Audio book query matching method and device based on AI model

Abstract

The invention belongs to the technical field of big data, and relates to an audio book query matching method and device based on an AI model, and the device can efficiently and accurately cluster and store multi-modal books by adopting a multi-level index and node clustering algorithm. The node construction of the books not only ensures that the storage is clear and ordered, but also reduces the time overhead during book query and improves the book storage efficiency through the implementation based on the importance weight and correlation degree of the books. Through incremental snapshot, regular snapshot and restoration point creation functions, an efficient book snapshot and restoration mechanism is provided for a client, through a past label restoration function, the client flexibly restores a book to any past state, and a label storage unit also provides label comparison and repeated solution functions, so that the book can be quickly restored to any past state. The method is suitable for label combination during parallel updating of multiple clients, and the integrity and consistency of book information are ensured.

Description

Technical Field

[0001] The present invention belongs to the field of big data technology and relates to a method and device for audio book query matching based on an AI model. Specifically, it relates to the efficient storage, storage, query, and restoration of multimodal books in large-scale multimodal book storage through multiple technical means such as multi-level integration, query statement processing, permission configuration, and tag storage. The present invention can be widely applied to multiple fields such as multimodal storage, digital media libraries, cloud storage services, and online multimodal libraries to improve the storage efficiency and security of multimodal books and improve book storage and reading methods. Background Art

[0002] With the advent of the information age, digital multimodal books have been widely used in various businesses, especially in the fields of the internet, cloud storage, and digital media. The number of multimodal books continues to increase, and the complexity of book storage has also increased accordingly. Traditional multimodal book storage devices mainly rely on simple book storage methods and lack functions such as multi-level integration of multimodal books, query statement processing, permission configuration, and tag storage. This leads to problems such as book disorganization, storage space waste, low query efficiency, and poor information security when storing large-scale information.

[0003] Currently, some book storage devices on the market can provide functions such as book clustering, querying, and snapshots, but they still have the following defects:

[0004] Insufficient multi-level storage capacity: Most devices only provide flat storage and cannot efficiently store books based on their priority relationships, resulting in confusion in book clustering and integration, affecting subsequent query and storage efficiency.

[0005] Limited query statement processing capabilities: Traditional devices fail to perform effective modality conversion, segmentation, clarity correction, and other processing for multimodality, and are unable to adapt to a variety of book modalities and storage instructions, resulting in low multimodal quality and storage efficiency. Summary of the Invention

[0006] This invention provides an AI-based audio book query and matching method and device. Through the collaboration of multiple units, this method enables efficient multi-level book storage, modality conversion, permission configuration, tag storage, query and matching, information snapshot and restoration, and inventory collection and analysis in large-scale multimodal book storage. These technical approaches address existing issues such as low multimodal book storage efficiency, poor information security, and difficulty locating tags, thereby enhancing the autonomy and security of book storage.

[0007] The present invention specifically includes the following contents:

[0008] 1. Multi-level book storage unit: used for multi-level integration and cluster storage of books based on the priority relationship and storage instructions of multimodal books, mainly including:

[0009] (1) Book node construction function based on multi-level index: Multi-node storage is obtained for multimodal books according to the tree. In the tree, each layer represents a book category or book group. The deeper the priority, the more specific the book clustering. In order to store the nodes of the books, the device adopts a node clustering algorithm:

[0010] The multimodal book set is F = {f1, f2, ..., f n}, the node tree is T = (V, E), then the node objective function is:

[0011]

[0012] Among them, ω i is the importance weight of the book, d is the correlation between the book and the node center, α is the correlation coefficient, and D(T) is the node depth of the node tree;

[0013] (2) Book clustering and annotation function: allows multimodal books to be clustered and annotated for quick matching and classification. Each book is assigned one or more keywords based on its attributes (book type, shooting location, shooting time). Keyword processing not only improves the storage efficiency of books, but also allows the target book to be quickly found based on keywords during query.

[0014] (3) Dynamic correction function of the relationship between book nodes: used to dynamically correct the position of books in the node tree. As the number of books and clustering instructions change, the node relationship of books may need to be reconfigured. In order to achieve this flexibility, a summary association table is adopted to collect the nodes where the books are located and their changes. When the nodes of a book change, the summary association table quickly locates and corrects the position of the book.

[0015] 2. Query statement processing unit: used for mode conversion, segmentation, clarity correction and color processing of multimodal books, mainly including:

[0016] (1) Multimodal parallel mode conversion function: Through a unified batch processing interface, a large number of multimodal images can be quickly and efficiently converted from one mode to another (JPEG, PNG, BMP conversion), ensuring the compatibility of query statements in different devices and application scenarios.

[0017] (2) Query statement quality improvement function: Use the neural network model to autonomously analyze the query statement, and improve the visual quality of the query statement through denoising, enhancing details, and super clarity.

[0018] (3) Query statement segmentation function: It provides two algorithms: lossless segmentation (no information is lost during the segmentation process, and the query statement quality remains unchanged, which is suitable for scenarios with extremely high quality requirements) and lossy segmentation (some information is lost during the segmentation process, which usually achieves a higher segmentation ratio and is suitable for scenarios with high requirements on book length (web page loading, limited storage space)). The segmentation method is adaptively selected based on the actual instructions. The query statement segmentation formula is:

[0019] Function word segment R = {r1, r2, ..., r m}, define the entropy value of each segment as H(r i ), then the maximum split ratio of the query statement is:

[0020]

[0021] Among them, P is the total function words of the query sentence, P(r i ) is the number of functional words in the segment.

[0022] 3. Permission configuration unit: As one of the core components of the device, it is mainly used to set reading, updating, and forwarding permissions for multimodal books based on client accounts and security instructions. It mainly includes:

[0023] (1) Client license level classification function: Adapts to three client account configurations:

[0024] Server: Has the highest permissions, performs device configuration, license scheduling, client storage and other analysis, and reads almost all functions in the device;

[0025] Updater: has certain permissions and can update content, update books, update information, etc., but cannot update device settings or schedule permissions;

[0026] Reader: The lowest level of permission, only able to read books or information, unable to update or configure analysis;

[0027] (2) Dynamic license modification function: based on the client's events and time range, the license is updated regularly. By analyzing the client's event pattern and reading history, the device can autonomously recommend license modifications to ensure the dynamic and real-time nature of license configuration. Specifically, a dynamic license recommendation algorithm is used. The client event sequence is S = {s1, s2, ..., s k}, the permission priority is L = {l1, l2, ..., l m}, then the probability that the client belongs to a certain permission priority is:

[0028]

[0029] (3) Information security audit function: This function ensures the collection and location of the license scheduling, change and adoption process. This function collects detailed information for each analysis and analyzes these information to discover potential security threats and violations. It mainly includes:

[0030] Permit scheduling and change collection: Whenever a device's permission configuration changes, a collection will be generated, detailing the time of the permission change, the person who performed the change, the content of the change, etc.

[0031] Analysis event positioning: collect the analysis of books, information or devices by each client, including reading, updating, and deleting. By collecting the analyzed client, time, and analysis content, we can locate the specific analysis;

[0032] Security risk analysis: By analyzing the client-side analysis and collection, the device identifies abnormal events such as excessive license enhancement, illegal reading, frequent license changes, etc., promptly issues an alarm, and notifies the server to handle it.

[0033] 4. Tag storage unit: used to collect and locate the update history of multimodal books, adapt to tag backtracking and multi-tag storage, mainly including:

[0034] (1) Update collection and timing saving function: The update collection and timing saving function will obtain an independent tag number every time a query statement book is updated, and collect the specific content of the update, including:

[0035] Book length: The updated book length reflects the amount of changes in the book content;

[0036] Update time: collect the timestamps of book updates for locating the past;

[0037] Updater: Collect the clients that perform update analysis for later review and location.

[0038] These updated collections will be stored in a dedicated tag information library to form a complete update history chain. Each time an update is made, the device will regularly save a copy of the update and obtain a unique tag number. The tag number is obtained using an incremental numbering or timestamp-based strategy. In this way, the client can easily locate the past status of the book, read the evolution of the book content, and ensure that any changes can be traced.

[0039] (2) Tag comparison and duplicate resolution function: When multiple people collaborate to update the same book, it may happen that multiple clients update the same book at the same time. In order to avoid duplicate book content, the tag storage unit provides a tag comparison and duplicate resolution function. The tag comparison function compares the differences between different tags and identifies which parts have changed at regular intervals. The device analyzes the differences in text, information or query statements between each tag and obtains a difference report. The report includes the updated part (marking the changed content between the two tags for easy identification by the client) and the updated time and updater (showing who updated the book and when). For duplicate tags, the device provides three options: latest tag priority (retain the update of the latest tag and ignore the update of the previous tag), merge updates of different tags (merge updates of different tags together, usually requiring trigger intervention, and the client decides how to merge based on the specific content), and manual confirmation (if the device cannot resolve the duplication at regular intervals, the client manually selects which tag's updated content should be retained). These three options allow the client to choose how to merge these tags.

[0040] (3) Timestamp backtracking function: allows the client to read all the past tags of the book through the timestamp. Each tag will be displayed on the timestamp. The client can select the tag at any time to restore. The timestamp display content includes the tag number (each tag has a unique number to facilitate identification of different tags), update time (collects the creation time of each tag to help the client quickly locate the specific tag), and a brief description of the updated content (the updated content of each tag will be briefly described so that the client can understand the changes in the book).

[0041] 5. Query and matching unit, used for rapid query and matching based on the name, keywords, attributes and content of multimodal books, mainly including:

[0042] (1) Multi-condition query function: Adapts the client to query based on multiple conditions and provides flexible query methods. The query conditions include:

[0043] Book name: Advanced or primary matching is performed based on the book name. The client searches for related books by entering some characters in the book name.

[0044] Keywords: The client searches based on the keywords attached to the book (such as "characters" and "scenery");

[0045] Creation time and update time: The client specifies a time range to search for books. For example, search for books created or updated on a specific date or within a date range.

[0046] Book Length: Adapts to filtering based on the length of books. The client sets the upper or lower limit of the book length to find books that meet the length conditions.

[0047] The combined query method is also suitable for advanced matching, basic matching and regular expression queries, and can handle a variety of complex query instructions.

[0048] (2) Keyword matching function: Books can be clustered based on keyword information, thereby improving the flexibility and convenience of queries. Books are given multiple keywords (theme, color, shooting location, etc.) and can be easily matched through these keywords.

[0049] (3) Query statement content query function: Using query statement recognition technology, relevance query is performed based on multimodal book themes (shape, color distribution, texture, etc.). The query statement content query function can not only query multimodal book names or keywords, but also query based on the characteristics of the query statement itself.

[0050] (4) Autonomous sorting and priority matching: Based on the transaction processing algorithm, the client's past query events are analyzed, and the priority of book display is predicted and automatically corrected. The device will sort the query results regularly based on the client's past queries and selections, thereby providing more personalized and relevant query results.

[0051] 6. Information snapshot and restoration unit, used to periodically snapshot the multi-level information of multimodal books and provide information restoration function after device failure, mainly including:

[0052] (1) Incremental snapshot function: When a book changes, only the changed part is snapshotted, rather than the entire book. By collecting the summary value of the book (each book gets a summary value, and periodically checks whether the summary value of the book has changed. Only books with changed summary values ​​are considered to have changed, and then snapshots are taken) or adopting timestamp identification (each time a book changes, the device collects the timestamp of the book. When taking a snapshot, the device compares the timestamps. If the timestamp of the book is updated, it is considered that the book has changed, and a snapshot is taken) the changed content, efficient information snapshots are achieved, saving storage space and bandwidth.

[0053] (2) Periodic snapshot and manual snapshot functions: The client sets the periodic execution of snapshots, such as daily, weekly, and other periodic snapshots. This method ensures that the client's books are snapshotted within a certain period of time, without the need for manual analysis, making it easier to keep the information up to date. The device also allows the client to manually initiate snapshot analysis. When needed, the client presses the snapshot button, selects the content and storage location of the snapshot, and takes an instant snapshot. Regardless of whether it is a periodic snapshot or a manual snapshot, the snapshot is stored in the local storage device or in the cloud.

[0054] 7. Inventory collection and analysis unit, used to collect client analysis events and device operating status, and obtain analysis reports for device performance analysis, mainly including:

[0055] (1) Client analysis list collection function: The device collects all analysis events of each client, including book upload, download, update, deletion, reading, etc. Each analysis collects the analysis time, client ID, analysis type and analysis object book information;

[0056] (2) Information analysis and recommendations: Through big information analysis technology, the client event list and device operation information are analyzed to mine the potential instructions of the device, such as loading speed bottlenecks, books with the highest reading frequency, etc. Based on the analysis results, the device will receive recommendations at regular intervals, such as cache strategies, book reading paths, etc., to improve overall performance.

[0057] An audio book search and matching method and device based on an AI model specifically includes the following steps:

[0058] Step 1: Based on the priority relationship and storage instructions of multimodal books, the device dynamically obtains multimodal book storage nodes through multi-level indexing, and uses node clustering algorithm to integrate books. Books are quickly queried and matched through keyword processing. When books are synchronized or reintegrated, a summary association table is used to collect node changes and adapt to cross-node parallel synchronization and information synchronization.

[0059] Step 2: Multimodal books are modally converted, segmented, clarity-corrected, and color-processed. Multimodal content is converted into different modalities in parallel. Query quality is independently improved through a neural network model to enhance detail expression and clarity. Query segmentation uses lossless and lossy segmentation algorithms. A lossless segmentation algorithm based on dynamic segmentation is used to correct the multimodal length to balance the volume and quality of the book.

[0060] Step 3. Based on the client account and security instructions, the device sets the reading, updating and forwarding permissions for multimodal books. The device configures permissions for different accounts, dynamically modifies permissions based on client event analysis and time range, and adopts a dynamic permission recommendation algorithm to predict and modify client permissions. The device collects and locates client analysis events and permission adoption status to ensure the transparency and locatability of the analysis.

[0061] Step 4: Collect and locate the update history of multimodal books, adapt to tag backtracking and multi-tag storage. Each time a book is updated, the device regularly saves the update collection, including the book length, update time and updater information, and obtains a new tag number. The device compares the content of books with different tags, detects duplications and prompts the client to select a duplication resolution strategy. The client reads all the past tags of the book through the timestamp and restores the book tag to a certain moment.

[0062] Step 5. Multimodal books are quickly searched and matched based on names, keywords, attributes and content. The client searches based on book names, keywords, creation time, update time, book length and other conditions, and is adapted to advanced and elementary matching and regular expression queries. Multimodal books are clustered by keywords, and the client can match books based on multiple keyword combination conditions. The device also uses query statement recognition technology to conduct content relevance queries based on book topics, analyzes the client's past query events through transaction processing algorithms, and autonomously corrects the sorting of query results to improve query accuracy and efficiency.

[0063] Step 6. The device provides periodic snapshot and restore functions for books to ensure the integrity and security of information. When a book changes, the device only snapshots the changed part, rather than the entire book. The summary value or timestamp is used to identify the changed part. The device is adapted to take regular snapshots and allows the client to manually initiate snapshot analysis.

[0064] Step 7. The device collects client analysis and device operation status in real time, and obtains information analysis reports to analyze device performance. The device collects all analysis events of each client and saves information such as analysis time, client ID, analysis type, etc. The device uses big information analysis technology to analyze client events and device operation information, explore potential instructions performed by the device, and obtain recommendations based on the analysis results in a timely manner to improve overall performance.

[0065] An audio book search and matching method and device based on an AI model has the following beneficial effects:

[0066] Improve the efficiency and accuracy of device book storage

[0067] By employing a multi-level indexing and node clustering algorithm, the device is able to efficiently and accurately cluster and store multimodal books. The node-based structure of books not only ensures clear and orderly storage but also reduces the time overhead of book searches by weighting and relevance based on book importance, improving storage efficiency. Dynamically modifying book node relationships and adapting to cross-node parallel synchronization further enhance the device's flexibility and adaptability.

[0068] Improve the level of autonomy in query statement processing

[0069] The device improves query quality by introducing a neural network model, enabling scheduled multimodal quality checks and reducing the need for triggered intervention. Furthermore, a lossless segmentation algorithm based on dynamic segmentation adaptively selects matching segmentation strategies to maximize the segmentation ratio while ensuring query quality, saving storage space and bandwidth for the client. This makes it particularly suitable for storing large volumes of query books.

[0070] Enhance information security and locatability

[0071] Through the inventory collection and analysis unit, the device comprehensively collects client analysis events and adapts to locate and analyze book uploads, downloads, updates, deletions, and readings. This function provides a detailed audit trail for the device, effectively monitoring and preventing inappropriate analysis. Furthermore, the information analysis function uses big data analysis technology to evaluate client events and device operating status, promptly identifying potential security risks or performance bottlenecks.

[0072] Improve the flexibility of book restoration and tag storage:

[0073] The device provides clients with an efficient book snapshot and restore mechanism through incremental snapshots, periodic snapshots, and restore point creation. Incremental snapshots only capture the changed portions of a book, saving storage space. The past tag restore feature allows clients to flexibly restore a book to any previous state. The tag storage unit also offers tag comparison and duplicate resolution, accommodating tag merging when multiple clients are updating concurrently, ensuring the integrity and consistency of book information. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 It is a flow chart of the method steps. DETAILED DESCRIPTION

[0075] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention:

[0076] 1. Multi-level book storage unit: used for multi-level integration and cluster storage of books based on the priority relationship and storage instructions of multimodal books, mainly including:

[0077] (1) Book node construction function based on multi-level index: Multi-node storage is obtained for multimodal books according to the tree. In the tree, each layer represents a book category or book group. The deeper the priority, the more specific the book clustering. In order to store the nodes of the books, the device adopts a node clustering algorithm:

[0078] The multimodal book set is F = {f1, f2, ..., f n}, the node tree is T = (V, E), then the node objective function is:

[0079]

[0080] Among them, ω i is the importance weight of the book, d is the correlation between the book and the node center, α is the correlation coefficient, and D(T) is the node depth of the node tree;

[0081] (2) Book clustering and annotation function: allows multimodal books to be clustered and annotated for quick matching and classification. Each book is assigned one or more keywords based on its attributes (book type, shooting location, shooting time). Keyword processing not only improves the storage efficiency of books, but also allows the target book to be quickly found based on keywords during query.

[0082] (3) Dynamic correction function of the relationship between book nodes: used to dynamically correct the position of books in the node tree. As the number of books and clustering instructions change, the node relationship of books may need to be reconfigured. In order to achieve this flexibility, a summary association table is adopted to collect the nodes where the books are located and their changes. When the nodes of a book change, the summary association table quickly locates and corrects the position of the book.

[0083] 2. Query statement processing unit: used for mode conversion, segmentation, clarity correction and color processing of multimodal books, mainly including:

[0084] (1) Multimodal parallel mode conversion function: Through a unified batch processing interface, a large number of multimodal images can be quickly and efficiently converted from one mode to another (JPEG, PNG, BMP conversion), ensuring the compatibility of query statements in different devices and application scenarios.

[0085] (2) Query statement quality improvement function: Use the neural network model to autonomously analyze the query statement, and improve the visual quality of the query statement through denoising, enhancing details, and super clarity.

[0086] (3) Query statement segmentation function: It provides two algorithms: lossless segmentation (no information is lost during the segmentation process, and the query statement quality remains unchanged, which is suitable for scenarios with extremely high quality requirements) and lossy segmentation (some information is lost during the segmentation process, which usually achieves a higher segmentation ratio and is suitable for scenarios with high requirements on book length (web page loading, limited storage space)). The segmentation method is adaptively selected based on the actual instructions. The query statement segmentation formula is:

[0087] Function word segment R = {r1, r2, ..., r m}, define the entropy value of each segment as H(r i ), then the maximum split ratio of the query statement is:

[0088]

[0089] Among them, P is the total function words of the query sentence, P(r i ) is the number of functional words in the segment.

[0090] 3. Permission configuration unit: As one of the core components of the device, it is mainly used to set reading, updating, and forwarding permissions for multimodal books based on client accounts and security instructions. It mainly includes:

[0091] (1) Client license level classification function: Adapts to three client account configurations:

[0092] Server: Has the highest permissions, performs device configuration, license scheduling, client storage and other analysis, and reads almost all functions in the device;

[0093] Updater: has certain permissions and can update content, update books, update information, etc., but cannot update device settings or schedule permissions;

[0094] Reader: The lowest level of permission, only able to read books or information, unable to update or configure analysis;

[0095] (2) Dynamic license modification function: based on the client's events and time range, the license is updated regularly. By analyzing the client's event pattern and reading history, the device can autonomously recommend license modifications to ensure the dynamic and real-time nature of license configuration. Specifically, a dynamic license recommendation algorithm is used. The client event sequence is S = {s1, s2, ..., s k}, the permission priority is L = {l1, l2, ..., l m}, then the probability that the client belongs to a certain permission priority is:

[0096]

[0097] (3) Information security audit function: This function ensures the collection and location of the license scheduling, change and adoption process. This function collects detailed information for each analysis and analyzes these information to discover potential security threats and violations. It mainly includes:

[0098] Permit scheduling and change collection: Whenever a device's permission configuration changes, a collection will be generated, detailing the time of the permission change, the person who performed the change, the content of the change, etc.

[0099] Analysis event positioning: collect the analysis of books, information or devices by each client, including reading, updating, and deleting. By collecting the analyzed client, time, and analysis content, we can locate the specific analysis;

[0100] Security risk analysis: By analyzing the client-side analysis and collection, the device identifies abnormal events such as excessive license enhancement, illegal reading, frequent license changes, etc., promptly issues an alarm, and notifies the server to handle it.

[0101] 4. Tag storage unit: used to collect and locate the update history of multimodal books, adapt to tag backtracking and multi-tag storage, mainly including:

[0102] (1) Update collection and timing saving function: The update collection and timing saving function will obtain an independent tag number every time a query statement book is updated, and collect the specific content of the update, including:

[0103] Book length: The updated book length reflects the amount of changes in the book content;

[0104] Update time: collect the timestamps of book updates for locating the past;

[0105] Updater: Collect the clients that perform update analysis for later review and location.

[0106] These updated collections will be stored in a dedicated tag information library to form a complete update history chain. Each time an update is made, the device will regularly save a copy of the update and obtain a unique tag number. The tag number is obtained using an incremental numbering or timestamp-based strategy. In this way, the client can easily locate the past status of the book, read the evolution of the book content, and ensure that any changes can be traced.

[0107] (2) Tag comparison and duplicate resolution function: When multiple people collaborate to update the same book, it may happen that multiple clients update the same book at the same time. In order to avoid duplicate book content, the tag storage unit provides a tag comparison and duplicate resolution function. The tag comparison function compares the differences between different tags and identifies which parts have changed at regular intervals. The device analyzes the differences in text, information or query statements between each tag and obtains a difference report. The report includes the updated part (marking the changed content between the two tags for easy identification by the client) and the updated time and updater (showing who updated the book and when). For duplicate tags, the device provides three options: latest tag priority (retain the update of the latest tag and ignore the update of the previous tag), merge updates of different tags (merge updates of different tags together, usually requiring trigger intervention, and the client decides how to merge based on the specific content), and manual confirmation (if the device cannot resolve the duplication at regular intervals, the client manually selects which tag's updated content should be retained). These three options allow the client to choose how to merge these tags.

[0108] (3) Timestamp backtracking function: allows the client to read all the past tags of the book through the timestamp. Each tag will be displayed on the timestamp. The client can select the tag at any time to restore. The timestamp display content includes the tag number (each tag has a unique number to facilitate identification of different tags), update time (collects the creation time of each tag to help the client quickly locate the specific tag), and a brief description of the updated content (the updated content of each tag will be briefly described so that the client can understand the changes in the book).

[0109] 5. Query and matching unit, used for rapid query and matching based on the name, keywords, attributes and content of multimodal books, mainly including:

[0110] (1) Multi-condition query function: Adapts the client to query based on multiple conditions and provides flexible query methods. The query conditions include:

[0111] Book name: Advanced or primary matching is performed based on the book name. The client searches for related books by entering some characters in the book name.

[0112] Keywords: The client searches based on the keywords attached to the book (such as "characters" and "scenery");

[0113] Creation time and update time: The client specifies a time range to search for books. For example, search for books created or updated on a specific date or within a date range.

[0114] Book Length: Adapts to filtering based on the length of books. The client sets the upper or lower limit of the book length to find books that meet the length conditions.

[0115] The combined query method is also suitable for advanced matching, basic matching and regular expression queries, and can handle a variety of complex query instructions.

[0116] (2) Keyword matching function: Books can be clustered based on keyword information, thereby improving the flexibility and convenience of queries. Books are given multiple keywords (theme, color, shooting location, etc.) and can be easily matched through these keywords.

[0117] (3) Query statement content query function: Using query statement recognition technology, relevance query is performed based on multimodal book themes (shape, color distribution, texture, etc.). The query statement content query function can not only query multimodal book names or keywords, but also query based on the characteristics of the query statement itself.

[0118] (4) Autonomous sorting and priority matching: Based on the transaction processing algorithm, the client's past query events are analyzed, and the priority of book display is predicted and automatically corrected. The device will sort the query results regularly based on the client's past queries and selections, thereby providing more personalized and relevant query results.

[0119] 6. Information snapshot and restoration unit, used to periodically snapshot the multi-level information of multimodal books and provide information restoration function after device failure, mainly including:

[0120] (1) Incremental snapshot function: When a book changes, only the changed part is snapshotted, rather than the entire book. By collecting the summary value of the book (each book gets a summary value, and periodically checks whether the summary value of the book has changed. Only books with changed summary values ​​are considered to have changed, and then snapshots are taken) or adopting timestamp identification (each time a book changes, the device collects the timestamp of the book. When taking a snapshot, the device compares the timestamps. If the timestamp of the book is updated, it is considered that the book has changed, and a snapshot is taken) the changed content, efficient information snapshots are achieved, saving storage space and bandwidth.

[0121] (2) Periodic snapshot and manual snapshot functions: The client sets the periodic execution of snapshots, such as daily, weekly, and other periodic snapshots. This method ensures that the client's books are snapshotted within a certain period of time, without the need for manual analysis, making it easier to keep the information up to date. The device also allows the client to manually initiate snapshot analysis. When needed, the client presses the snapshot button, selects the content and storage location of the snapshot, and takes an instant snapshot. Regardless of whether it is a periodic snapshot or a manual snapshot, the snapshot is stored in the local storage device or in the cloud.

[0122] 7. Inventory collection and analysis unit, used to collect client analysis events and device operating status, and obtain analysis reports for device performance analysis, mainly including:

[0123] (1) Client analysis list collection function: The device collects all analysis events of each client, including book upload, download, update, deletion, reading, etc. Each analysis collects the analysis time, client ID, analysis type and analysis object book information;

[0124] (2) Information analysis and recommendations: Through big information analysis technology, the client event list and device operation information are analyzed to mine the potential instructions of the device, such as loading speed bottlenecks, books with the highest reading frequency, etc. Based on the analysis results, the device will receive recommendations at regular intervals, such as cache strategies, book reading paths, etc., to improve overall performance.

[0125] Reference Figure 1, an audio book query and matching method and device based on an AI model, specifically comprising the following steps:

[0126] Based on the priority relationship and storage instructions of multimodal books, multimodal book storage nodes are dynamically obtained through multi-level indexing, and node clustering algorithm is used to integrate books. Books are quickly searched and matched through keyword processing.

[0127] Multimodal books are converted, segmented, sharpened, and colored. Multimodal books are converted into different modalities in parallel, and algorithmically corrected for multimodal length based on lossless segmentation of dynamic segments.

[0128] Set reading, updating, and forwarding permissions for multimodal books based on client accounts and security instructions, configure permissions for different accounts, and dynamically modify permissions based on client event analysis and time ranges;

[0129] The update history of multimodal books is collected and located, adapted to tag backtracking and multi-tag storage. Every time a book is updated, the update collection is saved regularly, including the book length, update time and updater information;

[0130] Multimodal books are quickly searched and matched based on names, keywords, attributes, and content. Clients can query based on multiple conditions, adapting to advanced and basic matching, as well as regular expression queries. Multimodal books are clustered by keywords, and clients can match books based on multiple keyword combinations. They also utilize query statement recognition technology to query content relevance based on book topics. Transaction processing algorithms analyze past client query events and autonomously adjust the order of query results.

[0131] It provides periodic snapshot and restore functions for books to ensure the integrity and security of information. When a book changes, only the changed part is snapshotted, rather than the entire book. The changed part is identified by a summary value or timestamp, and it is adapted to take regular snapshots. It also allows clients to manually initiate snapshot analysis.

[0132] Collect client analysis and operation status in real time, obtain information analysis reports for performance analysis, collect all analysis events of each client, and save information such as analysis time, client ID, analysis type, etc., analyze client events and operation information, explore potential instructions for execution, and obtain timely execution suggestions based on analysis results.

Claims

1. An audio book search and matching device based on an AI model, characterized in that: The following units are included: A multi-level book storage unit is used to perform multi-level integration and clustered storage of books based on the priority relationship and storage instructions of multimodal books. The multi-level integration is based on multi-level indexing and adopts a node clustering algorithm to integrate books. Books can be classified based on keywords, time, length, and book theme. Node changes are collected through a summary association table to adapt to cross-node parallel synchronization and information synchronization. A query statement processing unit, used for performing mode conversion, segmentation, clarity correction and color processing on multimodal books; A permission configuration unit, used to set reading, updating, and forwarding permissions for multimodal books based on client accounts and security instructions; The tag storage unit is used to collect and locate the update history of multimodal books, and is suitable for tag backtracking and multi-tag storage; A query and matching unit for fast query and matching based on the name, keywords, attributes and content of multimodal books; An information snapshot and restoration unit, which is used to periodically take snapshots of the multi-level information of multimodal books and provide information restoration after device failure; The inventory collection and analysis unit is used to collect client analysis events and device operation status, and obtain analysis reports for device performance.

2. The device according to claim 1, characterized in that The multi-level book storage unit also includes: (1) The book node construction function based on multi-level index is used to dynamically obtain multi-node multi-modal book storage according to the tree, which involves the node clustering algorithm: The multimodal book set is F = {f1, f2, ..., f n }, the node tree is T = (V, E), then the node objective function is: Among them, ω i is the importance weight of the book, d is the correlation between the book and the node center, α is the correlation coefficient, and D(T) is the node depth of the node tree; (2) Book clustering and annotation functions, which enable fast matching and classification through keyword processing; (3) Dynamic correction function between book nodes, using summary association table to collect node changes, adapting to cross-node parallel synchronization and information synchronization.

3. The device according to claim 1, characterized in that The query statement processing unit further includes: (1) Multi-modal parallel mode conversion function, adapting to multiple mode conversions; (2) Query statement quality improvement function, using neural network model to update autonomously; (3) Query sentence segmentation function, through the adaptive selection of lossless segmentation and lossy segmentation algorithms, taking into account the length and quality of the book, the query sentence segmentation function is based on the lossless segmentation of dynamic segments, and the function word segment R = {r1, r2, ..., r m }, define the entropy value of each segment as H(r i ), then the maximum split ratio of the query statement is: Among them, P is the total function words of the query sentence, P(r i ) is the number of functional words in the segment.

4. The device according to claim 1, characterized in that The license configuration unit further includes: (1) Client permission level division function, adapting to three account configurations: server, updater, and reader; (2) Dynamic license modification function, which updates the license regularly based on client events and time ranges. Using the dynamic permission recommendation algorithm, the client event sequence is S = {s1, s2, ..., s k }, the permission priority is L = {l1, l2, ..., l m }, then the probability that the client belongs to a certain permission priority is: (3) Information security audit function, which collects and analyzes the status of license scheduling and adoption.

5. The device according to claim 1, characterized in that The tag storage unit further includes: (1) Update collection and timing saving function: based on the query statement, each update of the book is saved regularly. Each update is given an independent tag number, and the specific content of the update is collected. The updated collection will be stored in a special tag information library for easy positioning and comparison in the future; (2) Tag comparison and duplication resolution function: When multiple clients update the same book in parallel, this unit will periodically compare the differences between tags and determine whether there are duplications based on the content duplication detection algorithm. For duplicate tags, the device will prompt the client to manually confirm or select a matching merge strategy; (3) The timestamp backtracking function is adapted to allow the client to read all the tags of a book through the timestamp and select the book tag at any time to restore. The timestamp display shows the update history of each tag, including the tag number, update time and a brief description of the update content. The client can easily go back to a certain tag and restore the past status.

6. The device according to claim 1, characterized in that The query and matching unit further includes: (1) Multi-condition query function: the client can query based on book name, keyword, creation time, update time, and book length, and is compatible with advanced matching, primary matching, and regular expression queries; (2) Keyword matching function: multimodal books are clustered by adding keywords, and the client can match books based on keyword combination conditions; (3) Query statement content query function, which uses query statement recognition technology to perform content relevance query based on multimodal book topics; (4) Autonomous sorting and priority matching: Through transaction processing algorithms, the client's past query events are analyzed, the priority of book display is independently corrected, and personalized query results are provided to improve query efficiency and accuracy.

7. The device according to claim 1, characterized in that The information snapshot and restoration unit further includes: (1) Incremental snapshot function, which only takes a snapshot of the changed part of a book when it changes, rather than a full snapshot. By collecting the summary value of the book or using a timestamp to identify the changed content, efficient information snapshot is achieved, saving storage space and bandwidth; (2) Periodic snapshot and manual snapshot functions. The device is adapted to take regular snapshots and also allows the client to manually initiate snapshot analysis. The snapshot content is stored locally or in the cloud.

8. The device according to claim 1, characterized in that The inventory collection and analysis unit also includes: (1) Client analysis list collection function: the device collects all analysis events of each client, including book upload, download, update, deletion, and reading analysis. Each analysis collects the analysis time, client ID, analysis type, and analysis object book information; (2) Information analysis and recommendations: Through big information analysis technology, the client event list and device operation information are analyzed to mine the potential instructions of the device, including loading speed bottlenecks and the most frequently read books. Based on the analysis results, the device will receive recommendations at regular intervals, including cache strategies and book reading paths, to improve overall performance.

9. An audio book query and matching method based on an AI model, characterized in that: The following steps are involved: Step 1: Based on the priority relationship and storage instructions of multimodal books, the device dynamically obtains multimodal book storage nodes through multi-level indexing and uses node clustering algorithm to integrate books. Books are quickly searched and matched through keyword processing. When books are synchronized or reintegrated, a summary association table is used to collect node changes and adapt to cross-node parallel synchronization and information synchronization; Step 2: Multimodal books undergo modal conversion, segmentation, clarity correction, and color processing. Multimodal content is converted to different modalities in parallel and adaptively. Query quality is independently improved through a neural network model to enhance detail and clarity. Query segmentation uses both lossless and lossy segmentation algorithms. A lossless segmentation algorithm based on dynamic segmentation is used to correct the multimodal length to balance book volume and quality. Step 3: Based on the client account and security instructions, the device sets the reading, updating, and forwarding permissions for multimodal books. The device configures permissions for different accounts and dynamically modifies permissions based on client event analysis and time range. A dynamic permission recommendation algorithm is used to predict and modify client permissions. The device collects and locates client analysis events and permission adoption status to ensure transparency and localizability of the analysis. Step 4: Collect and locate the update history of multimodal books, adapt to tag backtracking and multi-tag storage. Every time a book is updated, the device regularly saves the update collection, including the book length, update time and updater information, and obtains a new tag number. The device compares the content of books with different tags, detects duplications and prompts the client to select a duplication resolution strategy. The client reads all the book's tags in the past through the timestamp and restores the book tag at a certain moment. Step 5: Multimodal books are quickly searched and matched based on names, keywords, attributes, and content. The client searches based on book names, keywords, creation time, update time, and book length, and is compatible with advanced and elementary matching and regular expression queries. Multimodal books are clustered by keywords, and the client can match books based on multiple keyword combinations. The device also uses query statement recognition technology to search for content relevance based on book topics. It analyzes past client query events through transaction processing algorithms and autonomously modifies the order of query results to improve query accuracy and efficiency. Step 6: The device provides periodic snapshot and restore functions for books to ensure the integrity and security of information. When a book changes, the device only snapshots the changed parts, not the entire book. The changed parts are marked with a summary value or timestamp. The device is adapted to take regular snapshots and allows the client to manually initiate snapshot analysis. Step 7. The device collects client analysis and device operation status in real time, and obtains information analysis reports to analyze device performance. The device collects all analysis events of each client and saves analysis time, client ID, and analysis type information. The device uses big information analysis technology to analyze client events and device operation information, explore potential instructions performed by the device, and obtain recommendations based on the analysis results in a timely manner to improve overall performance.

Images