the tags of documents in the preset document database in real time according to the number of documents corresponding to the target node, achieving adaptive optimization of the tag system and improving document classification efficiency and retrieval accuracy.

[0109] This application provides a document tag management device 6, such as... Figure 6 As shown, it includes: The determination module 61 (corresponding to the aforementioned tag tree management module, document monitoring module, and threshold judgment logic module) is used to determine the target node from the document tag tree; each node in the document tag tree corresponds to a tag; the preset document database stores documents corresponding to different tags; the number of documents corresponding to the target node in the preset document database meets the preset conditions, and the document database is used to provide reference information for the inference of the target language model; Adjustment module 62 (corresponding to the tag refinement module and tag merging module mentioned above) is used to adjust the document tags of the document corresponding to the target node.

[0110] In one embodiment of this application, the determining module 61 is further configured to, in response to the number of documents corresponding to the target node being greater than the first target number, obtain multiple sub-tags corresponding to the target node; and adjust the tags of the documents corresponding to the target node according to the multiple sub-tags.

[0111] In one embodiment of this application, the determining module 61 is further configured to determine multiple sub-tags corresponding to the target node from a preset tag tree, wherein the preset tag tree has tags corresponding to each node in the document tag tree, and the number of tags in the preset tag tree is greater than that in the document tag tree; determine multiple sub-tags corresponding to the target node from the leaf nodes of the target node in the document tag tree; and generate multiple sub-tags corresponding to the target node through the target model.

[0112] In one embodiment of this application, the determining module 61 is further configured to add leaf nodes to the target node based on multiple sub-labels.

[0113] In one embodiment of this application, the adjustment module 62 is further configured to classify the documents corresponding to the target node based on multiple sub-tags to obtain the documents corresponding to each sub-tag; and adjust the tags of the documents corresponding to each sub-tag to the corresponding sub-tags.

[0114] In one embodiment of this application, the adjustment module 62 is further configured to adjust the tags of the documents corresponding to the target node to the tags of the parent node corresponding to the target node in response to the number of documents corresponding to the target node being less than the second target number.

[0115] In one embodiment of this application, the adjustment module 62 is further configured to adjust the tags of all documents under the parent node corresponding to the target node to the tags of the parent node corresponding to the target node in response to the total number of documents corresponding to all child nodes under the parent node corresponding to the target node being less than the third target number.

[0116] In one embodiment of this application, the determining module 61 is further configured to obtain document quantity information corresponding to multiple nodes in the document tag tree according to a preset time period; and determine the target node from the multiple nodes based on the document quantity information.

[0117] In one embodiment of this application, the determining module 61 is further configured to determine the first target quantity based on the total number of documents corresponding to the document tag tree.

[0118] This application provides a document tag management device that determines a target node from a document tag tree; each node in the document tag tree corresponds to a tag; a preset document database stores documents corresponding to different tags; the number of documents corresponding to the target node in the preset document database meets a preset condition; the document database is used to provide reference information for inference of the target language model; and the document tags of the documents corresponding to the target node are adjusted. The document tag management device provided in this application combines a tree-like tag structure with a preset document database, and adjusts the tags of documents in the preset document database in real time according to the number of documents corresponding to the target node, achieving adaptive optimization of the tag system and improving document classification efficiency and retrieval accuracy.

[0119] Figure 7 A schematic diagram of the structure of a document tag management device 7 provided in an embodiment of this application includes: a processor 71, a memory 72, and a communication bus 73; Communication bus 73 is used to realize the communication connection between processor 71 and memory 72; The processor 72 is used to execute the computer program stored in the memory 72 to implement the document tag management method described above.

[0120] This application provides a computer-readable storage medium storing one or more computer programs, which can be executed by one or more processors to implement the document tag management method described above. The computer-readable storage medium can be transient or non-transient.

[0121] This application provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program. When the computer program is read and executed by a computer, it implements some or all of the steps in the interface generation method described above. This computer program product can be implemented specifically through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied as a computer storage medium; in another optional embodiment, the computer program product is specifically embodied as a software product, such as a software development kit (SDK), etc. In some embodiments, the storage medium may be a computer-readable storage medium, which may be volatile memory, such as random-access memory (RAM); or non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), solid-state drive (SSD), ferromagnetic random access memory (FRAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory, magnetic surface memory, optical disk, or compact disk-read-only memory (CD-ROM); or it may be various devices including one or any combination of the above-mentioned memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc.

[0122] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0123] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0124] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0125] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0126] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0127] As an example, executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file containing other programs or data, for example, in one or more scripts within a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple co-located files (e.g., files storing one or more modules, subroutines, or code sections). As an example, executable instructions may be deployed to execute on a single computing device, or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0128] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

[0129] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the above-described embodiments are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0130] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components may be combined, or integrated into another system, or some features may be ignored or not performed.

[0131] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for managing document tags, the method comprising: Identify the target node from the document tag tree; Each node in the document tag tree corresponds to one tag; The default document database stores documents corresponding to different tags; The number of documents corresponding to the target node in the preset document database meets a preset condition, and the document database is used to provide reference information for the inference of the target language model; Adjust the document tags of the document corresponding to the target node.

2. The document tag management method according to claim 1, wherein adjusting the document tags of the document corresponding to the target node includes: In response to the fact that the number of documents corresponding to the target node is greater than the first target number, multiple sub-tags corresponding to the target node are obtained; Adjust the tags of the document corresponding to the target node based on the multiple sub-tags.

3. The document tag management method according to claim 2, wherein obtaining the multiple sub-tags corresponding to the target node includes at least one of the following: Multiple sub-tags corresponding to the target node are determined from a preset tag tree, wherein the preset tag tree contains tags corresponding to each node in the document tag tree, and the number of tags in the preset tag tree is greater than that in the document tag tree; Determine multiple sub-tags corresponding to the target node from the leaf nodes of the target node in the document tag tree; Multiple sub-labels corresponding to the target node are generated using the target model.

4. The document tag management method according to claim 2, after obtaining the multiple sub-tags corresponding to the target node, the method further includes: Add leaf nodes to the target node based on the multiple sub-labels.

5. The document tag management method according to claim 2, wherein adjusting the tags of the document corresponding to the target node according to the plurality of sub-tags includes: Based on the multiple sub-tags, the documents corresponding to the target node are classified to obtain the documents corresponding to each sub-tag; Adjust the document tags corresponding to each sub-tag to the corresponding sub-tag.

6. The document tag management method according to claim 1, wherein adjusting the document tags of the document corresponding to the target node includes: In response to the fact that the number of documents corresponding to the target node is less than the second target number, the tags of the documents corresponding to the target node are adjusted to the tags of the parent node corresponding to the target node.

7. The document tag management method according to claim 1, wherein adjusting the tags of the document corresponding to the target node includes: In response to the fact that the total number of documents corresponding to all child nodes under the parent node of the target node is less than the third target number, the tags of all documents under the parent node of the target node are adjusted to the tags of the parent node of the target node.

8. The document tag management method according to claim 1, wherein determining the target node from the document tag tree includes: Obtain the document quantity information corresponding to multiple nodes in the document tag tree according to a preset time period; The target node is determined from multiple nodes based on the document quantity information.

9. The document tag management method according to claim 2, further comprising: The first target quantity is determined based on the total number of documents corresponding to the document tag tree.

10. A document tag management device, comprising: The determination module is used to identify the target node from the document tag tree; Each node in the document tag tree corresponds to one tag; The default document database stores documents corresponding to different tags; The number of documents corresponding to the target node in the preset document database meets a preset condition, and the document database is used to provide reference information for the inference of the target language model; The adjustment module is used to adjust the document tags of the document corresponding to the target node.

Images