Ownership Claim Release Request System

Abstract

The present invention relates to an automated system for efficiently resolving ownership claim issues arising from the use of sound sources on online video platforms (e.g., YouTube). The effects of the present invention are as follows. First, repetitive and high-volume claim requests can be listed, managed, and tracked within the system, enabling efficient processing. Second, even if the song title is the same, the exact audio source can be identified by utilizing stored metadata such as the composer's name, label, and original publisher. Third, the system can be accessed on mobile devices, enabling immediate response without spatial or temporal constraints. Fourth, the allowlist registration request function can prevent repetitive claims occurring on specific channels in advance.

Description

Technology Field

[0001] The present invention relates to an automated system for efficiently resolving ownership claim issues arising from the use of sound sources on online video platforms (e.g., YouTube). Background Technology

[0002] Recently, the online video platform YouTube has been operating a proprietary software system called 'Content ID' to protect creators' copyrights and support monetization. This system identifies matching content by comparing fingerprints of registered audio and video, and content owners can choose one of three options for the matched videos: blocking views, generating ad revenue, or tracking viewing statistics.

[0003] However, due to these automated systems, problems frequently arise where advertising revenue is attributed to rights holders or videos are blocked, even for users who have obtained a legal license, because ownership claims are raised. In particular, under a recently implemented policy, the damage to licensed users is increasing, as videos are immediately blocked if ownership claims are detected in short videos longer than one minute.

[0004] In order for the rightful rights holder to clear these claims, they must contact the original publisher or an authorized monitoring agency (such as Adrev, Fuga, or Orfium) directly to request a manual release. However, there is the inconvenience of having to manually verify contact information for each track, repeatedly write emails in a fixed format, as the original publisher and monitoring agency differ for each track. Furthermore, revenue losses resulting from the absence of the person in charge or delays in response directly translate into damages for the broadcasters or production companies.

[0005] To address these issues, there is an allowlist method that pre-registers specific channels; however, this does not serve as a practical solution because it requires individual requests for each label and is difficult to manage depending on the license duration. Furthermore, it is realistically very difficult for music users to manually create a list of planned uses before performing any music-related tasks, and such a procedure is rarely carried out in practice.

[0006] Therefore, there is an urgent need to develop an automated system that combines a sound source database and an email sending system to automatically identify rights information and quickly request the withdrawal of ownership claims. Prior art literature

[0007] Korean Patent Publication No. 10-2024-0093227 The problem to be solved

[0008] The objectives of the present invention are as follows.

[0009] First, we aim to prevent damage to the manufacturer caused by the absence of personnel or delays in response by automating the complex claim resolution process that was previously done manually.

[0010] Second, it is to protect the rights and interests of license users by quickly resolving issues where advertising revenue is blocked or attributed to rights holders due to ownership claims.

[0011] Third, the purpose is to establish a rapid response system for special situations where Shorts videos, particularly those 1 to 3 minutes in length, are immediately blocked upon the occurrence of a claim.

[0012] Fourth, for each song, accurately match the contact information of the different original publishers and the various monitoring companies they have commissioned, and send emails.

[0013] Fifth, unlike existing work methods, the continuous system does not rely on a specific person in charge and provides an environment where relevant personnel can easily perform tasks and process them without errors, even in the absence of the person in charge. means of solving the problem

[0014] To solve the above-mentioned problem, the present invention includes a sound source integrated management server (100) that controls the entire process of the system and manages the transmission of claim withdrawal requests by matching sound source data with user requests. The sound source integrated management server (100) includes a sound source metadata database (110) that stores unique rights information of the sound source including song title, composer name, publisher name, and label name; a sound source identification and storage module (120) for selecting a specific track from search results and temporarily storing it in a claim request target list; an email template and automatic generation engine (130) for managing standardized email forms and dynamically combining them with sound source data to construct essential information required by the rights holder; and a claim management and tracking dashboard (140) for managing all claim withdrawal request details processed through the system by database and checking the processing status in real time. The sound source metadata database (110) manages multiple lengths and edited versions of the same sound source as individual data according to the purpose of use, and each sound source data and The system provides a request for withdrawal of ownership claim characterized by storing corresponding sound source usage approval records and royalty settlement information in conjunction, and, if YouTube claim notification information is insufficient, providing an auxiliary identification function to identify the sound source by referring to history data including previously processed claim request records and response records, and the claim management and tracking dashboard (140) visually lists and provides whether each request has been sent and whether processing has been completed, provides a function to manage modification and resending of previously sent records, and tracks viewership statistics and analysis data of the video to determine the impact of the claim occurrence on the business. Effects of the invention

[0015] The effects of the present invention are as follows.

[0016] First, repetitive and high-volume claim requests can be listed, managed, and tracked within the system, enabling efficient processing.

[0017] Second, even if the song title is the same, the exact audio source can be identified by utilizing stored metadata such as the composer's name, label, and original publisher.

[0018] Third, the system can be accessed on mobile devices, enabling immediate response without spatial or temporal constraints.

[0019] Fourth, the allowlist registration request function can prevent repetitive claims occurring on specific channels in advance. Brief explanation of the drawing

[0020] FIG. 1 is a drawing showing the overall configuration of the present invention. FIG. 2 is a diagram showing the configuration of the sound source integrated management server (100) of the present invention. FIG. 3 is a diagram showing the sequence of operations according to the system of the present invention. FIGS. 4 to 12 are drawings relating to embodiments according to the present invention. Specific details for implementing the invention

[0021] Hereinafter, various embodiments are described in more detail with reference to the attached drawings. The embodiments described in this specification may be modified in various ways. Specific embodiments may be depicted in the drawings and described in detail in the detailed description. However, specific embodiments disclosed in the attached drawings are intended only to facilitate understanding of various embodiments. Accordingly, the technical concept is not limited by specific embodiments disclosed in the attached drawings, and it should be understood that it includes all equivalents or substitutions that fall within the spirit and scope of the invention.

[0022] Terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but these components are not limited by the aforementioned terms. The aforementioned terms are used solely for the purpose of distinguishing one component from another.

[0023] Functions related to artificial intelligence according to the present disclosure are operated through a processor and memory. The processor may be composed of one or more processors. In this case, the one or more processors may be general-purpose processors such as CPUs, APs, and DSPs (Digital Signal Processors), graphics-dedicated processors such as GPUs and VPUs (Vision Processing Units), or artificial intelligence-dedicated processors such as NPUs. The one or more processors control the processing of input data according to predefined operation rules or artificial intelligence models stored in memory. Alternatively, if the one or more processors are artificial intelligence-dedicated processors, the artificial intelligence-dedicated processors may be designed with a hardware structure specialized for processing a specific artificial intelligence model.

[0024] The predefined rules of operation or artificial intelligence models are characterized by being created through learning. Here, being created through learning means that a basic artificial intelligence model is trained using a number of training data by a learning algorithm, thereby creating predefined rules of operation or artificial intelligence models configured to perform desired characteristics (or objectives). Such learning may be performed on the device itself where the artificial intelligence according to the present disclosure is executed, or it may be performed through a separate server and / or system. Examples of learning algorithms include supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning, but are not limited to the examples described above.

[0025] An artificial intelligence model can be composed of multiple neural network layers. Each of the multiple neural network layers has multiple nodes and weight values, and performs neural network operations through calculations between the results of previous layers and the multiple weights. The multiple weights possessed by the multiple neural network layers can be optimized based on the learning results of the artificial intelligence model. For example, multiple weights can be updated so that the loss value or cost value obtained by the artificial intelligence model during the learning process is reduced or minimized. Additionally, to minimize the loss value or cost value, multiple weights can be updated in a direction that minimizes the gradient associated with the loss value or cost value. Artificial neural networks may include deep neural networks (DNNs), such as Convolutional Neural Networks (CNNs), Deep Neural Networks (DNNs), Recurrent Neural Networks (RNNs), Restricted Boltzmann Machines (RBMs), Deep Belief Networks (DBNs), Bidirectional Recurrent Deep Neural Networks (BRDNNs), or Deep Q-Networks, but are not limited to the examples mentioned above.

[0026] A network is a network that serves as a transmission path for web pages; it may be a closed network such as a LAN (Local Area Network) or WAN (Wide Area Network), but it is desirable for it to be an open network such as the Internet. The Internet refers to a global open computer network structure that provides the TCP / IP protocol and various services existing at its upper layers, namely HTTP (HyperText Transfer Protocol), Telnet, FTP (File Transfer Protocol), DNS (Domain Name System), SMTP (Simple Mail Transfer Protocol), SNMP (Simple Network Management Protocol), NFS (Network File Service), and NIS (Network Information Service).

[0027] The terminal can be implemented in various forms. For example, the terminal described in this specification may be a mobile terminal such as a smartphone, tablet PC, PDA (Personal Digital Assistant), PMP (Portable Multimedia Player), or MP3 Player, as well as a fixed terminal such as a smart TV or desktop computer.

[0028] In this specification, terms such as “comprising” or “having” are intended to specify the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof. When a component is described as being “connected” or “connected” to another component, it should be understood that it may be directly connected to or connected to that other component, or that there may be other components in between. On the other hand, when a component is described as being “directly connected” or “directly connected” to another component, it should be understood that there are no other components in between.

[0029] Meanwhile, a "module" or "part" for a component as used in this specification performs at least one function or operation. Furthermore, a "module" or "part" may perform a function or operation by hardware, software, or a combination of hardware and software. Additionally, a plurality of "modules" or a plurality of "parts," excluding a "module" or "part" that must be performed on specific hardware or on at least one processor, may be integrated into at least one module. A singular expression includes a plural expression unless the context clearly indicates otherwise.

[0030] In addition, power, power transmission, and control therefor for the following assembly configurations and embodiments, including "by control," follow conventional technology including terminals, applications, hardware control modules, etc., so they are omitted to avoid redundancy.

[0031] In addition, the operation embodiments and configurations described in a general manner without being explained in detail below follow the prior art and are omitted in order to focus on describing the purpose of the present invention and the resulting effects.

[0032] Furthermore, in describing the present invention, if it is determined that a detailed description of related known functions or configurations may unnecessarily obscure the essence of the invention, such detailed description is abbreviated or omitted.

[0034] The present invention relates to an automated system for efficiently resolving ownership claim issues arising from the use of sound sources on online video platforms (e.g., YouTube).

[0035] Referring to FIG. 1, the present invention allows a user terminal (10) and a sound source integrated management server (100) to communicate via a network.

[0037] The user terminal (10) is a device that transmits and receives data to and from the sound source integrated management server (100) via a communication network, and includes not only personal computers (PCs) but also mobile devices. This supports relevant personnel in processing tasks without errors even if they lack understanding of the tasks and historical information, and provides an environment that allows for a quick response to claim release requests regardless of location.

[0038] Specifically, the user terminal (10) is equipped with the following interface and functions.

[0039] The user accesses the Hi-Five online system through the user terminal (10) and performs a login, thereby accessing a personalized claim management environment.

[0040] Additionally, the user searches for the necessary audio source by entering at least one keyword among the track name, composer name, or publisher name through the search column provided on the screen. By clicking the 'Add' button of a matching track among the search results, the user requests that the corresponding audio source information be temporarily stored in the audio source identification and storage module (120). At this time, a pop-up window indicating 'Adding successful' is displayed at the top right of the terminal screen to provide visual feedback to the user.

[0041] Additionally, the user enters the mail claim creation screen based on the list of sound sources searched through the user terminal (10) and selects an email template through the dropdown menu. Then, the user enters or modifies the following required information.

[0042] - Video URL: Directly enter or add the link address of the YouTube video where the claim occurred.

[0043] - User and Recipient Information: Enter the sender's name (in English) and email address, the email addresses of CC recipients, and business partner (customer) information.

[0044] - Body Editing: In addition to the audio list and label information automatically entered by the system, the email subject and body content can be manually modified as needed.

[0046] The user is equipped with a control button that can send commands to the server to save or send the entered content. Additionally, through the claim management list screen, the user checks a list sorted in order of recent requests and monitors the processing status of each request in real time, such as whether it is in progress or completed.

[0048] The sound source integrated management server (100) controls the entire process of the system and is a higher-level server that manages the transmission of claim withdrawal requests by matching sound source data with user requests.

[0049] The above-mentioned integrated sound source management server (100) centrally manages sources such as sound source files, license approval history, and royalty settlement information, and automates the sound source identification and email sending procedures that were previously performed manually by logically connecting them.

[0051] FIG. 2 is configured to include a sound source metadata database (110), a sound source identification and storage module (120), an email template and automatic generation engine (130), and a claim management and tracking dashboard (140) as components of the sound source integrated management server (100) of the present invention.

[0053] The specific details of each component are as follows.

[0055] The sound source metadata database (110) is a data storage that stores unique rights information of sound sources, including song title, composer name, publisher name, and label name.

[0056] If there is a song with the same title in the above-mentioned audio metadata database (110), uniqueness is determined by utilizing composer or label information, and contact and email address information of monitoring companies (Adrev, Fuga, Orfium, etc.) corresponding to the audio source is retained.

[0058] The audio metadata database (110) is a core repository of a system that manages a complex copyright ecosystem and technical identification information, going beyond simply listing audio information.

[0060] Specifically, it is equipped with the following advanced data structures and functions.

[0061] First, in addition to general song title and composer name information, it is possible to apply major identifiers such as unique identification codes from domestic and international copyright management organizations (KOMCA code, KOSCAP code), ISRC, ISWC, and UCI codes, as well as BMAT IDs used in global music monitoring systems, and it also supports integration with other monitoring systems.

[0062] In addition, it links with HiFive NAS filename information to identify physical locations within the server, synchronizing the logical information in the database with the actual audio resources.

[0063] Second, even for the same song, various versions depending on the purpose of use (e.g., Full Mix, Alt Mix, 60 sec, 30 sec, 15 sec, etc.) are managed as individual data. This is to support the inclusion of the most accurate metadata in claim release requests, tailored to the actual length or format of the audio used by the user in the YouTube video.

[0064] Third, based on label and original publisher information for each song, the system stores a 1:1 or 1:N mapping of the dedicated receiving email addresses of monitoring companies (e.g., Adrev, Fuga, Orfium, HAAWK, etc.) to which the rights holder has delegated YouTube Content ID operations. This allows the system to automatically identify the correct recipient simply by the user selecting a song.

[0065] Fourth, this database goes beyond being a simple information repository and is organically linked with music usage approval records and royalty settlement information. This serves as supporting data that allows for the immediate internal verification of whether the user possesses a legitimate license upon a claim release request.

[0066] Fifth, to prepare for cases where it is difficult to identify the audio source due to insufficient YouTube claim notification information, records of previous claim requests and responses are stored as historical data. By utilizing this, supplementary grounds for judgment are provided to identify the accurate audio source even for claims with missing information, by referring to past successful resolution cases.

[0068] The sound source identification and storage module (120) is for selecting a specific track from the search results and temporarily storing it in the claim request target list.

[0069] The audio source identification and storage module (120) matches fragmentary information confirmed by the user through YouTube claim notifications with sophisticated metadata within the system and acts as a bridge to safely transmit this to the subsequent process, the mail generation stage. Specifically, it includes the following detailed functions.

[0070] - Multi-condition search and filtering function: Users can enter a search term by selecting one or more conditions among track name, composer name, and publisher name, and the system extracts a list of audio files that meet the conditions in real time by linking with the audio metadata database (110).

[0071] - User feedback-based confirmation process: When a user clicks the 'Add' button for a track that matches the song with the actual claim in the search results list, a notification popup with 'Add successful: (Track name)' is displayed at the top right of the screen to visually confirm to the user that the data has been successfully temporarily saved.

[0072] - Automatic data loading and synchronization: Song information saved via the 'Add' function is automatically retrieved on the mail claim creation page. At this time, not only the song title but also the composer name, label name, and original publisher information are retrieved together and mapped in real-time to the automatic entry areas (#{track_title}, #{licensing_information}, etc.) within the email body.

[0073] - Support for listing for bulk request processing: Even when claims occur for multiple songs, users can organize a list of items to be processed by sequentially searching and adding them. This enables the efficient processing and listing of repetitive and bulk claim requests within a single system without omission.

[0074] - Auxiliary identification function when information is insufficient: Even when information such as the song title is provided incompletely in YouTube claim notifications, this module provides an interface that can identify the correct audio source by comparing the stored metadata with content location information verified directly by the user.

[0076] The email template and automatic generation engine (130) is a module that manages standardized email forms and dynamically combines them with audio data to accurately and quickly configure essential information required by the rights holder (monitoring company). Specifically, it includes the following detailed functions.

[0077] - Management of customized templates for various rights holders: Since monitoring companies (e.g., Syncvault, UJOYSOUND, Epic Elite, HAAWK, etc.) may have different preferred or required email formats, multiple templates optimized for each company are pre-registered and can be selected via a dropdown menu.

[0078] - Sophisticated Variables System: In addition to just song titles and URLs, detailed information such as the following is managed in the form of variables and automatically converted into real-time data at the time of transmission.

[0079] #{receiver_name}: Recipient or receiving company name

[0080] #{licensing_information}: Name of the label or original publisher used to prove the rights to the audio file

[0081] #{sender_name} and #{sender_email}: English name and contact information of the person requesting the release of the claim

[0082] #{company_name}: Name of the sending service operator (e.g., Hi-Five)

[0083] - Automatic mapping and entry of recipients: Based on label information identified in the sound source metadata database (110), the dedicated recipient email address (e.g., ugc@fuga.com, etc.) of the monitoring company managing the label is automatically entered to prevent the risk of misdelivery due to manual entry.

[0084] - Flexible editing and final review features: Email subjects and body text automatically generated by the system can be made final through the user interface (UI) before sending. This is intended to allow for flexible responses to situations where additional explanations are required for specific cases or where formatting needs to be fine-tuned.

[0085] - Minimizing damage by shortening response time: In particular, in special cases where Shorts videos of 1 to 3 minutes in length are immediately blocked upon the occurrence of a claim, the time of content exposure interruption is minimized by generating an immediate release request email through a pre-configured template.

[0086] The configuration of such an engine automates the complex email composition process that was previously done manually, preventing damage caused by the absence of personnel or delays in manual response, and enabling the efficient processing of repetitive bulk requests.

[0088] The claim management and tracking dashboard (140) is an integrated management interface that databases and manages all claim withdrawal requests processed through the system and checks the processing status in real time to ensure business continuity and accuracy. Specifically, it includes the following detailed functions.

[0089] - Multidimensional Data Listing and Visualization: For all sent requests, a detailed list is provided in a table format, including YouTube video titles, client names, label names, track names, composer names, and sender information. This allows administrators to grasp the details of individual requests at a glance.

[0090] - Real-time Status Tracking System: Displays real-time updates on the 'Completion Status (In Progress / Completed)' and 'Dispatch Status (Sent / Not Sent)' based on the progress stage of each request. Additionally, it supports timeline-based management by recording both the creation date and the actual processing completion date.

[0091] - Sophisticated search and filtering functions: Various search columns such as subject, business partner, label, track, and sender are provided to quickly retrieve specific records from a large amount of data, and request history at a specific point in time can be filtered by setting a period such as creation date or processing date.

[0092] - Function to supplement and compare incomplete information: When it is difficult to identify the audio source because the information of the reporter or delegator (original publisher) is missing from the YouTube claim notification information, the past request records and processing history accumulated within the dashboard are searched to find the accurate rights holder information and used as a control group dataset to respond.

[0093] - History-based modification and resending management: Provides a 'modification' function for previously sent records, supporting a re-response by supplementing content based on existing data in the event of information errors or additional requests.

[0094] - Utilization of Statistical and Analytical Data: By tracking viewership statistics and analytical data generated by the video in question, it provides metrics that enable data-driven assessment of the business impact of claims.

[0096] FIG. 3 is a diagram showing the sequence of operations according to the system of the present invention. With reference to FIG. 3, the operation of the present invention is explained as follows.

[0098] S1) Sound source search and identification step

[0099] The sound source search and identification step (S1) is a step for transmitting a search query, such as a track name, composer name, or publisher name entered by the user through the user terminal (10), to the sound source integrated management server (100).

[0100] Through the above-mentioned sound source search and identification step (S1), the sound source integrated management server (100) receives the query and compares it with the sound source information stored in the sound source metadata database (110), and the sound source metadata database (110) extracts and returns detailed metadata such as the KOMCA code, KOSCAP code, and label name of the sound source that meets the search conditions.

[0102] S2) Select target audio source and add step

[0103] The target sound source selection and storage step (S2) is a step in which the sound source identification and storage module (120) is activated when the user selects a specific track among the search results that requires claim release through the user terminal (10).

[0104] Through the above target sound source selection and storage step (S2), the sound source identification and storage module (120) stores the song title, composer name, and label / publisher information of the selected sound source in a temporary storage, and upon success, sends a 'storage success' notification to the user terminal (10).

[0106] S3) Email Template Retrieval and Auto-Generation Steps

[0107] The email template calling and automatic generation step (S3) is a step in which, when a user requests the creation of a claim through a user terminal (10), the email template and automatic generation engine (130) calls the track information stored in the sound source identification and storage module (120) and the rights holder information of the sound source metadata database (110).

[0108] Through the email template calling and automatic generation step (S3), the email template and automatic generation engine (130) converts pre-set replacement characters (e.g., #{track_title}, #{licensing_information}) into actual data and, at the same time, automatically maps and enters the email address of the monitoring company (e.g., Adrev, Fuga, etc.) to be the recipient based on the label information of the sound source.

[0110] S4) Step of entering and sending additional user information

[0111] The user additional information input and sending step (S4) is a step in which the user finally inputs the YouTube video URL and this information to be sent directly through the user terminal (10).

[0112] After final confirmation through the above user additional information input and sending step (S4), when the user issues a sending command, a withdrawal request email is sent to an external rights holder monitoring company through the integrated sound source management server (100).

[0114] S5) Claim Status Update and Trace Step

[0115] In the claim status update and tracking step (S5), the request is registered in the claim management and tracking dashboard (140) at the same time as the email is sent.

[0116] The claim management and tracking dashboard (140) updates and lists status information such as whether it was sent (sent / not sent), whether it was completed (in progress / completed), and the date of processing in real time.

[0118] FIG. 4 is a track selection screen through a sound source search and identification module as an embodiment of the present invention.

[0119] Referring to FIG. 4, the user accesses the sound source identification and storage module (120) through the user terminal (10) and searches for a sound source by track name, composer name, or publisher name. When the user clicks the 'Store' button of a track that matches the rights relationship in the result list extracted from the sound source metadata database (110), the system stores the unique information (KOMCA code, etc.) of the corresponding sound source in a temporary storage and outputs a success message on the screen.

[0121] Fig. 5 is a mail claim management list and creation start screen, and Fig. 6 is a metadata confirmation popup screen for a selected sound source.

[0122] Referring to FIGS. 5 and 6, when a user requests 'creation of mail claim,' the song title and composer name stored in the sound source identification and storage module (120) are called. As shown in FIG. 5, the label name of the corresponding song is automatically displayed through a pop-up window, which is based on the rights holder mapping information stored in the sound source metadata database (110).

[0125] Fig. 7 is the top screen of the email template for entering user information and video URL, and Fig. 8 is the bottom screen of the automatically generated email body and sending control.

[0126] Referring to FIGS. 7 and 8, the email template and automatic generation engine (130) automatically fills in the recipient's email address (e.g., ugc@fuga.com) and the email subject according to the selected template. When the user enters a YouTube video URL through the terminal (10), the engine (130) generates a completed English email body by replacing variables such as #{track_title}, #{youtube_url}, etc., with actual data.

[0128] Fig. 9 is a dashboard screen for monitoring the claim processing status after sending is complete, and Fig. 10 is an example screen of an actual email finally sent to the rights holder through the system.

[0129] Referring to FIGS. 9 and 10, the sent request is registered in real time on the claim management and tracking dashboard (140). The administrator can visually check and manage the sending status (sent / not sent), completion status (in progress / completed), and creation date through the list in FIG. 9. FIG. 10 shows the final result reached to the rights holder through this process, demonstrating that the request was sent while maintaining data consistency.

[0131] Fig. 11 is a list screen for managing customized mail claim templates by rights holder, and Fig. 12 is a new mail template setting screen using a variable substitution function.

[0132] Referring to FIGS. 11 and 12, the email template and automatic generation engine (130) provides a template management function to respond to different request forms for each rights holder. As shown in FIG. 12, if a replacement character (#) is set according to the body writing guidelines, it is automatically replaced with data registered in the system when transmitted in the future, thereby maximizing work efficiency.

[0134] The sound source integrated management server (100) of the present invention further includes an audio fingerprinting-based automatic sound source identification module (150), and the audio fingerprinting-based automatic sound source identification module (150) is an advanced identification means that analyzes the unique acoustic characteristics of a sound source and matches it with a sound source in a database in order to technically solve the problem of misidentification caused by insufficient text information or duplicate song titles. This module is included as a sub-component of the sound source integrated management server (100) and has the following detailed functions and technical features.

[0136] The above audio fingerprinting-based automatic sound source identification module (150) extracts an audio signal from a video file or audio file (mp3, wav, etc.) uploaded through a user terminal (10), converts the extracted signal into a frequency domain to extract unique acoustic feature points such as energy distribution over time, beats, and melody lines, and generates audio fingerprint data in the form of a digital fingerprint.

[0137] The generated fingerprint data is compared with waveform data and identification information of managed songs stored in advance in the sound source metadata database (110), and a search is performed based on the matching rate of physical data rather than text (song title), so that even if the song title is the same, sound sources with different composers or versions (Full Mix, Alt Mix, etc.) are distinguished and confirmed without error.

[0139] Specifically, this system defines and calculates the following factors to quantify the physical characteristics of a sound source.

[0140] - Data for comparison: Multidimensional vector data generated by converting an audio signal uploaded from a user terminal (10) into the frequency domain through a Fast Fourier Transform (FFT), etc., and then extracting the energy peak point. This represents the sound information of the actual video for which the current claim is made.

[0141] Here, the multidimensional vector data refers to a numeric matrix having frequency, amplitude, and time interval information of extracted energy peak points as multiple dimensional components. Specifically, the audio fingerprinting-based automatic sound source identification module (150) divides the signal converted to the frequency domain into n specific bands and generates an n-dimensional vector by arranging the feature values ​​of each band, thereby expressing the unique acoustic fingerprint of the sound source as quantified coordinate values. This configuration enables the comparison of the directionality of the frequency pattern rather than the size of the data, thereby providing a technical basis for identifying the target sound source without error even within a video containing background noise.

[0143] - Reference data: This is the frequency characteristic vector of the original sound source that is pre-stored in the sound source metadata database (110). This is an acoustic fingerprint in an optimal state without noise.

[0144] - Similarity Score: Calculated using the cosine similarity method, which calculates the angular difference between the two vectors above. As the directions pointed to by the two vectors align, the result converges to 1, which means that the patterns of the two sound signals are physically identical. Conversely, as the value approaches 0, they are interpreted as different sound sources.

[0145] Here, cosine similarity is a method that considers two data points as 'arrows (vectors)' pointing in a specific direction in a high-dimensional space, and determines similarity by measuring the 'angle' between the two arrows.

[0146] The primary reason for adopting cosine similarity in this technology is that it is not affected by the absolute size of the data. As an example of audio application, the original audio source (reference data) may be very loud and clear, whereas the audio source in a user-uploaded video (comparison data) may have low volume or be mixed with ambient noise, resulting in a much shorter arrow length (data size). While conventional distance measurement methods can misidentify two data sources as different if there is a large difference in size, cosine similarity compares only the frequency energy distribution pattern (the direction of the arrow). Therefore, a high similarity score is calculated if the core characteristics of the sound match, regardless of volume level.

[0147] The resulting value is displayed as a numerical value within a specific range, and its meaning is as follows.

[0148] When the similarity score converges to 1, the two arrows point in the exact same direction. This means that the physical patterns, such as the frequency composition and melody lines of the two sound sources, are identical, and the system determines this as a 'match'.

[0149] When the similarity score approaches 0, the two arrows are perpendicular to each other and point in completely different directions. This means that there are no common features between the two sounds, and the system judges this as a 'mismatch'.

[0150] In the audio fingerprinting module (150), cosine similarity acts as a pattern matching verification engine. In a database containing tens of thousands of sound sources, instead of simply finding songs with the same title, it calculates how geometrically similar the extracted frequency feature points are to reference data and provides objective numerical data that replaces human auditory comparison.

[0152] - Identification threshold: A logical boundary value used to determine whether a match is made. Setting this value higher (e.g., 0.98) lowers the false recognition rate but may fail to identify when background music (BGM) is quietly mixed in, while setting it lower (e.g., 0.85) detects images mixed with ambient noise well but carries the risk of misidentifying different songs with similar rhythms as the same song.

[0154] Sound source identification is performed sequentially through the following steps according to the flow of time.

[0155] 1) Signal extraction step: Separate the audio stream from the video file received from the user, and primarily sample only the signal from the timestamp interval (e.g., 1:25-2:14) provided in the YouTube claim notification.

[0157] 2) Data Normalization Step: The volume and frequency range of the sampled signal are normalized to a level comparable to reference data and processed.

[0159] 3) Feature point matching step: A 1:N similarity operation is performed on the comparison target vector extracted from the processed data and the candidate vectors in the database.

[0161] 4) Multi-condition evaluation step: Verify whether the calculated similarity score meets a preset threshold, and simultaneously check whether the license for the corresponding sound source is held.

[0163] 5) Final confirmation step: Items that satisfy all conditions are set as final identification targets and transmitted to the email generation engine (130).

[0165] Final identification applies the following composite judgment criteria rather than a single score.

[0166] - Condition 1 (Required): The similarity score must exceed the system's initial value of 0.95 (based on weighted average). This threshold value is set based on general acoustic statistics during the initial stages of system operation and is subsequently automatically adjusted in increments of 0.01 by learning from actual successful unlocking cases and false positive feedback data.

[0167] - Second condition (auxiliary): The frequency of matching feature points within the timestamp interval must be 70% or more of the total.

[0169] The two conditions above are combined using the AND method. That is, no matter how high the similarity score is, if the frequency of matching feature points is low, it is considered temporary pseudo-noise and excluded. Conversely, if the frequency is high but the similarity score falls below the threshold, it is determined that there is excessive background noise, and a 'Manual Review' notification is displayed to the administrator.

[0171] In the present invention, a feature point refers to a peak point (peak points or constellation points) that has a relatively high energy intensity compared to surrounding data on a spectrogram representing an audio signal in a time-frequency plane. This is the core minimum unit of information that constitutes the fingerprint of sound. The extraction of feature points is performed through a feature point extraction engine provided within an audio fingerprinting-based automatic sound source identification module (150). That is, the received audio data is divided into a fixed time unit (Frame), and a Fast Fourier Transform (FFT) is applied to each frame to analyze the frequency components and their amplitudes for each time period. From the analyzed results, to enhance the sound identification capability, the top N frequency points with high energy relative to background noise are selected as feature points. Each feature point is quantified in the form of a coordinate value having components of [occurrence time (t), frequency (f), energy intensity (m)].

[0173] The matching frequency of feature points mentioned in the second condition (auxiliary) is calculated by the following technical logic.

[0174] Let N_total be the total number of feature points of the comparison target data existing within a specific interval designated by a timestamp, and let N_match be the number of feature points among them that are determined to have the same frequency and temporal relative position as the reference data; then, the ratio (N_match / N_total) is calculated.

[0175] In other words, while the similarity score (first condition) evaluates the overall directionality of the signal pattern, the matching frequency of feature points (second condition) quantitatively verifies how densely and accurately the key markers of the sound overlap.

[0177] It is assumed that, according to one embodiment of the present configuration, a specific section of a video uploaded by a user was analyzed and the similarity score was calculated to be 0.97.

[0179] If the current system's identification threshold is set to 0.95, the first condition is satisfied because the calculated value (0.97) exceeds the reference value (0.95).

[0180] At the same time, the frequency of matching feature points within the corresponding section is measured at 82%, satisfying the second condition (70% or more).

[0181] Since both conditions are satisfied, the system finalizes the audio source as 'Swing and Sing (Full Mix)' and immediately calls the original publisher '2 Red Jokers' dedicated template to proceed with the claim release request procedure.

[0183] Conventional text-based search methods have slow response speeds and a high risk of misjudgment because users must listen to and compare multiple songs with the same title individually when such songs exist. The present invention quantifies the physical waveform of audio into a multidimensional vector and automatically evaluates it according to strict multiple threshold logic to achieve the following effects.

[0184] Since judgments are made solely based on quantitative similarity standards (0.95 or higher) without subjective human intervention, requests to lift erroneous claims caused by misjudgment are blocked at the source. In particular, when claims arise for Shorts videos ranging from 1 to 3 minutes in length, the audio source is identified and an email is automatically generated within seconds of file upload, minimizing damage caused by video blocking. Furthermore, despite incomplete YouTube claim notification information (such as missing song titles), the technical reliability of the overall processing result is maximized by analyzing the actual video signal to trace back the rights relationship.

[0187] In addition, the above-mentioned audio fingerprinting-based automatic sound source identification module (150) receives information on the 'location where content was found (timestamp)' included in the YouTube claim notification information and intensively analyzes only a specific section (e.g., 1:25-2:14) within the uploaded file. At this time, by extracting and comparing only the audio features of the section where the claim was made, rather than the entire file, the identification speed is increased, and even sound sources that are mixed quietly, such as background music (BGM), are found with high accuracy.

[0188] Once identification is complete, the exact song title, composer name, and label name information of the corresponding song are automatically retrieved from the audio metadata database (110). The extracted information is immediately transmitted to the email template and automatic generation engine (130) without passing through the audio identification and storage module (120), thereby completely replacing the existing manual process that required a person to listen and make a judgment.

[0190] As such, the audio fingerprinting-based automatic sound source identification module (150) can make objective and accurate sound source identification possible because, unlike simple text search methods which rely on the user's subjective judgment when there are songs with the same name, this configuration is based on physical data (fingerprint). It also provides an advanced function that can backtrack rights relationships within the system database by analyzing the sound of the actual video even when the song name is omitted or incorrect information is provided in the YouTube claim information.

[0191] In addition, the step of the user manually searching for, listening to, and comparing songs is omitted, and the process proceeds automatically to the stage of generating a claim release email with only file upload and timestamp input, thereby dramatically improving the response speed. Furthermore, in the event of a claim for Shorts videos that are immediately blocked, which are 1 to 3 minutes long, the fastest identification and email sending are guaranteed, thereby minimizing damage caused by the interruption of content exposure.

[0193] The sound source integrated management server (100) may further include a response optimization and automatic feedback loop module (170), and the response optimization and automatic feedback loop module (170) is an intelligent control means that learns past claim response history to suggest an optimal response strategy for each rights holder and dynamically improves the success rate of the system through real-time result feedback.

[0194] The functions of the above response optimization and automatic feedback loop module (170) are as follows.

[0196] The above response optimization and automatic feedback loop module (170) constructs a probability model in which past claim data (time of dispatch, template used, final release status) is used as an independent variable and the claim release success rate is used as a dependent variable. When multiple templates (Type A, Type B) are tested for a specific rights holder (e.g., Adrev), if the success rate of a specific format is statistically significantly higher (e.g., 87% vs 62%), the template is automatically classified as the top priority recommendation target when a claim is made against the same rights holder in the future.

[0198] The above response optimization and automatic feedback loop module (170) learns the average release time required according to the rights holder, sound genre, and video length through time series analysis. In particular, when it detects a special situation where it is immediately blocked upon the occurrence of a claim, such as a Shorts video, the system classifies it as an 'urgent processing target' and calculates the expected release time and expected loss due to the blocking, and visualizes it to the person in charge as a top priority.

[0199] For the above time series analysis, a Recurrent Neural Network (RNN) or Long Short-Term Memory (LSTM) algorithm is utilized to learn the seasonality and variability of response patterns for each rights holder. Through this, release delays occurring at specific times (e.g., at the time of major copyright policy changes) are predicted in advance and reflected in the processing time.

[0201] The above response optimization and automatic feedback loop module (170) patterns common causes of failure (metadata errors, old song information, etc.) from rejected claim data through natural language processing (NLP). If the reason for rejection is determined to be "video URL format error," the system standardizes the URL format upon re-request and automatically refines it, and generates a modified template that corrects the defect to increase the success rate of resending.

[0202] The pattern recognition through the aforementioned Natural Language Processing (NLP) utilizes Named Entity Recognition (NER) technology to extract key reasons for rejection from the body of rejection emails, and enhances the correlation between the extracted reasons and the success of the supplemented re-request through reinforcement learning.

[0204] The above response optimization and automatic feedback loop module (170) collects all processing results (completion / modification, etc.) occurring in the claim management and tracking dashboard (140) in real time. The collected result data is immediately incorporated into the training dataset of the ML model, and the model is automatically redeployed on a weekly or monthly basis to dynamically improve the success rate and prediction accuracy of the system.

[0205] To prevent data drift, where the performance of existing training models degrades due to changes in rights holder policies or updates to YouTube's algorithms, the system performs incremental learning based on the latest 6 months of data collected in real time, and ensures the reliability of the system by automatically redeploying the model when performance metrics fall below a threshold.

[0207] The above response optimization and automatic feedback loop module (170) calculates a priority index (PI) to determine where to first concentrate limited human and technical resources among the large number of copyright claims. This is used to minimize damage to high-risk content, such as YouTube Shorts, where videos are immediately blocked when a claim of ownership is made, and to maximize work efficiency by selecting cases with a high probability of successful unlocking based on past data.

[0209] This system quantifies the following three independent factors to derive a priority index, which is a comprehensive score.

[0211] - Urgency Factor (U): A value that quantifies the risk of content visibility restrictions under YouTube's copyright policy, representing the scale of immediate damage based on the video format. Under YouTube policy, if a claim occurs on a vertical video (Shorts) between 1 and 3 minutes in length, it is immediately blocked and becomes unviewable; therefore, a higher weight is assigned compared to regular videos. This value is assigned as a constant when the system analyzes the video metadata of a claim notification and finds that a 'Shorts' tag exists or the video aspect ratio and length meet the criteria.

[0212] To determine the above urgency factor, the system parses video metadata transmitted via claim notification emails received from YouTube or APIs. It then checks the video_length field within the metadata to verify if it is 180 seconds (3 minutes) or less, determines whether it is vertical (e.g., 9:16 aspect ratio) based on the video resolution information, and checks whether the #Shorts tag is included in the description or title, or if it has a Shorts-only URL structure on the YouTube system.

[0213] For Shorts videos that meet all of the above conditions (length 3 minutes or less and vertical format), the highest grade constant (e.g., 10) is assigned because the immediate blocking policy is applied, whereas for general horizontal videos with common revenue sharing or statistics tracking options, a lower grade constant (e.g., 2) is assigned to set the relative urgency low.

[0215] - Probability of Success (P): This represents the success rate of unlocking email templates sent in the past to the rights holder and the sound source. This value is calculated by a machine learning model learning the 'sent' versus 'completed' data accumulated in the claim management and tracking dashboard (140).

[0216] The above success probability factor is a data-based indicator that indicates how effective a specific template was for a specific rights holder or monitoring company. To obtain this, historical data accumulated in the claim management and tracking dashboard (140) is fully examined, and historical records are classified by 'monitoring company ID' and 'template ID' used, and the total number of emails sent with that combination (Total) is tallied. Among the total number, the number of cases where 'completion status' is marked as 'Completed' (Completed) is counted, and a probability value between 0 and 1 is calculated using the formula P = Completed / Total.

[0217] This goes beyond simply calculating a past average; it applies an exponential moving average technique that weights the success rate over the past three months to reflect the company's latest policy changes in the success probability in real time.

[0218] The input dataset for training the above machine learning model includes identification information of the sound source (genre, country of release), the time of claim filing (day of the week, time of day), the size of the YouTube channel, and keyword frequency of rejection texts received from the same company in the past. By setting these multifaceted variables as independent variables, a success probability is derived that reflects not only the company name but also the situational context in which the claim occurred.

[0221] - Expected Time (T): This is the historical average time it takes for the monitoring company to resolve the actual claim after receiving the email. A smaller value indicates that it is a 'short-term issue' that can be resolved quickly.

[0222] The above expected processing time factor is a value that predicts the physical time required from the time the request is sent until the actual claim is released. To obtain this value, timestamp data of the creation date (sending date) and the processing completion date recorded in the dashboard (140) is utilized. Specifically, for each case completed in the past, the individual processing time is calculated by subtracting the creation date from the processing completion date, and the average time required is calculated at the level of the monitoring company (e.g., Adrev, Fuga, etc.) managing the corresponding sound source. At this time, non-business days such as weekends or public holidays are excluded from the calculation to prevent data distortion.

[0223] The smaller this factor is (the larger the exponent when located in the denominator), the more likely the case is to be resolved quickly. Through this, the system optimizes the overall speed of claim resolution by prioritizing requests from companies with high processing efficiency.

[0225] Priorities are determined through the following steps according to the flow of time.

[0226] 1) Data collection stage

[0227] When a new claim occurs, the system first collects the technical characteristics and historical data of the relevant content to prepare basic data for analysis.

[0228] Analyzes the metadata of the video transmitted via claim notification information or API. Specifically, it checks the video length (whether it is 180 seconds or less), aspect ratio (e.g., 9:16), and the presence of specific tags (e.g., #Shorts) in the title to identify whether the video is a Shorts video to which the blocking policy is immediately applied.

[0229] Based on the information of the identified sound source, the sound source metadata database (110) is queried to retrieve the past average response success rate of the company monitoring the sound source (e.g., Adrev, Fuga, etc.) and the average time taken from sending the request to actually releasing it.

[0231] 2) Weight application step

[0232] Predefined weights are applied to adjust the influence of each collected factor on the final priority.

[0233] To prevent damage caused by video blocking making viewing impossible, the highest weight is set to the Urgency Factor (U). This is a logical setting designed to prioritize cases requiring immediate action for business protection purposes.

[0234] Past success probability (P) and expected processing time (T) are auxiliary indicators that increase the efficiency of the system and are given relatively lower weights than the urgency factor, but the machine learning model fine-tunes these weight values ​​in increments of 0.1 according to the latest success data to increase precision.

[0235] The machine learning model sets the objective function to minimize the total time required until actual unlocking is completed. It calculates the error between the actual result value fed back from the dashboard (140) and the system's prediction value, and maintains prediction accuracy by dynamically updating the weights (w_1, w_2, w_3) within the priority index (PI) calculation formula through an optimization algorithm such as gradient descent.

[0237] 3) Composite Index Calculation Step

[0238] A quantified priority index (PI) is calculated by combining weights for each factor.

[0239] The 'urgency score' and 'success probability score' are added together to increase the score, and the result is converted into the reciprocal of the expected processing time (1 / T) and reflected. In other words, the shorter the processing time (the smaller the T value), the larger the reciprocal value becomes; thus, 'short-term solutions' that can be resolved within a short time receive a higher priority score. Consequently, the system derives the highest priority for cases that are urgent, have a high probability of resolution, and can be processed quickly.

[0241] 4) Ranking Confirmation Stage

[0242] Based on the calculated final score, the system's user interface is controlled to induce the actions of the person in charge.

[0243] Claims are automatically placed at the top of the list on the claim management and tracking dashboard (140) in order of highest calculated index (PI). This provides a processing order based on risk and efficiency rather than a first-in, first-out (FIFO) method.

[0244] For ultra-urgent cases where the index exceeds a specific threshold (e.g., 8.5 points or higher), a separate highlight or alarm is generated on the dashboard, allowing the person in charge to intuitively identify and take action on key content that requires immediate response among a list of thousands of items.

[0246] The reference value used when calculating the priority index initially starts with an average value set by the manager (e.g., urgency constant 5, weights 0.3 each), but as operational data accumulates, the machine learning model automatically adjusts it in increments of 0.1.

[0247] Example Case A (Shorts Video): When the urgency factor is 10 (Shorts special exception applied), the past success rate is 0.9 (90%), and the expected processing time is 1 hour. The final PI calculated by applying the weighted average is very high.

[0248] Example Case B (General Image): When the urgency factor is 2 (General Image), the success rate is 0.5 (50%), and the expected processing time is 48 hours. The PI value is calculated to be relatively low.

[0250] The system generates an 'urgent response' alarm for items where the calculated PI exceeds a reference threshold (e.g., 7.0). If multiple items have the same PI, it performs an OR auxiliary logic to place the item with the shorter expected processing time at a higher priority.

[0252] Unlike the conventional method of simply processing claims in the order they occurred (FIFO), this configuration quantitatively evaluates both the actual risk factor of video blocking (urgency) and the feasibility of resolution (success rate and time) simultaneously. Through this, administrators can selectively process short videos that can prevent immediate revenue loss and audio files with high processing efficiency from among hundreds of claims, based on technical grounds. As a result, the overall time required to clear claims is shortened, and technical benefits are achieved in that the rate of damage to content creators caused by blocking is drastically reduced.

[0254] Although the present invention has been described above with reference to one embodiment, those skilled in the art will understand that various modifications and changes can be made to the invention without departing from the spirit and scope of the invention as described in the following claims. Explanation of the symbols

[0255] User terminal (10) Sound source integrated management server (100) Sound source metadata database (110) Sound source identification and storage module (120) Email template and automatic generation engine (130) Claim management and tracking dashboard (140)

Claims

Claim 1 The system includes a sound source integrated management server (100) that communicates with a user terminal (10) via a network, controls the entire process of the system, and manages the transmission of claim withdrawal requests by matching sound source data with user requests. The sound source integrated management server (100) includes a sound source metadata database (110) that stores unique rights information of a sound source including song title, composer name, publisher name, and label name; a sound source identification and storage module (120) for selecting a specific track from search results and temporarily storing it in a claim request target list; an email template and automatic generation engine (130) for managing standardized email forms and dynamically combining them with sound source data to construct essential information required by the rights holder; a claim management and tracking dashboard (140) for managing all claim withdrawal request details processed through the system by database and checking the processing status in real time; an audio fingerprinting-based automatic sound source identification module (150); and a response optimization and automatic feedback loop module (170). The above audio metadata database (110) includes, and for the same audio source, manages multiple length and edited versions according to the purpose of use as individual data, stores audio usage approval records and royalty settlement information corresponding to each audio data in conjunction, and if YouTube claim notification information is insufficient, provides an auxiliary identification function to identify the audio source by referring to history data including previously processed claim request records and response records, and the above claim management and tracking dashboard (140) is characterized by listing and visually providing whether each request has been sent and whether processing has been completed, providing a function to manage modification and resending of previously sent records, and tracking viewership statistics and analysis data of the video to determine the impact of the claim occurrence on the business, and the above audio fingerprinting-based automatic audio identification module (150) extracts an audio signal from a file uploaded from a user terminal (10), converts the extracted signal into the frequency domain, and the energy distribution, bit,Audio fingerprint data is generated by extracting acoustic feature points including a melody line, and the generated fingerprint data is compared with waveform data and identification information of the original sound source stored in the sound source metadata database (110). A search is performed based on the matching rate of physical data rather than text information to distinguish the composer or version within the same song title and finally confirm the corresponding sound source. The audio fingerprinting-based automatic sound source identification module (150) identifies the sound source by combining a first condition for calculating a similarity score and a second condition for calculating the matching frequency of feature points using a logical AND method. The first condition determines whether the similarity score calculated through a cosine similarity calculation method, which calculates the angle difference between the audio fingerprint data and reference data within the sound source metadata database (110), exceeds a preset threshold. The second condition determines whether the matching frequency of feature points within the timestamp interval included in the YouTube claim notification information is 70% or more of the total. Only when both the first and second conditions are satisfied is the corresponding sound source finally confirmed. The method is characterized by confirming the confirmed sound source information and transmitting it to an email template and automatic generation engine (130) to automatically generate an email requesting the withdrawal of a claim of ownership, and the response optimization and automatic feedback loop module (170) utilizes Named Entity Recognition (NER) technology from the body of a rejection email received from a rights holder or monitoring company to extract and pattern the core reason for rejection through natural language processing, and if the extracted core reason for rejection corresponds to any one of metadata errors, old song information, or video URL format errors, the error is standardized and automatically refined upon re-request, and a modified template that compensates for the defect is generated and made resent through the email template and automatic generation engine (130), and to process a large number of copyright claims, the method determines whether it is a short video based on at least one of video length, aspect ratio, and title tag.Quantifying an urgency factor (U) that is assigned a higher value than a general video when determined to be a short video, a success probability factor (P) calculated by investigating the number of completed releases relative to the number of past dispatches per monitoring company and applying an exponential moving average technique that assigns weights to the success rate of the last three months within the entire historical dataset, and an expected processing time factor (T) calculated based on the difference between the creation date or dispatch date and the processing completion date recorded in the claim management and tracking dashboard (140), applying respective weights to the urgency factor (U), the success probability factor (P), and the expected processing time factor (T), while reflecting the expected processing time factor (T) in the form of an inverse to calculate a priority index (PI), automatically placing the items at the top of the list in the claim management and tracking dashboard (140) in order of highest calculated priority index (PI), and for ultra-urgent cases where the priority index (PI) exceeds a preset threshold, generating a separate highlight or alarm on the claim management and tracking dashboard (140). A system for requesting withdrawal of ownership claim, characterized by calculating a loss value based on the error between the actual processing result value fed back from the claim management and tracking dashboard (140) and the predicted value based on the calculation of the priority index (PI), and dynamically updating the weights applied to the calculation of the priority index (PI) using a gradient descent algorithm so as to reduce the loss value. Claim 2 delete Claim 3 delete Claim 4 delete Claim 5 delete Claim 6 delete

Images