Operation server for providing ai-based customized feedback through emotion analysis to support user routine execution and operation method thereof

Abstract

An operating server and a method of operation are disclosed for providing AI-based personalized feedback to support the execution of a user's routine. The method of operation of the present invention includes the steps of: receiving target data from a user terminal through an interface provided by the operating server; generating routine data through the analysis of the target data; generating planning data by linking the routine data with the user's schedule data and providing the planning data to the user terminal; receiving result data for the planning data; determining an emotional state related to the execution of the user's routine by analyzing the received result data; analyzing the characteristics of the user's routine execution based on the emotional state, routine data, and result data; and generating personalized feedback based on the emotional state and routine execution characteristics through an AI-based feedback generation model and providing the generated personalized feedback to the user terminal.

Description

Technology Field

[0001] The present invention relates to an operating server and a method of operation that provide AI-customized feedback based on sentiment analysis to support the execution of a user's routine. More specifically, the invention relates to an operating server and a method of operation that analyzes an emotional state based on a user's goal data and execution result data, and generates and provides customized feedback according to the analyzed emotional state and routine execution characteristics. Background Technology

[0002] In modern society, establishing and executing routines is widely utilized for personal self-development, health management, and the improvement of learning and work performance. For example, the way users perform daily exercise, follow study plans, and achieve goals in accordance with work schedules has established itself as a core means for daily management and self-directed achievement. However, for users to consistently execute these routines, simple schedule management functions alone are insufficient; psychological motivation, management of adherence, and emotional support must also be provided.

[0003] Conventional technology has primarily focused on schedule management systems based on calendars, alarms, and checklists. These systems were limited to simply recording planned schedules or providing notifications. Consequently, they had limitations in that they failed to account for psychological fatigue, decreased motivation, and negative emotions resulting from failure experienced by users when actually executing routines. As a result, it was common for many users to abandon routines midway after a certain period, unable to maintain consistency even if they initially established them.

[0004] Furthermore, while AI-based recommendation features are being utilized in various fields such as wellness services, learning management platforms, and fitness applications, most focus on content recommendation or simple statistical analysis. Functions that quantitatively analyze a user's emotional state and generate personalized feedback based on this to support routine execution are still lacking. In particular, even if users set the same goal, individual tendencies, emotional responses, and execution patterns differ, so the effectiveness of uniform feedback is inevitably limited.

[0005] Therefore, there is a need for technology that analyzes users' goal and execution result data to quantify their emotional states, derives routine execution characteristics by comparing them with other user groups, and provides personalized feedback using AI-based models. In other words, there is a demand for systems or methods that can play a key role in reinforcing users' motivation to sustain their routines, helping them overcome emotional difficulties, and assisting them in achieving higher levels of performance. Prior art literature

[65535] Korean Published Patent Application No. 10-2025-0125597 (Aug. 22, 2025) Korean Published Patent Application No. 10-2020-0068057 (June 15, 2020) Korean Registered Patent Application No. 10-2672317 (June 3, 2024) The problem to be solved

[0006] The objective of the present invention, which is to solve the above-mentioned problems, is to provide an operating server and a method of operation thereof that analyzes an emotional state based on user goal data and execution result data, and generates and provides customized feedback according to the analyzed emotional state and routine execution characteristics. means of solving the problem

[0007] One aspect of the present invention for achieving the above objective provides an operating server and a method of operation thereof that provides AI-customized feedback based on sentiment analysis to support the execution of a user's routine.

[0008] The above method of operation is a method of operation of an operating server that provides AI-customized feedback based on sentiment analysis to support the execution of a user's routine, comprising: receiving goal data related to the goal of a routine input from the user's terminal through an interface provided by the operating server; analyzing the goal data based on the type of goal, available execution time, and difficulty level, and generating routine data including a plurality of execution items for achieving the goal and the order of each execution item based on the analysis results; linking the generated routine data with the user's schedule data to generate planning data in which the routine execution items are placed within the user's available time interval, and providing the planning data to the user terminal; receiving result data from the user terminal including execution status data related to the execution status of each execution item included in the planning data and review data related to the user's review following the execution of each execution item; analyzing the execution status data and the review data among the received result data to extract sentiment expression data from the review data and derive an execution pattern from the execution status data, and determining the sentiment state related to the execution of the user's routine based on the extracted sentiment expression data and execution pattern; and determining the user's sentiment state, routine data, and result data among a plurality of users registered with the operating server The method includes the steps of: analyzing the routine execution characteristics of the user by comparing the routine similarity with the user with data of a user whose routine similarity is greater than or equal to a preset standard similarity; generating customized feedback through an AI-based feedback generation model that is pre-trained to generate customized feedback for the user's routine execution based on the emotional state and the routine execution characteristics; and providing the generated customized feedback to the user terminal.

[0009] The step of receiving the goal data includes receiving the user's goal data through the interface in the form of text input, optional input, numerical input, and voice input, and classifying each input value received as the goal data into at least one item among the type of goal, goal achievement period, performance frequency, preferred time interval, difficulty level, and priority.

[0010] The step of generating the routine data comprises: classifying the user's goal into one of the goal types among health management, learning, work, and lifestyle management based on input values ​​classified as goal type items in the goal data; obtaining the user's schedule data through the user terminal and determining a time interval allottable for the routine by comparing the schedule data with input values ​​classified as available time items in the goal data; calculating a difficulty score based on pre-set criteria and weights for converting the number of executions, estimated time, and ratio to the target value into a difficulty level, using input values ​​classified as difficulty level items in the goal data, the number of executions required to achieve the goal, the estimated time required, and the ratio to the target value relative to the period, and determining the calculated difficulty score as one of the difficulty grades corresponding to a pre-set threshold interval; deriving candidate execution items for achieving the goal based on the goal type, the allottable time interval, and the difficulty grade, wherein the execution items are defined as units of action that the user must actually perform, and each execution item includes the estimated time required for the corresponding action, the number of executions, necessary resources, and an indicator of contribution to goal achievement; and for each candidate execution item, the estimated time required The method further includes the steps of calculating, determining whether each execution item is executable by comparing the calculated estimated time required with the allocatable time interval, determining the execution order of each execution item based on the executable status, the difficulty level, and the required execution frequency within the goal achievement period, and generating routine data that includes each determined execution item and the execution order of each execution item.

[0011] The method further includes the steps of: providing the above-mentioned planning data to a user terminal; comparing each execution item of the routine data with an available time interval included in the user's schedule data, and generating planning data such that each execution item among the available time intervals is placed in a non-conflicting time slot based on the comparison result; and including the time slots, estimated time required, and execution completion times for each execution item in the planning data, and transmitting the included planning data to the user terminal to provide it for visual display.

[0012] The step of receiving the result data further comprises: receiving from the user terminal execution status data including whether each execution item included in the plan data is performed, the degree of progress expressed as a preset progress unit value, the time required, and whether there is a delay; receiving from the user terminal review data regarding the execution of each execution item, including review content that includes at least one of a satisfaction score entered through the user terminal, text input related to the review, and voice input related to the review; and configuring the received execution status data and review data into result data by linking them with an identifier of each execution item and a timestamp indicating the time information when each data was received.

[0013] The step of determining the emotional state further includes: a step of calculating the user’s emotional expression values ​​by identifying emotional keywords, matching emotional categories, and determining voice tone in order to extract emotional expression data from the review data, thereby calculating emotional classification values, emotional intensity values, and arousal values; a step of calculating the delay rate, completion rate, average execution speed, and deviation from the plan for each execution item to determine the execution pattern from the execution status data, and determining a quantified execution pattern value by normalizing each calculated indicator into either an aggregate value per period unit or an aggregate value per recent predetermined number of times unit; and a step of calculating an emotional index quantifying the user’s emotional state based on the weighted average of the emotional expression values ​​and the execution pattern values, and determining the emotional state related to the user’s routine execution as one of positive, neutral, or negative by comparing the calculated emotional index with a predefined threshold interval.

[0014] The step of analyzing the routine execution characteristics of the user further comprises: calculating routine similarity between the routine data of the user and the routine data of a plurality of users registered on the operating server, and determining users whose routine similarity is greater than or equal to a preset standard similarity as a similar user set; comparing the result data and emotional state of each user included in the similar user set with the result data and emotional state of the user, and calculating at least one of the user's completion rate, delay rate and persistence index, distribution of positive, neutral, and negative emotional indices, and success rate relative to difficulty grade as a routine characteristic indicator based on the comparison result; and comparing the routine characteristic indicator with a type-specific threshold interval that classifies routine execution characteristics into a plurality of types, and classifying the user's routine execution characteristics into at least one of an execution ability type, an emotional response type, and a difficulty suitability type based on the comparison result.

[0015] The step of providing the customized feedback to the user terminal comprises providing the emotional state and the routine execution characteristics as inputs to a pre-trained AI-based feedback generation model to generate customized feedback for the user's routine execution, and transmitting and providing the customized feedback to the user terminal.

[0016] The above artificial intelligence-based feedback generation model is a multi-layer neural network-based natural language generation model configured to map the routine data, result data, and sentiment index between inputs and outputs, and is pre-trained using a training data set that includes target data, routine data, result data, sentiment state labels, routine execution characteristics, and customized feedback for each user, for multiple users accumulated in advance on the operating server.

[0017] In addition, the AI-based feedback generation model applies a learning condition during the learning process that maps the feedback candidate set to include at least one of a motivational message, a guideline for improving the execution method, and a suggestion for adjusting the difficulty level corresponding to the aforementioned emotional state and routine execution characteristics.

[0018] In addition, the AI-based feedback generation model outputs customized feedback by configuring the feedback with the highest preset suitability among the feedback candidate group corresponding to the user's emotional state and routine execution characteristics input to the feedback generation model in the form of a natural language sentence. Effects of the invention

[0019] The operating server according to the present invention as described above can quantitatively calculate the emotional state by collecting and analyzing the user's goal data and execution result data, and can provide personalized feedback by identifying routine execution characteristics based on this. Accordingly, the operating server of the present invention allows the user to receive not only notifications based on a schedule regarding the routine, but also motivational messages, improvement guidelines, and difficulty adjustment suggestions suitable for their emotional state and execution patterns, thereby enhancing the continuity and efficiency of routine execution.

[0020] In addition, the operating server of the present invention can generate feedback suitable for user characteristics by utilizing multiple user data to compare and analyze the execution results and emotional responses of a group with routines similar to those of an individual user. Through this, the operating server of the present invention allows users to receive more realistic and reliable feedback by referring to a group of users in similar situations, and effectively overcome emotional and psychological barriers that may occur during the execution of routines.

[0021] Therefore, the operating server of the present invention overcomes the limitations of existing simple schedule management systems and supports routine execution by providing AI-based customized feedback tailored to the user's sentiment analysis results and execution characteristics, thereby providing tangible effects in various fields such as self-development, health management, learning, and work. Brief explanation of the drawing

[0022] FIG. 1 is a schematic diagram showing an environment in which an operation method of an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment is performed. FIG. 2 is a diagram illustrating an exemplary configuration of hardware for an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment. FIG. 3 is a flowchart illustrating an operation method performed in an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment. FIG. 4 is a flowchart illustrating a method for analyzing target data in an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment. FIG. 5 is a flowchart illustrating a method for generating routine data in an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment. FIG. 6 is a flowchart illustrating a method for generating planning data in an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment. FIG. 7 is a flowchart illustrating a method of configuring result data in an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment. FIG. 8 is a flowchart illustrating a method for determining an emotional state in an operating server that provides emotion analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment. FIG. 9 is a flowchart illustrating a method for analyzing routine execution characteristics in an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment. Specific details for implementing the invention

[0023] The present invention is susceptible to various modifications and may have various embodiments; specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the invention to specific embodiments, and it should be understood that the invention includes all modifications, equivalents, and substitutions that fall within the spirit and scope of the invention. Similar reference numerals have been used for similar components in the description of each drawing.

[0024] Terms such as first, second, A, B, etc., may be used to describe various components, but said components shall not be limited by said terms. These terms are used solely for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be named the second component, and similarly, the second component may be named the first component. The terms "and" and "or" include a combination of a plurality of related described items or any of a plurality of related described items.

[0025] When it is stated that one component is "connected" or "connected" to another component, it should be understood that while it may be directly connected or connected to that other component, there may also be other components in between. On the other hand, when it is stated that one component is "directly connected" or "directly connected" to another component, it should be understood that there are no other components in between.

[0026] The terms used in this application are used merely to describe specific embodiments and are not intended to limit the invention. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this application, terms such as "comprising" or "having" are intended to specify the presence 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.

[0027] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this application.

[0028] Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the attached drawings.

[0030] FIG. 1 is a schematic diagram showing an environment in which an operation method of an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment is performed.

[0031] Referring to FIG. 1, a method of operation of an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment can be performed at an operating server (10) and a user terminal (20).

[0032] The operating server (10) may refer to a server device that performs sentiment analysis to support the execution of a user's routine, or a computing device capable of performing the same or similar functions. More specifically, the operating server (10) may receive target data, schedule data, execution result data, and review data from a user terminal (20), and after analyzing each of the received data to calculate the sentiment state and routine execution characteristics, it may generate customized feedback through an artificial intelligence-based feedback generation model. The operating server (10) may provide the customized feedback generated for the user to the user terminal (20) so that the user can utilize it immediately during the execution of their routine.

[0033] Meanwhile, the user terminal (20) may be a user's smartphone, tablet, smartwatch, laptop, or similar mobile terminal device, and may perform data transmission and reception in conjunction with the operation server (10). The user terminal (20) may provide an interface for the user to input target data and may receive and display plan data and customized feedback provided by the operation server (10). Additionally, the user terminal (20) may transmit execution status data and review data reflecting the user's execution results to the operation server (10), and may operate to enable the operation server (10) to continuously perform analysis of the emotional state and analysis of routine execution characteristics.

[0034] The operating server (10) and the user terminal (20) can be connected via a network (e.g., the Internet, a wireless communication network, a mobile communication network, etc.) and can transmit and receive data to and from each other as shown in FIG. 1. For example, the user terminal (20) can transmit target data to the operating server (10) in the form of text input, voice input, selectable options, numeric input, etc., and the operating server (10) can analyze the received target data to generate routine data, and then convert the routine data into planning data linked to the user's schedule data and transmit it to the user terminal (20).

[0035] Additionally, the user terminal (20) can transmit to the operation server (10) performance status data, such as whether the item is performed, the degree of progress, the time taken, and whether there is a delay, as well as review data, such as satisfaction scores, review text, and voice input, while performing each execution item according to the plan data. The operation server (10) can comprehensively analyze each data received or collected from the user terminal (20) to quantify the emotional state, evaluate routine execution characteristics, and provide customized feedback back to the user terminal (20) based on the results.

[0037] FIG. 2 is a diagram illustrating an exemplary configuration of hardware for an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment.

[0038] Referring to FIG. 2, in one embodiment, an operating server (100) that provides sentiment analysis-based AI customized feedback to support the execution of a user routine may include at least one processor (110) and a memory (120) that stores instructions that instruct at least one processor (110) to perform at least one operation.

[0039] Here, at least one processor (110) may mean a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor on which methods according to embodiments of the present invention are performed.

[0040] The memory (120) may be composed of at least one of a volatile storage medium and a non-volatile storage medium. For example, the memory (120) may be one of read-only memory (ROM) and random access memory (RAM).

[0041] An operating server (100) that provides AI-customized feedback based on sentiment analysis to support the execution of a user's routine may further include a storage device (160) for storing input data, intermediate processing temporary data, output data, etc. for performing at least one operation. For example, the storage device (160) may be a flash memory, a hard disk drive (HDD), a solid-state drive (SSD), or various memory cards (e.g., a micro SD card).

[0042] Additionally, the operating server (100) that provides AI-customized feedback based on sentiment analysis to support the execution of a user's routine may include a transceiver (130) that performs communication via a wireless network. Additionally, the operating server (100) that provides AI-customized feedback based on sentiment analysis to support the execution of a user's routine may further include an input interface device (140), an output interface device (150), a storage device (160), etc. Each component included in the operating server (100) that provides AI-customized feedback based on sentiment analysis to support the execution of a user's routine may be connected by a bus (170) to communicate with each other.

[0043] Examples of operating servers (100) that provide AI-customized feedback based on sentiment analysis to support the execution of a user's routine may include a communicable desktop computer, laptop computer, notebook, smartphone, tablet PC, mobile phone, smart watch, smart glass, e-book reader, PMP (portable multimedia player), portable game console, navigation device, digital camera, DMB (digital multimedia broadcasting) player, digital audio recorder, digital audio player, digital video recorder, digital video player, PDA (Personal Digital Assistant), etc.

[0045] FIG. 3 is a flowchart illustrating an operation method performed in an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment.

[0046] FIG. 4 is a flowchart illustrating a method for analyzing target data in an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment, and FIG. 5 is a flowchart illustrating a method for generating routine data in an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment.

[0047] FIG. 6 is a flowchart illustrating a method for generating plan data in an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment, and FIG. 7 is a flowchart illustrating a method for configuring result data in an operating server that provides sentiment analysis-based AI customized feedback to support the execution of a user's routine according to one embodiment.

[0048] FIG. 8 is a flowchart illustrating a method for determining an emotional state in an operating server that provides emotion-based AI customized feedback to support the execution of a user's routine according to one embodiment, and FIG. 9 is a flowchart illustrating a method for analyzing routine execution characteristics in an operating server that provides emotion-based AI customized feedback to support the execution of a user's routine according to one embodiment.

[0050] First, referring to FIG. 3, an operating server according to one embodiment can receive target data from a user terminal to support the execution of a user routine (S100).

[0051] At this time, the operating server can receive goal data related to the goal of the routine input from the user terminal through an interface provided by the operating server to the user terminal, and can analyze the received goal data. In this regard, the specific process of receiving and analyzing goal data from the operating server can be explained in more detail below with reference to FIG. 4.

[0052] Referring to FIG. 4, the operating server can receive target data having multiple input forms through an interface (S110).

[0053] Specifically, the operating server can receive the user's target data through an interface in the form of text input, optional input, numerical input, and voice input.

[0054] For example, text input may refer to text entered by a user on the interface of a user terminal, and when a sentence such as "walking 30 minutes every day for 3 months" is entered, the operating server can tokenize the entered text to separate it into multiple entities and identify each separated entity by semantic units such as time ("3 months"), frequency ("every day"), duration ("30 minutes"), and activity item ("walking"). Subsequently, each identified entity can be converted into structured items such as the goal achievement period, performance frequency, estimated duration, and goal type.

[0055] Additionally, the optional input may refer to a result selected by the user from among predefined items displayed on the interface of the user terminal, and for example, if options such as "Goal Type: Health Management", "Difficulty Level: Medium", and "Priority: High" are selected, the operation server can receive the selected result as goal data by mapping it to a predefined code value.

[0056] Numerical input may refer to a form in which numerical values ​​and units are entered from a user terminal, for example, when input such as "3 times a week," "45 minutes per session," or "for a total of 60 days," the operating server can extract the entered numerical values ​​and units and convert them into standard units (e.g., number of times performed per week, time taken per minute, total period per day) to receive them in a standardized form.

[0057] In addition, voice input may refer to content spoken by a user to a user terminal, and for example, when a speech such as "memorize 50 English words once a day for the next two months" is input, the operating server can convert the voice into text through a speech recognition module (e.g., may refer to an STT module) and then structure it into target data by performing tokenization and entity identification in the same way as text input.

[0058] Afterwards, the operation server can perform analysis on the target data received from the user terminal (S120).

[0059] Specifically, the operating server can classify each input value received as goal data into at least one of the following items: goal type, goal achievement period, execution frequency, preferred time interval, difficulty level, and priority.

[0060] First, in the goal data, the goal type item can be classified into one of the categories, such as health management, learning, work, or lifestyle management, based on activity content extracted from input text, multiple-choice options, or voice recognition results. That is, "walking for 30 minutes every day" can be classified as a health management type, and "memorizing 50 English words a day" can be classified as a learning type. Next, the goal achievement period item can be identified based on period-related data extracted from text or numerical input. For example, if input such as "3 months" is received, the operating server can convert this into 90 days to quantify the goal achievement period.

[0061] Additionally, the execution frequency item can be analyzed based on the repetitive execution conditions included in the target data. For example, if there is input such as "3 times a week" or "twice a day," the operation server can calculate the periodic frequency by separating this into the execution unit (e.g., week / day) and the number of times. The preferred time interval item can determine the execution time period preferred by the user by analyzing specific time zones entered from the user terminal (e.g., 6 AM to 8 AM, after 8 PM, etc.). In this case, the operation server can store the corresponding time zone in conjunction with schedule data to allow comparison with the actual feasibility of execution.

[0062] Meanwhile, the difficulty level item can be calculated based on the number of executions, time spent, and ratio relative to the target value, either by direct input from the user as a selectable option or by extracting numerical or text inputs. For example, the difficulty level of a goal such as "30 minutes every day for 3 months" can be classified by comparing the number of executions (90 times), total time spent (2,700 minutes), and target value relative to the period (daily baseline of 30 minutes) with a difficulty level table categorized by intervals for the number of executions, total time spent, and target value relative to the period. Additionally, the priority item can be analyzed as a factor that assigns relative importance within the goal data based on values ​​directly entered or selected by the user. For instance, if "Priority: High" is selected, that goal can be reflected in the planning data before other goals. In this way, the operational server can classify the input values ​​included in the received goal data by type and convert each item into a standardized indicator, structuring it for use in the subsequent routine data generation stage.

[0064] Referring again to FIG. 3, an operating server according to one embodiment can generate routine data for a user using the analysis results of target data (S200).

[0065] At this time, the operation server can analyze the target data based on the type of the target, the time available for execution, and the level of difficulty, and can generate routine data that includes multiple execution items for achieving the target and the order of each execution item based on the analysis results. In this regard, the specific process of generating routine data in the operation server can be explained in more detail below with reference to FIG. 5.

[0067] Referring to FIG. 5, the operation server can distinguish the type of user's goal based on the goal data (S210).

[0068] Specifically, the operation server can classify a user's goal into one of the following goal types: health management, learning, work, and lifestyle management, based on input values ​​classified by goal type items in the goal data. Here, the operation server can classify the user's goal type based on the goal type classified through the analysis of the goal data in step S120. In other words, the operation server can utilize the goal type classified in step S120.

[0069] Afterwards, the operation server can determine a time interval that can be allocated for the routine using the target data and the user's schedule data (S220).

[0070] Specifically, the operation server can acquire the user's schedule data through the user terminal and determine the time intervals available for the routine by comparing the schedule data with input values ​​classified as available time items in the target data. Here, the schedule data is data containing schedule information registered in a calendar or scheduling application managed by the user, and each schedule may include attributes such as "start time," "end time," "category (work, personal, appointment, etc.)," ​​"location," and "priority." The schedule data can be utilized as a criterion for determining the user's availability by time zone, and based on this, the operation server can distinguish between time intervals where the user can actually perform the routine and time intervals where they cannot.

[0071] Meanwhile, the input value categorized as the available time item may refer to the routine execution conditions specified by the user during the process of inputting target data. For example, a time slot preferred or available by the user, such as "weekdays 6 AM to 8 AM" or "weekends after 2 PM," may be received as the available time item. In this way, the operation server can determine the time interval that can be allocated for the routine based on the time interval in which the routine can actually be executed as determined through schedule data and the available time specified by the user.

[0072] Afterwards, the operation server can analyze the difficulty of the target data and determine the difficulty level of the user's routine (S230).

[0073] Specifically, the operation server can calculate a difficulty score based on pre-set criteria and weights for converting the number of executions, estimated time required, and ratio relative to the target value into difficulty levels, using input values ​​classified as difficulty level items in the target data, the number of executions required to achieve the goal, the estimated time required, and the ratio relative to the target value over the period. Subsequently, the operation server can determine the calculated difficulty score as one of the difficulty grades corresponding to a pre-set threshold range. For example, the operation server can extract the number of executions per week f, the estimated time required per execution d, the ratio of the target value relative to the period for achieving the goal r, and the ratio of available time relative to the weekly required time (which may mean time density) ρ from the target data as indicators for calculating the difficulty score. Subsequently, the operation server can normalize each extracted indicator in the manner shown in Equation 1 below.

[0074]

[0076] According to mathematical formula 1, x can represent each indicator, and x norm represents a normalized indicator, U represents a preset maximum value, and L represents a preset minimum value, and can be normalized to the range of 0 to 1 through Equation 1. Subsequently, the operation server can calculate a difficulty score D by combining each normalized indicator into a weighted sum as shown in Equation 2 below.

[0077]

[0079] In mathematical equation 2, f norm can mean the normalized number of executions per week, and d norm can refer to the normalized estimated time per iteration, and r norm can mean the ratio relative to the normalized target value, and ρ normrepresents the ratio of available time to normalized weekly required time. Additionally, in Equation 2, w1 to w4 may represent weights pre-set by the operation server for each indicator. Subsequently, the operation server can determine the difficulty grade G by comparing the difficulty score D, calculated in the manner of Equation 3 below, with the threshold interval L1 to L4, which is determined by a pre-set threshold (τ1 to τ3).

[0080]

[0082] In mathematical formula 3, L1 can mean "easy" among the difficulty grades, L2 can mean "normal" among the difficulty grades, L3 can mean "hard" among the difficulty grades, and L4 can mean "very hard" among the difficulty grades.

[0083] Afterwards, the operation server can derive candidate execution items for achieving the goal in relation to the user's routine (S240).

[0084] Specifically, the operations server can derive candidate execution items for achieving a goal based on the goal type, allocable time intervals, and difficulty grade. Here, an execution item can be defined as a unit of action that the user must actually perform, and each execution item may include the estimated time required for the action, the number of times it is performed, the necessary resources, and an indicator of contribution to goal achievement.

[0085] For example, an execution item can refer to a unit of action that the user must actually perform, and can be a specific action such as "30 minutes of walking," "25 minutes of study session," or "40 minutes of focused work block." Each execution item may include an estimated time required (τ), number of executions (n), required resources (R), and a contribution indicator (C) to goal achievement. Here, the contribution indicator (C) may refer to a numerical value representing the extent to which each execution item contributes to the overall goal achievement. For health management goals, the contribution may be calculated based on calorie expenditure, heart rate maintenance time, exercise intensity, etc.; for learning goals, the contribution may be calculated based on the amount of study, review rate, and number of problems solved. Additionally, for work goals, the contribution may be calculated based on the amount of work completed, quality of deliverables, and performance relative to focus time; and for lifestyle management goals, the contribution may be calculated based on the number of completed lifestyle improvement items and the repetition rate.

[0086] Meanwhile, the operational server may have candidate execution items for each goal type stored in the operational server's database in the form shown in Table 1 below.

[0087]

[0089] According to Table 1, the estimated time required may refer to the average time taken to complete the corresponding execution item, the number of executions may refer to the number of times repeated executions are required within a given period, the required resources may refer to the tools or environmental elements required to perform the execution item, and the contribution indicator may refer to a value quantified in the range of 0 to 1 regarding the relative contribution to achieving the goal, as previously explained.

[0090] Afterwards, the operation server can determine whether each candidate execution item is executable (S250).

[0091] Specifically, the operation server can calculate the estimated time required for each candidate execution item and determine whether each execution item is executable by comparing the calculated estimated time with allocable time intervals. For example, if the estimated time required for a specific execution item is "45 minutes," the operation server may determine that the execution item is "unexecutable" if only a continuous 30-minute time interval exists in the user's schedule data. On the other hand, the operation server may determine that the execution item is "executable" if there is a time interval of "45 minutes" or longer among the continuous time intervals found in the user's schedule data.

[0092] Afterwards, the operation server can generate routine data associated with each execution item for the user's routine (S260).

[0093] Specifically, the operation server can determine the execution order of each execution item based on execution feasibility, difficulty level, and the required execution frequency within the goal achievement period, and can generate routine data that includes each determined execution item and the execution order of each execution item.

[0094] For example, the operation server may first determine execution items that are judged to be "executable" as targets for routine data generation. Subsequently, the operation server may perform a first sorting of the determined execution items based on a difficulty grade (D). At this time, execution items assigned a high difficulty grade may be placed ahead of execution items assigned a low difficulty grade. This may mean placing relatively difficult items during the initial time interval when the user's concentration and energy are high.

[0095] Subsequently, the operation server can perform a second adjustment sort based on the execution frequency (f) required within the goal achievement period for the first sorted execution items. Specifically, the operation server can distribute items with high execution frequency within the same routine so that the repetitive placement interval is constant, and can sequentially place items with low execution frequency into the remaining time slots. Subsequently, the operation server can verify whether there is a conflict by comparing the estimated time required (T) of each execution item with the user's available time interval, and if it is determined that there is a conflict based on the verification result, it can reallocate the time slots.

[0096] For example, if a conflict occurs, the operations server can first check other available time intervals to search for an alternative time slot where the execution item can be placed. Subsequently, if no alternative time slot exists, the operations server can determine whether to accommodate the execution item by extending adjacent time intervals in the user's schedule. If the operations server determines that the execution item cannot be accommodated, it can handle the conflict with the user's available time intervals regarding the estimated time required (T) by carrying over the lower-priority execution item to a later time interval.

[0098] Referring again to FIG. 3, an operating server according to one embodiment can generate planning data using routine data and provide it to a user terminal (S300).

[0099] At this time, the operation server can generate planning data in which the execution items of the routine are placed within the user's available time interval by linking the generated routine data with the user's schedule data, and can provide the planning data to the user terminal. In this regard, the specific process of generating planning data from the operation server and providing it to the user terminal can be explained in more detail below with reference to FIG. 6.

[0101] Referring to FIG. 6, the operation server can generate planning data for routine execution using the user's schedule data (S310).

[0102] Specifically, the operation server can request and obtain the user's schedule data from the user terminal and compare each execution item of the routine data with the available time intervals included in the user's schedule data. Subsequently, based on the comparison results, the operation server can generate planning data by placing each execution item into a time slot within the available time interval where there is no conflict. At this time, the planning data may include the time slots assigned to each execution item, the estimated duration of each execution item, and the execution completion time. The operation server can integrate this data for each execution item to construct the final planning data, and the generated planning data can be output in a form that can be visually displayed on the user terminal.

[0103] Afterwards, the operation server can provide the plan data to the user terminal by transmitting the generated plan data to the user terminal (S320).

[0104] Specifically, the operation server may include time slots, estimated duration, and execution completion times allocated for each execution item in the planning data, and transmit the planning data containing each item to a user terminal so that it can be visually displayed. For example, the operation server may format and transmit the time slots, estimated duration, and execution completion times for each execution item included in the planning data so that they are visually displayed on the screen of the user terminal. Accordingly, the schedule and detailed execution conditions for each execution item included in the planning data can be displayed on the user terminal's interface, and the user can intuitively understand and perform routine execution based on the planning data.

[0106] Referring again to FIG. 3, an operation server according to one embodiment can receive result data for execution items included in the plan data (S400).

[0107] At this time, the operation server may receive result data from the user terminal, including execution status data related to the execution status of each execution item included in the plan data and review data related to the user's review following the execution of each execution item. In this regard, although it has been described that the operation server receives the result data, it is not necessarily limited thereto, and the result data may be constructed based on each of the received data after receiving the execution status data and review data, and this may be explained in more detail below with reference to FIG. 7.

[0109] Referring to FIG. 7, the operation server can receive execution status data for each execution item from the user terminal (S410).

[0110] Specifically, the operation server can receive execution status data from the user terminal, including whether each execution item included in the plan data is executed, the degree of progress expressed as a preset progress unit value, the time required, and whether there is a delay.

[0111] For example, the execution status may refer to an indicator showing whether an execution item has actually been initiated or not performed at all, and the user may input the initiated or unexecuted status through an execution indicator element of an interface provided by the operation server, and the input value may be transmitted to the operation server. Additionally, the progress level may refer to an indicator showing the degree to which an execution item has been implemented, and may be input as a pre-set progress unit value (e.g., 10% unit or checkpoint unit, etc.). In this case, the user may select a progress unit value through a progress input element implemented such as a slider or checkbox via a user terminal, and the operation server may receive the progress unit value selected by the user via the user terminal.

[0112] Meanwhile, elapsed time refers to the time used or spent on the actual execution of an execution item. If a user records the start and end times of execution through start and end inputs on an interface provided via a user terminal, the operation server may receive the elapsed time calculated as the difference between the two times; alternatively, if the user directly inputs the elapsed time in the form of minutes or seconds, the server may receive that value. Furthermore, the presence of a delay can be determined by comparing the timing of the start time of the execution item included in the plan data with the actual start time recorded by the user. If the actual start occurs after the planned start, it is judged as a delay, and the delay status may be determined as "True." In this case, the operation server may receive execution status, progress level, elapsed time, and delay status combined into a single execution status data.

[0114] Afterwards, the operating server can receive review data related to reviews for each execution item from the user terminal (S420).

[0115] Specifically, the operating server can receive review data resulting from the execution of each execution item, which includes review content containing at least one of a satisfaction score entered through the user terminal, text input related to the review, and voice input related to the review.

[0116] For example, a satisfaction score may refer to an indicator that numerically expresses the level of satisfaction a user feels regarding the results of each execution item through a user terminal, and when a user selects a score through a score input element of the interface provided by the operating server (e.g., a selectable input in the range of 1 to 5 points or 1 to 10 points), the selected value may be transmitted to the operating server. Additionally, text input may refer to content freely written by a user through the user terminal regarding the perceived difficulty, challenges encountered, suggestions for improvement, and willingness for the next execution during the process of executing each execution item. Text input may be transmitted to the operating server along with an execution item identifier (which may refer to an identifier used to distinguish execution items).

[0117] Meanwhile, voice input may refer to data submitted by a user that records a review spoken through the user. A voice file generated through the recording function of the user terminal may be transmitted to the operating server, or the result of converting the voice to text from the user terminal may be transmitted together. In this case, the operating server may receive at least one of the satisfaction score, text input, and voice input included in the review data as the content of the review.

[0118] Afterwards, the operation server can configure result data for each execution item based on execution status data and follow-up data (S430).

[0119] Specifically, the operation server can organize the received execution status data and review data into result data by associating them with the identifier of each execution item and a timestamp indicating the time information when each data was received. For example, the operation server can distinguish which execution item the data belongs to by associating the received execution status data and review data with the unique identifier of each execution item. Additionally, by recording the time of receipt—indicating the time information when each data was received by the operation server—as a timestamp, the server can manage when the execution status data and review data for the same execution item were submitted, along with the time information. In this case, the result data may include both quantitative values, such as whether the execution item was performed, the degree of progress expressed as a preset unit value, the actual time taken, and the presence of delay, as well as qualitative values, such as satisfaction scores, text input, and voice input; these values ​​can be linked to the identifier and timestamp to form a single integrated data structure.

[0121] Referring again to FIG. 3, an operating server according to one embodiment can analyze result data to determine an emotional state related to the execution of a user's routine (S500).

[0122] At this time, the operation server analyzes the execution status data and review data among the received result data to extract emotion expression data from the review data and derive execution patterns from the execution status data. Subsequently, the operation server can determine the emotion state related to the execution of the user's routine based on the extracted emotion expression data and the derived execution patterns. In this regard, the specific process of determining the user's emotion state by the operation server can be explained in more detail below with reference to FIG. 8.

[0123] Referring to FIG. 8, the operating server can calculate an emotional expression value from the review data included in the user's result data (S510).

[0124] Specifically, the operating server can calculate the user's emotional expression values, emotional classification values, emotional intensity values, and arousal values ​​by identifying emotional keywords, matching emotional categories, and determining voice tone to extract emotional expression data from review data.

[0125] For example, the production server counts the number of positive keywords N in the text included in the review data in the same way as Mathematical Formula 4 below. pos and number of negative keywords N neg The polarity score Pol can be calculated by comparing it with a normalized keyword value KW in the range of 0 to 1. norm It can be converted to. In mathematical formula 4, may mean a pre-set positive constant so that the denominator does not become zero.

[0126]

[0128] Subsequently, the operation server can perform emotion category matching to calculate the probability of belonging to a specific category (e.g., joy, anger, sadness, etc.), and that probability is the emotion classification value E c It can be defined as. In addition, the operating server converts the satisfaction score entered by the user into a value S ranging from 0 to 1. norm and normalized keyword value KW norm The emotional intensity value E is weightedly combined. i It can be calculated in the same way as in Equation 5 below. In Equation 5, λ1 and λ2 may represent pre-set weights assigned to satisfaction scores and keyword values.

[0129]

[0131] Meanwhile, if the operating server includes voice input in the review data, the volume Loud norm Pitch Variation Rate PitchVar norm , Ignition rate normEach can be normalized to a range of 0 to 1, and the arousal value E is obtained by weighted combination thereof. a It can be calculated in the same way as in Equation 6 below. In Equation 6, μ1 to μ3 may represent pre-set weights for each speech feature.

[0132]

[0134] Subsequently, the operating server can calculate the final emotion expression value E by weightedly combining the emotion classification value Ec, the emotion intensity value Ei, and the arousal value Ea, as shown in Equation 7 below. In Equation 7, α, β, and γ may represent pre-set weights for the category, intensity, and arousal components.

[0135] Subsequently, the operation server can determine the execution pattern value through the execution status data included in the user's result data (S520).

[0136] Specifically, the operation server can calculate the delay rate, completion rate, average execution speed, and deviation from the plan for each execution item to determine execution patterns from the execution status data, and can determine a quantified execution pattern value by normalizing each calculated indicator into either an aggregate value per period unit or an aggregate value per recent predetermined number of times unit. For example, the operation server can determine the completion rate C from the execution status data. r , latency D r , average execution speed V norm Deviation from plan P d ...can be calculated. Subsequently, the operation server can normalize each metric to a range of 0 to 1 and assign weights to calculate the execution pattern value P as shown in Equation 7 below.

[0137]

[0139] In mathematical equation 7, C r can mean the completion rate (number of items performed / total number of items), and D rcan mean the delay rate (as the ratio of delay occurrence to the plan, which can be calculated as a value from 0 to 1), and V norm can mean the value obtained by normalizing the average execution speed, and P d represents the deviation from the plan (which may mean the value obtained by normalizing the absolute error of actual progress compared to the plan progress), and η1 to η4 may represent pre-set weights for each indicator.

[0140] Afterwards, the operation server can determine the user's emotional state based on the emotional expression value and the execution pattern value (S530).

[0141] Specifically, the operation server can calculate a sentiment index that quantifies the user's emotional state based on a weighted average of sentiment expression values ​​and execution pattern values, and can determine the emotional state associated with the user's routine execution as positive, neutral, or negative by comparing the calculated sentiment index with a predefined threshold range. For example, the operation server can calculate the sentiment index EI by weightedly combining the sentiment expression value E and the execution pattern value P in a manner similar to Equation 8 below. In Equation 8, EI may represent the sentiment index, and w e and w p may refer to pre-set weights for a weighted combination of emotion expression values ​​and execution pattern values.

[0142]

[0144] Using the sentiment index EI calculated as in Equation 8, the operating server uses a preset threshold, the positive reference threshold θ pos and neutral reference threshold θ neu The emotional state can be determined as one of positive, neutral, or negative by comparing it with the threshold intervals distinguished by . In other words, the operating server determines the emotional index EI based on the positive threshold θ posIf above, it can be judged as positive, and the sentiment index EI is the positive threshold θ pos Less than, and neutral reference threshold θ neu If above, it can be judged as neutral, and the sentiment index EI is the neutral threshold value θ neu If it is less than, it can be judged as negative.

[0145] Referring again to FIG. 3, an operating server according to one embodiment can analyze the user's routine execution characteristics based on the user's routine data (S600).

[0146] At this time, the operation server can analyze the user's routine execution characteristics by comparing the user's emotional state, routine data, and result data with the data of a user among multiple users registered on the operation server whose routine similarity with the user is greater than or equal to a preset standard similarity. In this regard, the specific process of analyzing the user's routine execution characteristics on the operation server can be explained in more detail below with reference to FIG. 9.

[0147] Referring to FIG. 9, the operation server can determine a set of similar users for a user based on the user's routine data (S610).

[0148] Specifically, the operation server can calculate routine similarity between a user's routine data and the routine data of multiple users registered on the operation server, and can determine users whose routine similarity is greater than or equal to a preset threshold similarity as a set of similar users. For example, the operation server can calculate and utilize an integrated similarity score Sim based on the similarity of execution item types, similarity of batch order, execution frequency, and average time taken.

[0149] In this case, the type match of execution items can refer to an indicator of how much the execution items included in the routines of two users overlap, and a higher value may be calculated as the number of common items increases. That is, if one user's routine is "walking, reading, meditation" and another user's routine is "walking, reading, exercise," "walking" and "reading" are considered common items, and a high match score may be calculated. Additionally, the similarity of arrangement order can indicate how similarly the same execution items are arranged in the two routines; for example, if both users have organized their routines in the order of "walking → reading → meditation," a high similarity score may be calculated, while a low value may be calculated if the order is significantly different.

[0150] In addition, the similarity of execution frequency can indicate how similarly the same execution item is repeated in the routines of two users. For example, if User A performs "walking" 5 times a week and User B performs "walking" 4 times a week, the difference in frequency is small, so a high similarity can be calculated. Furthermore, the similarity of average time spent indicates how similarly each execution item is performed in the two routines. For example, if one user performs "walking" for an average of 30 minutes and the other user performs it for an average of 28 minutes, a high similarity can be calculated; on the other hand, if one performs it for 10 minutes and the other for 60 minutes, a low similarity can be calculated.

[0151] The operating server can calculate an integrated similarity score by combining the type match of the execution items, the similarity of the placement order, the similarity of the execution frequency, and the similarity of the average time taken, as shown in Equation 9 below.

[0152]

[0154] In mathematical equation 9, Sim type can mean the similarity of execution item types, and Sim ordercan mean the similarity of the execution item order, and Sim freq can mean the similarity of execution frequency, and Sim time represents the similarity of the average time taken, ρ1 to ρ4 represent pre-set weights for each item, and the sum may be pre-set to 1. Subsequently, the operation server may determine a set of similar users as users whose routine similarity, which is the calculated similarity Sim, is greater than or equal to a pre-set reference similarity.

[0155] Afterwards, the operation server can calculate routine characteristic indicators for users based on a set of similar users (S620).

[0156] Specifically, the operating server can compare the result data and emotional state of each user included in a similar set of users with the result data and emotional state of the user, and based on the comparison result, can calculate at least one of the completion rate, delay rate and persistence index for the user's execution item, the distribution of positive, neutral, and negative emotional indices, and the success rate relative to the difficulty grade as a routine characteristic indicator.

[0157] First, the completion rate can be defined as the ratio of execution items actually completed out of the planned execution items. For example, if a total of 20 execution items are planned for a user and 16 of them are completed, the completion rate can be calculated as 80%. Subsequently, the calculated completion rate can be defined as an indicator for evaluating whether the user's performance fidelity is relatively high or low by comparing it with the average completion rate of users belonging to a similar user set.

[0158] Additionally, latency can be defined as the proportion of execution items that started later than their planned start time. For example, if five of the total execution items recorded an actual start time later than their planned start time, the latency can be calculated as 25% of the total. The operations server can compare the calculated latency with the average latency of a similar set of users to determine whether the user is superior or inferior to the average in terms of time compliance.

[0159] Additionally, the persistence index can be defined as an indicator of how consistently a routine is maintained throughout the entire target period. For example, if the target period is 30 days and an execution item is performed normally for 24 days, the persistence index can be calculated as 24 / 30 = 0.8. The operations server can determine whether the user's routine retention is relatively high by comparing the persistence index with the average persistence index of a similar set of users.

[0160] Additionally, the sentiment index distribution can be defined based on the proportions of positive, neutral, and negative emotions included in the review data. For example, if a total of 20 review data points are collected, with 12 positive emotions, 5 neutral emotions, and 3 negative emotions, the respective proportions can be calculated as 60%, 25%, and 15%, respectively. The operational server can analyze whether the user experienced relatively more positive emotions during the performance of the routine by comparing the sentiment index distribution with the average sentiment distribution of a similar set of users.

[0161] Additionally, the success rate relative to the difficulty level can be defined as the ratio of successfully completed items among execution items classified at a specific difficulty level. For example, if there were a total of 10 execution items of the high difficulty level and 7 of them were completed, the success rate relative to the difficulty level can be calculated as 70%. Subsequently, the operation server can compare this value with the average success rate of execution items corresponding to the same difficulty range within a set of similar users to determine whether the user is achieving accomplishments appropriately corresponding to their difficulty level.

[0162] Subsequently, the operation server can construct routine characteristic indicators that reflect the relative differences between the user and a set of similar users by comprehensively utilizing the calculated completion rate, latency rate, persistence index, sentiment index distribution, and success rate relative to the difficulty grade, and these routine characteristic indicators can subsequently be used as basic data to classify the routine execution characteristics of the user.

[0163] Here, the operation server can calculate the difference value for each metric based on the comparison results between the user and a set of similar users, and the operation server can determine the metric whose difference value is greater than or equal to a preset threshold difference standard as the routine characteristic metric. In this case, if multiple metrics are greater than or equal to the threshold difference standard, the operation server can adopt all multiple metrics as routine characteristic metric indicators; conversely, if the difference of all metrics is less than the threshold difference standard, the operation server can determine the metric with the largest difference value as the representative routine characteristic metric. In other words, for example, if the user's completion rate is calculated to be 85% and the average completion rate of the set of similar users is calculated to be 70%, resulting in a difference of 15%, the completion rate can be determined as the routine characteristic metric if that difference is greater than or equal to the threshold difference standard (e.g., 10%).

[0164] Afterwards, the operation server can classify the user's routine execution characteristics into at least one of a plurality of preset types based on routine characteristic indicators (S630).

[0165] Specifically, the operation server can compare routine characteristic indicators with type-specific threshold ranges that classify routine execution characteristics into multiple types, and based on the comparison results, classify the user's routine execution characteristics into at least one of the execution ability type, emotional response type, and difficulty suitability type. Here, the type-specific threshold range may refer to a pre-set numerical range for defining routine execution characteristic types such as execution ability, emotional response, and difficulty suitability.

[0166] For example, the type of action ability can be determined by indicators such as the completion rate and persistence index. If the completion rate meets a certain standard (e.g., 80% or higher) and the persistence index meets a standard value (e.g., 0.7 or higher), the user may be classified as a "high action ability" type. Conversely, if both the completion rate and the persistence index are lower than a certain standard, the user may be classified as a "low action ability" type.

[0167] Additionally, the type of emotional response can be determined by the distribution of the emotional index, and if the proportion of positive emotion exceeds a certain threshold of the total (e.g., 60% or more), it can be classified as a "positive response" type, and if the proportion of negative emotion exceeds a certain standard (e.g., 40% or more), it can be classified as a "negative response" type. If neutral emotion is relatively high, it can be classified as a "neutral response" type.

[0168] In addition, the difficulty suitability type can be defined by the success rate relative to the difficulty grade, and if the success rate of the higher difficulty range satisfies a preset standard (e.g., 70% or more), it can be classified as the "difficulty suitability" type, and conversely, if it is below the standard, it can be classified as the "difficulty mismatch" type.

[0169] Accordingly, the operation server can classify the user's routine execution characteristics into at least one of the execution ability type, emotional response type, and difficulty suitability type by comparing routine characteristic indicators with preset type-specific threshold intervals. In this process, if multiple indicators simultaneously satisfy different type conditions, the user may be classified into two or more types in an overlapping manner.

[0171] Referring again to FIG. 3, an operating server according to one embodiment can generate and provide customized feedback for a user based on the user's emotional state and routine execution characteristics (S700).

[0172] At this time, the operation server can generate customized feedback through a pre-trained AI-based feedback generation model that generates customized feedback for the user's routine execution based on emotional state and routine execution characteristics, and can provide the generated customized feedback to the user terminal. Specifically, the operation server can generate customized feedback for the user's routine execution by providing the emotional state and routine execution characteristics as input to the pre-trained AI-based feedback generation model, and can transmit and provide the customized feedback to the user terminal.

[0173] Here, the AI-based feedback generation model can be implemented as a natural language generation model based on a multilayer neural network configured to map routine data, result data, and sentiment index between inputs and outputs. In this case, the feedback generation model can be pre-trained and supervised fine-trained using a training dataset for multiple users accumulated in advance on an operating server (i.e., a training dataset including target data, routine data, result data, sentiment state labels and routine execution characteristics, and customized feedback for each user). During the training process, a learning condition can be set to map to a feedback candidate group that includes at least one of a motivational message, a guideline for improving execution methods, and a difficulty adjustment suggestion corresponding to the sentiment state and routine execution characteristics. Additionally, a goodness-of-fit signal can be calculated based on an indicator of the user's acceptance performance (e.g., the rate of initiation of the corresponding execution item within 48 hours, the change in the completion rate for the following week, etc.) and applied to a loss function.

[0174] Additionally, the AI-based feedback generation model may be a model trained to output customized feedback by composing the feedback with the highest preset suitability in the form of a natural language sentence from a pool of feedback candidates corresponding to the user's emotional state and routine execution characteristics input to the feedback generation model. Here, suitability can be calculated by combining predictions of subsequent performance signals (e.g., increased subsequent completion rate, reduced latency, improved sentiment index, etc.) from the user and a set of similar users with penalties for rule-based constraints (e.g., prevention of schedule conflicts, overload risk, duplicate messages, etc.). If the operational server determines that the suitability is below a preset threshold, it may regenerate customized feedback by selecting alternative candidates or automatically adjusting the tone, length, and action suggestions.

[0175] Additionally, the operation server may provide the generated customized feedback by transmitting it to the user terminal. In this case, the feedback may include a message body, a related execution item identifier and scheduled time slot, metadata for interaction elements such as an execution button, a rescheduled button, and a difficulty adjustment button, and an explanatory tag such as the reason for provision (e.g., "3 consecutive delays detected"). The metadata may be configured to lead to actions such as moving to a routine screen, accepting a schedule rescheduling proposal, or lowering the difficulty level when the user terminal selects or taps the customized feedback.

[0176] Additionally, the operational server may collect short-term fluctuations in user responses (e.g., button clicks, acceptance / rejection of a proposal, whether to proceed within 24 to 48 hours, changes in the completion rate for the following week, etc.) and sentiment indices as post-hoc logs after providing personalized feedback, and these can be used to update personalization rules to improve the timing, tone, and type of future provision, and to accumulate training samples used as data for periodic retraining of the feedback generation model. In this case, the operational server may apply only re-ranking and rule correction without changing model parameters during the online provision process, and parameter updates may be performed according to the server's offline retraining cycle.

[0178] The methods according to the present invention may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the computer-readable medium may be those specifically designed and configured for the present invention, or they may be those known and available to those skilled in the art of computer software.

[0179] Examples of computer-readable media may include hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions may include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The aforementioned hardware devices may be configured to operate as at least one software module to perform the operation of the present invention, and vice versa.

[0180] In addition, the above-described method or device may be implemented by combining all or part of its configuration or function, or by implementing it separately.

[0181] Although the present invention has been described above with reference to preferred embodiments, 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

[0182] 10: Operation Server 20: User Terminal 100: Operation Server 110: Processor 120: Memory 130: Transmitter / Receiver 140: Input interface device 150: Output interface device 160: Storage device 170: Bus

Claims

Claim 1 A method of operation for an operating server that provides AI-customized feedback based on sentiment analysis to support the execution of a user's routine, comprising: receiving goal data related to the goal of a routine input from the user's terminal through an interface provided by the operating server; analyzing the goal data based on the type of goal, available execution time, and difficulty level, and generating routine data including a plurality of execution items for achieving the goal and the order of each execution item based on the analysis results; linking the generated routine data with the user's schedule data to generate planning data in which the routine execution items are placed within the user's available time interval, and providing the planning data to the user terminal; receiving result data from the user terminal including execution status data related to the execution status of each execution item included in the planning data and review data related to the user's review following the execution of each execution item; analyzing the execution status data and the review data among the received result data to extract sentiment expression data from the review data and derive an execution pattern from the execution status data, and determining the sentiment state related to the execution of the user's routine based on the extracted sentiment expression data and execution pattern; and determining the user's sentiment state, routine data, and result data among a plurality of users registered with the operating server, routine with the user A step of analyzing the routine execution characteristics of said user by comparing the similarity with data of a user whose similarity is greater than or equal to a preset standard similarity; and a step of generating customized feedback through an AI-based feedback generation model pre-trained to generate customized feedback for the execution of said user's routine based on said emotional state and said routine execution characteristics, and providing the generated customized feedback to said user terminal;The step of receiving the goal data includes: receiving the user's goal data through the interface in the form of text input, selectable option input, numeric input, and voice input; and classifying each input value received as the goal data into at least one item among the type of goal, goal achievement period, execution frequency, preferred time interval, difficulty level, and priority. The method includes the step of generating routine data, wherein the step of classifying the user's goal into one of the goal types among health management, learning, work, and lifestyle management based on an input value classified as a goal type item in the goal data; the step of obtaining the user's schedule data through the user terminal and determining a time interval allottable for the routine by comparing the input value classified as an actionable time item in the goal data with the schedule data; the step of calculating a difficulty score based on a pre-set standard and weight for converting the number of executions, estimated time required, and ratio of the target value to the period, using an input value classified as a difficulty level item in the goal data, the number of executions required to achieve the goal, the estimated time required, and the ratio of the target value to the period, and determining the calculated difficulty score as one of the difficulty grades corresponding to a pre-set threshold interval; the step of deriving candidate execution items for achieving the goal based on the goal type, the allottable time interval, and the difficulty grade, wherein the execution item is defined as a unit of action that the user must actually perform, and each execution item includes the estimated time required for the corresponding action, the number of executions, the required resources, and an indicator of contribution to achieving the goal; the candidate execution A step of calculating the estimated time required for each item and determining whether each execution item is feasible by comparing the calculated estimated time required with the allocatable time interval;A method of operation of an operating server providing AI-customized feedback based on sentiment analysis, further comprising: a step of determining the execution order of each execution item based on the feasibility of execution, the difficulty level, and the required execution frequency within the goal achievement period, and generating routine data that includes each determined execution item and the execution order of each execution item; a step of providing the planning data to a user terminal; a step of comparing each execution item of the routine data with an available time interval included in the user's schedule data, and generating planning data such that each execution item is placed in a non-conflicting time slot within the available time interval based on the comparison result; and a step of including the time slot placed for each execution item, the estimated time required, and the execution completion time in the planning data, and transmitting the included planning data to the user terminal to provide it for visual display. Claim 2 delete Claim 3 ◈Claim 3 was abandoned upon payment of the registration fee.◈ In Claim 1, the step of receiving result data comprises: receiving from the user terminal execution status data including whether each execution item included in the planning data is performed, the degree of progress expressed as a preset progress unit value, the time required, and whether there is a delay; receiving from the user terminal review data regarding the execution of each execution item including review content comprising at least one of a satisfaction score entered through the user terminal, text input related to the review, and voice input related to the review; and configuring the received execution status data and review data into result data by linking them with an identifier of each execution item and a timestamp indicating time information when each data was received. The method further includes: a step of determining the emotional state, wherein the step of calculating the emotional classification value, emotional intensity value, and arousal value as the user's emotional expression value through emotional keyword identification, emotional category matching, and voice tone determination to extract emotional expression data from the review data; a step of calculating the delay rate, completion rate, average execution speed, and deviation from the plan for each execution item to determine the execution pattern in the execution status data, and determining a quantified execution pattern value by normalizing each calculated indicator into one of the aggregated values ​​per period unit and the aggregated values ​​per recent predetermined number unit; and a step of calculating an emotional index quantifying the user's emotional state based on the weighted average of the emotional expression value and the execution pattern value, and determining the emotional state related to the user's routine execution as one of positive, neutral, and negative by comparing the calculated emotional index with a predefined threshold interval; and the step of analyzing the user's routine execution characteristics further includes: a step of calculating the routine similarity between the user's routine data and the routine data of a plurality of users registered on the operation server, and determining users whose routine similarity is greater than or equal to a pre-set standard similarity as a set of similar users.A step of comparing the result data and emotional state of each user included in the aforementioned set of similar users with the result data and emotional state of said user, and based on the comparison result, calculating at least one of said user's completion rate, delay rate and persistence index, distribution of positive, neutral, and negative emotional indices and success rate relative to difficulty grade as a routine characteristic indicator; and a step of comparing said routine characteristic indicator with a type-specific threshold interval that classifies routine execution characteristics into multiple types, and based on the comparison result, classifying said user's routine execution characteristics into at least one of execution ability type, emotional response type, and difficulty suitability type; A method of operation of an operating server providing AI-based personalized feedback for sentiment analysis, comprising further including the step of providing the personalized feedback to the user terminal, wherein the emotional state and the routine execution characteristics are provided as inputs to a pre-trained AI-based feedback generation model to generate personalized feedback for the user's routine execution, and the personalized feedback is transmitted to and provided to the user terminal; wherein the AI-based feedback generation model is a multi-layer neural network-based natural language generation model configured to map the routine data, result data, and sentiment index between input and output, and is pre-trained using a learning data set containing target data, routine data, result data, emotional state labels, routine execution characteristics, and personalized feedback for each user accumulated in advance on the operating server, wherein during the learning process, a learning condition is applied to map to a feedback candidate group that includes at least one of a motivational message, execution method improvement guidelines, and difficulty adjustment suggestions corresponding to the emotional state and routine execution characteristics, and wherein the feedback with the highest pre-set suitability among the feedback candidate group corresponding to the user's emotional state and routine execution characteristics input to the feedback generation model is constructed in the form of a natural language sentence to output the personalized feedback.

Images