Method and apparatus for determining automated interactions with content

A hierarchical interaction decision structure using rule-based filters and AI semantic determination addresses challenges in conventional automation by improving content item selection and enabling iterative policy adjustments.

KR103043639B1Active Publication Date: 2026-11-11EN HANCE
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
KR1020260125949
Authority / Receiving Office
KR · KR
Patent Type
Patents
Current Assignee / Owner
Filing Date
2026-07-09
Publication Date
2026-11-11
Estimated Expiration
2046-07-09

AI Technical Summary

Technical Problem

Conventional automation methods for interacting with content items face challenges such as determining contextual relevance, filtering inappropriate content, preventing monotonous interactions, managing interaction patterns, and lacking a structured decision history for improving interaction criteria.

Method used

A hierarchical interaction decision structure combining rule-based filters, AI semantic determination, and feedback-based decision history to select and store interaction targets and actions, adjusting policies iteratively based on stored decision history.

Benefits of technology

Improves content item selection quality by reducing unsuitable interactions, prevents monotonous patterns, and enables continuous policy adjustments using stored decision history.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure R1020260125949_ABST
    Figure R1020260125949_ABST
Patent Text Reader

Abstract

The present invention relates to a method and apparatus for automatically determining interaction with content targets. The apparatus for automatically determining interaction with content targets comprises: a collection unit that collects a plurality of candidate content items from a content source; a determination unit that selects interaction target candidates by applying rule-based judgment to the candidate content items to exclude some and using the results of artificial intelligence semantic judgments for the candidate content items that are not excluded; an action determination unit that determines one or more actions among a plurality of action types for the selected interaction target candidates; a history storage unit that stores a decision history including a judgment result, a reason for judgment, and a determined action; and a policy adjustment unit that aggregates the decision history to generate an adjustment signal for a policy applied to one or more of the selection and the decision of the action. Accordingly, the quality and potential for improvement of the interaction determination can be enhanced.
Need to check novelty before this filing date? Find Prior Art

Description

Technology Field

[0001] The present invention relates to a technology for automatically determining interactions with content targets, and more specifically, to a method and apparatus for determining the target and action of an interaction for candidate content items collected from one or more content sources, and storing the determination process as a decision history to adjust subsequent interaction policies. Background Technology

[0003] Automation involving external content through accounts can be utilized to enhance awareness, expertise, and trust by appropriately participating in relevant conversations. However, conventional automation methods based on search terms or single rules have the following limitations.

[0004] First, even if the search term matches, interaction may occur with content items unrelated to the operational interests. This is because it is difficult to determine the actual meaning or contextual relevance of a content item based solely on whether the search term matches.

[0005] Second, users may interact with promotional content, financial content, specific topics prohibited by operational policies, or low-quality, engagement-inducing content. Simple rules alone are insufficient to adequately filter out such inappropriate content in advance.

[0006] Third, repeated contact with the same author or content items may lead to a negative perception of the operating account.

[0007] Fourth, interaction patterns can become monotonous by performing the same action for all candidates, such as always responding. This becomes a factor that causes automated operations to be perceived mechanically.

[0008] Fifth, only the results of the finally performed interactions are recorded, making it difficult to explain retrospectively why a specific candidate was excluded or selected. Consequently, it is difficult to secure grounds for improving interaction criteria.

[0009] On the other hand, while generative AI is useful for determining the meaning of content items, relying solely on AI to determine all candidates increases judgment costs and leads to the problem of depending on the model's judgment even for items that are clearly prohibited by policy and should be excluded.

[0010] Furthermore, automated interaction decision systems need to manage not only which candidates were selected in a single execution, but also which candidates were excluded for what reasons and how those reasons are reflected in subsequent interaction criteria. Since conventional automation methods only store the results of performed interactions, it is difficult to utilize information regarding excluded candidates for adjusting subsequent policies.

[0011] Therefore, a hierarchical interaction decision structure is required that organically combines low-cost policy filters, AI semantic determination, author impact verification, and feedback based on action selection and decision history. Prior art literature

[0013] Korean Registered Patent No. 10-2953166 (April 13, 2026), System for providing user-customized SNS marketing information based on AI-based automatically generated content and method of operation thereof The problem to be solved

[0014] The objective of the present invention is to provide a method and apparatus for automatically determining content target interactions by hierarchically combining a low-cost policy filter and an artificial intelligence semantic judgment to remove unsuitable candidate content items prior to AI judgment, and to determine the target of interaction and the action for the passed candidates.

[0015] Another objective of the present invention is to provide a method and apparatus for automatically determining interaction with content, which stores whether candidate content items are passed or excluded, the reason therefor, selected actions, and execution results as a decision history, and aggregates the stored decision history to adjust one or more of policy filters, artificial intelligence semantic judgments, and action decision criteria.

[0016] However, the technical problems that the present invention aims to solve are not limited to those described above, and other unmentioned problems can be clearly understood by those skilled in the art from the description of the invention below. means of solving the problem

[0018] According to one embodiment of the present invention to achieve the above objective, an automatic interaction determination device for content targets is disclosed, comprising: a collection unit that collects a plurality of candidate content items from a content source; a determination unit that applies a rule-based determination to the candidate content items to exclude some of the candidate content items, obtains an artificial intelligence semantic determination result for the candidate content items not excluded by the rule-based determination, and selects interaction target candidates using the artificial intelligence semantic determination result; an action determination unit that determines one or more actions among a plurality of action types for the selected interaction target candidates; a history storage unit that stores a decision history including the result and reason for the rule-based determination, the artificial intelligence semantic determination result, and the determined action; and a policy adjustment unit that aggregates the decision history stored in the history storage unit and generates an adjustment signal for a policy applied to one or more of the selection by the determination unit and the action determination by the action determination unit.

[0019] In one embodiment, the history storage unit may store the decision history, which further includes the reason for exclusion for the candidate content item excluded by the rule-based judgment, the judgment of the judgment unit, and the version of the policy applied to the action decision of the action decision unit.

[0020] In one embodiment, the policy adjustment unit may generate an adjustment signal that changes the exclusion criteria corresponding to the exclusion reason so that the range of candidate content items excluded by the rule-based judgment is reduced when the frequency of occurrence of the exclusion reason is above a standard and the human approval result for the excluded candidate content item differs from the rule-based judgment.

[0021] In one embodiment, the policy adjustment unit may calculate an adjustment score by combining at least one of the occurrence frequency of exclusion reasons within an analysis window, the reversal rate between human approval results and automatic judgment results, and execution results by behavior type, and generate the adjustment signal if the adjustment score is greater than or equal to a standard.

[0022] In one embodiment, the action decision unit determines the action among writing a response, displaying a reaction, author relationship action, requesting approval, and skipping, and may decide to perform displaying a reaction instead of writing a response based on at least one of an action budget and an action portfolio policy.

[0023] In one embodiment, the determination unit can convert each of the candidate content items into a common schema including one or more of a content identifier, an author identifier, a body text, a time of creation, a reaction indicator, and an author influence indicator.

[0024] In one embodiment, the determination unit determines duplicate contact conditions for the same content item or the same author based on past interaction history, and excludes candidate content items corresponding to the duplicate contact conditions from the interaction target candidates, or restricts some of the plurality of behavior types for candidate content items corresponding to the duplicate contact conditions and allows the behavior to be determined from among the remaining behavior types.

[0025] In one embodiment, the judgment unit verifies at least one of the author influence indicator and the reaction indicator for candidate content items not excluded by the rule-based judgment, calculates a target score and a risk score based on the verification result and the artificial intelligence semantic judgment result, and can select the interaction target candidates using the target score and the risk score.

[0026] In one embodiment, the content target automatic interaction determination device may further include an execution unit that transmits the action determined by the action determination unit to a human approval queue if the action corresponds to a risk criterion, transmits the action to a delay queue if the action corresponds to an uncertainty criterion, and executes the action on a target content source if the action does not correspond to either the risk criterion or the uncertainty criterion.

[0027] In one embodiment, the content target automatic interaction determination device further includes an execution unit that executes the action determined by the action determination unit, and the history storage unit may store the decision history that further includes the result of executing the action by the execution unit.

[0028] According to one embodiment of the present invention, to achieve the above objective, a method for automatically determining interaction with content targets is disclosed, comprising: a step of collecting a plurality of candidate content items from a content source; a step of excluding some of the candidate content items by applying a rule-based judgment to the candidate content items; a step of obtaining an artificial intelligence semantic judgment result for the candidate content items not excluded by the rule-based judgment and selecting interaction target candidates using the artificial intelligence semantic judgment result; a step of determining at least one action among a plurality of action types for the selected interaction target candidates; a step of storing a decision history including the result and reason for the rule-based judgment, the artificial intelligence semantic judgment result, and the determined action; and a step of aggregating the stored decision history to generate a adjustment signal for a policy applied to one or more of the selection of interaction target candidates and the determination of the action.

[0029] In one embodiment, the step of storing the decision history may include storing the decision history, which further includes the reason for exclusion for the candidate content item excluded by the rule-based judgment, and the version of the policy applied to the selection and the decision of the action.

[0030] In one embodiment, the step of generating the adjustment signal may include the step of generating the adjustment signal to change the exclusion criteria corresponding to the exclusion reason so that the range of candidate content items excluded by the rule-based judgment is reduced when the frequency of occurrence of the exclusion reason is greater than or equal to a standard and the human approval result for the excluded candidate content item differs from the rule-based judgment.

[0031] In one embodiment, the step of generating the adjustment signal may include calculating an adjustment score by combining at least one of the occurrence frequency of exclusion reasons within an analysis window, the reversal rate between human approval results and automatic judgment results, and execution results by behavior type, and generating the adjustment signal when the adjustment score is greater than or equal to a standard.

[0032] In one embodiment, the step of determining the action may include determining the action among writing a response, displaying a reaction, author relationship behavior, requesting approval, and skipping, and determining to perform displaying a reaction instead of writing a response based on at least one of an action budget and an action portfolio policy.

[0033] In one embodiment, the automatic interaction determination method for content targets may further include the step of converting each of the candidate content items into a common schema including one or more of a content identifier, an author identifier, a body text, a time of creation, a reaction indicator, and an author influence indicator.

[0034] In one embodiment, the method for automatically determining interaction with content targets further includes a step of determining duplicate contact conditions for the same content item or the same author based on past interaction history, and the step of selecting interaction target candidates may include a step of excluding candidate content items corresponding to the duplicate contact conditions from the interaction target candidates, or restricting some of the plurality of behavior types for candidate content items corresponding to the duplicate contact conditions and determining the behavior from among the remaining behavior types.

[0035] In one embodiment, the step of selecting the interaction target candidates may include: verifying at least one of the author influence indicator and the reaction indicator for candidate content items not excluded by the rule-based judgment; calculating a target score and a risk score based on the verification result and the artificial intelligence semantic judgment result; and selecting the interaction target candidates using the target score and the risk score.

[0036] In one embodiment, the automatic interaction determination method for content targets may further include: a step of transmitting the determined action to a human approval queue if the action corresponds to a risk criterion; a step of transmitting the action to a delay queue if the action corresponds to an uncertainty criterion; and a step of executing the action on a target content source if the action does not correspond to either the risk criterion or the uncertainty criterion.

[0037] In one embodiment, the method for determining automatic interaction with content targets further includes the step of executing the determined action, and the step of storing the decision history may include the step of storing the decision history including the result of executing the action. Effects of the invention

[0039] According to one embodiment of the present invention, by first applying rule-based judgment to candidate content items to exclude some of them and obtaining AI semantic judgment results only for the candidate content items that were not excluded, it is possible to determine the topical relevance of content items that are difficult to judge by keyword matching alone, while reducing the consumption of computational resources for AI semantic judgment. Furthermore, by removing unsuitable content items that should be clearly excluded at the rule stage prior to AI judgment, interaction with unsuitable content items can be reduced and the quality of content item selection can be improved.

[0040] According to one embodiment of the present invention, the selection of interaction targets and the determination of behavior types are combined in a single decision process, and by determining one or more of a plurality of behavior types based on a behavior budget and a behavior portfolio policy, it is possible to prevent the interaction pattern from becoming monotonous due to the same behavior being performed for all candidate content items. Furthermore, by storing as a decision history which candidate content item was passed or excluded at which stage and for what reason, and which behavior was selected for what reason, it is possible to explain retrospectively not only the selection of interaction targets but also the reasons for selecting behavior types.

[0041] According to one embodiment of the present invention, a decision history including the exclusion reason for excluded candidate content items and the version of the applied policy is aggregated, and an adjustment signal is generated to adjust one or more of the exclusion criteria for rule-based judgment, AI semantic judgment, and policies applied to action decision based on an adjustment score calculated therefrom, thereby enabling the use of information regarding excluded candidate content items as well as the results of the performed interaction for the adjustment of subsequent policies. Accordingly, an interaction decision structure in which the criteria for selecting interaction targets and determining actions are continuously corrected through iterative execution, rather than being limited to one-time automatic execution, can be provided in a manner independent of a specific platform.

[0042] However, the effects obtainable through the present invention are not limited to those described above, and other unmentioned technical effects will be clearly understood by those skilled in the art from the description of the invention below. Brief explanation of the drawing

[0044] FIG. 1 is a block diagram showing the configuration of a content target automatic interaction determination device related to one embodiment of the present invention. FIG. 2 is a flowchart illustrating a method for determining automatic interaction with content targets related to an embodiment of the present invention. FIG. 3 is a flowchart illustrating the determination and action decision process of candidate content items related to one embodiment of the present invention. FIG. 4 is a diagram showing the formulas for calculating the target score and risk score related to one embodiment of the present invention. FIG. 5 is a diagram showing the calculation formula of an adjustment signal related to one embodiment of the present invention. FIG. 6 is a diagram showing the configuration of a decision history record related to one embodiment of the present invention. FIG. 7 is a diagram showing the hardware configuration of a content target automatic interaction determination device related to one embodiment of the present invention. Specific details for implementing the invention

[0045] An embodiment of the present invention will be described in detail below with reference to the attached drawings. Prior to this, terms and words used in this specification and claims should not be interpreted as being limited to their ordinary or dictionary meanings. Instead, based on the principle that the inventor can appropriately define the concept of a term to best describe their invention, they should be interpreted in a meaning and concept consistent with the technical spirit of the present invention. Therefore, the embodiments described in this specification and the configurations illustrated in the drawings are merely partial embodiments of the present invention and do not represent all of the technical spirit of the present invention. It should be understood that various equivalents and modifications capable of replacing them may exist at the time of filing this application.

[0046] Additionally, as used herein, “comprise, include” and / or “comprising, including” specify the presence of the mentioned features, numbers, steps, actions, parts, elements, and / or groups thereof, and do not exclude the presence or addition of one or more other features, numbers, actions, parts, elements, and / or groups.

[0047] Additionally, to aid in understanding the invention, the attached drawings are not drawn to actual scale, and the dimensions of some components may be exaggerated. Furthermore, the same reference numerals may be assigned to identical components in different embodiments.

[0048] The statement that two subjects of comparison are 'identical' means that they are 'substantially identical.' Therefore, substantial identity may include deviations considered low in the industry, for example, deviations within 5%. Additionally, the statement that a parameter is uniform in a given area may mean that it is uniform from an average perspective.

[0049] Although terms such as "first," "second," etc., are used to describe various components, it goes without saying that these components are not limited by these terms. These terms are used merely to distinguish one component from another, and unless specifically stated otherwise, the first component may also be the second component.

[0050] Throughout the specification, unless specifically stated otherwise, each component may be singular or plural.

[0051] The fact that any configuration is placed on the "upper (or lower)" of a component or on the "upper (or lower)" of a component may mean not only that any configuration is placed in contact with the upper (or lower) surface of said component, but also that another configuration may be interposed between said component and any configuration placed on (or below) said component.

[0052] Furthermore, where one component is described as being "on," "connected to," or "coupled to" another component, it should be understood that while the components may be directly connected or coupled to each other, another component may be "interposed" between each component, or that each component may be "connected," "coupled," or "coupled" through another component.

[0053] As used herein, the term “and / or” includes any and all combinations of one or more of the associated listed items. Additionally, when describing embodiments of the invention, the use of “may” relates to “one or more embodiments of the invention.” Expressions such as “one or more” preceding a list of elements modify the entire list of elements and do not modify individual elements of the list.

[0054] Throughout the specification, "A and / or B" means A, B, or A and B unless specifically stated otherwise, and "C to D" means C or more and D or less, unless specifically stated otherwise.

[0055] When syntax such as "at least one of A, B, and C", "at least one of A, B, or C", "at least one selected from the group of A, B, and C", or "at least one selected from A, B, and C" is used to specify a list of elements A, B, and C, the syntax can refer to any suitable combination.

[0056] The term "use" may be considered synonymous with the term "utilize." As used herein, terms such as "substantially," "about," and similar terms are used as terms of approximation rather than degree, and are intended to account for the inherent variation of a measured or calculated value that would be recognized by a person skilled in the art.

[0057] In this specification, terms such as first, second, third, etc. may be used to describe various elements, components, regions, layers, and / or sections, but these elements, components, regions, layers, and / or sections should not be limited by these terms. These terms are used to distinguish one element, component, region, layer, or section from another element, component, region, layer, or section. Accordingly, the first element, component, region, layer, or section discussed below may be named the second element, component, region, layer, or section without departing from the teachings of the exemplary embodiments.

[0058] Spatial relative terms such as "beneath," "below," "lower," "above," and "upper" may be used in the specification for ease of description to explain the relationship between one element or feature and other element(s) or feature(s) as illustrated in the drawings. Spatially relative positions are to be understood as encompassing different orientations of the device in use or operation, in addition to the orientations depicted in the drawings. For example, if the device in the drawings is inverted, the other element is understood as "below" or "below," and the described element is understood as "above" or "upper" of the other element. Thus, the term "below" may encompass both the up and down directions.

[0059] The terms used in this specification are intended to describe embodiments of the invention and are not intended to limit the invention.

[0060] In this specification, a content item may refer to a unit of information created by an author and containing a body. A content item may include, but is not limited to, social media posts, and may include any unit of information accompanied by one or more of an author identifier, a body, and a time of creation, such as reviews, Q&A, comments, community posts, or messages on a collaboration stream.

[0061] In this specification, a content source may mean a service, system, or data providing means capable of providing one or more content items. A content source may include one or more of search results, streams, notifications, recommendation feeds, official interfaces, third-party data providers, or internal collection systems.

[0062] In this specification, an interaction may mean an act of engagement performed toward a content item or its author. An interaction may include one or more of writing a response, displaying a reaction, author relationship actions, requesting approval, and skipping.

[0063] In this specification, rule-based determination may mean determining a content item based on whether it meets pre-set conditions. Artificial intelligence semantic determination may mean determining the semantic relevance between the body of a content item and a topic of interest. Automatic determination may be understood to encompass determinations performed without human intervention, including the rule-based determination and the artificial intelligence semantic determination.

[0064] In this specification, the decision history may refer to data that records one or more of a judgment step, a judgment result, a reason for judgment, and a selected action regarding a content item in a structured manner.

[0065] In this specification, an analysis window may refer to an aggregation interval that serves as the unit of calculation for a adjustment signal. An analysis window may be defined as a time interval, as the number of processed content items, or as an interval during which a specific policy version is maintained.

[0066] FIG. 1 is a block diagram showing the configuration of a content target automatic interaction determination device related to one embodiment of the present invention.

[0067] Referring to FIG. 1, the content target automatic interaction determination device (100) may include a collection unit (110), a judgment unit (120), an action determination unit (130), an execution unit (140), a history storage unit (150), and a policy adjustment unit (160). The content target automatic interaction determination device (100) receives content items from a content source, determines the target and action of interaction for the received content items, and can adjust the policy applied to subsequent interactions based on the decision history regarding the determination. Each of the above components is distinguished according to function and does not necessarily mean physically separated components, and functions performed by one or more processors may be implemented corresponding to the above components.

[0068] The collection unit (110) can collect multiple candidate content items from a content source. In one embodiment, the collection unit (110) can collect candidate content items from a single content source and can also collect candidate content items from multiple different content sources. The collection unit (110) can collect candidate content items based on search terms, topics, or descriptions of interest, and can collect candidate content items using one or more of browser rendering results, official interfaces, third-party data providers, or internal collection systems.

[0069] In one embodiment, the collection unit (110) can convert each collected candidate content item into a common schema. The common schema may include one or more of a content identifier, an author identifier, a body text, a time of creation, a reaction indicator, and an author influence indicator. The common schema can represent content items of different formats for each content source in a structure independent of the content source, and accordingly, the content target automatic interaction determination device (100) can process candidate content items without being dependent on the format of a specific content source. Although the conversion to the common schema is described as being performed by the collection unit (110), in other embodiments it may be performed by the determination unit (120).

[0070] The judgment unit (120) can select candidates for interaction targets by applying multiple judgments to the candidate content items. Specifically, the judgment unit (120) can apply rule-based judgments to the candidate content items to exclude some of the candidate content items, obtain AI semantic judgment results for the candidate content items that were not excluded by the rule-based judgments, and select candidates for interaction targets using the AI ​​semantic judgment results. In obtaining the AI ​​semantic judgment results, the judgment unit (120) can generate the results by directly performing AI semantic judgments, or in other embodiments, it may obtain the results by receiving them from an external judgment means. By applying the rule-based judgments prior to the AI ​​semantic judgments, the number of candidate content items requiring the acquisition of AI semantic judgment results can be reduced. The detailed judgment process of the judgment unit (120) will be described later with reference to FIG. 3.

[0071] The action decision unit (130) may determine one or more actions among a plurality of action types for the selected interaction target candidates. The plurality of action types may include one or more of writing a response, displaying a reaction, author relationship behavior, requesting approval, and skipping. In one embodiment, the action decision unit (130) may determine the actions based on one or more of a target score, a risk score, an action budget, and an action portfolio policy, and may determine different actions for each candidate rather than determining the same action for all interaction target candidates.

[0072] The execution unit (140) can execute the action determined by the action decision unit (130). In one embodiment, the execution unit (140) may transfer the action to a human approval queue if the action corresponds to a risk criterion, transfer the action to a delay queue if the action corresponds to an uncertainty criterion, and execute the action on a target content source if the action does not correspond to either the risk criterion or the uncertainty criterion. The human approval queue may refer to a queue where actions to be executed after human confirmation are waiting, and the delay queue may refer to a queue where actions whose execution time is deferred are waiting.

[0073] The history storage unit (150) may store a decision history including the decision result and reason for the decision of the decision unit (120) and the action determined by the action decision unit (130). In one embodiment, the decision history stored in the history storage unit (150) may include one or more of the result and reason for the decision of the rule-based decision, the result of the artificial intelligence semantic decision, the determined action, the reason for exclusion of candidate content items excluded by the rule-based decision, the version of the policy applied to the decision of the decision unit (120) and the action decision of the action decision unit (130), and the result of execution of the action by the execution unit (140). The history storage unit (150) may be implemented as one or more of a JSONL file, a relational database, an event stream, or an immutable log storage. The decision history may be used as an input for adjusting subsequent policies by recording not only the result of the performed interaction but also information regarding the excluded candidate content items.

[0074] The policy adjustment unit (160) can aggregate the decision history stored in the history storage unit (150) to generate an adjustment signal for a policy applied to one or more of the selection of the judgment unit (120) and the action decision of the action decision unit (130). The adjustment signal may include an exclusion criterion for a policy exclusion filter, a prompt used for artificial intelligence semantic determination, a creator influence criterion, and an adjustment signal for one or more of the action portfolio policy. As illustrated in FIG. 1, the adjustment signal generated by the policy adjustment unit (160) may be provided to the judgment unit (120) and the action decision unit (130), and the execution of the execution unit (140) itself may not be included in the target of the adjustment signal. Accordingly, the content target automatic interaction decision device (100) may not be limited to one-time automatic execution, but may repeatedly adjust the policy applied to the selection of interaction targets and action decisions based on the stored decision history.

[0075] Meanwhile, the policy adjustment unit (160) may further include a policy version management unit or be linked with the policy version management unit. The policy version management unit may store one or more versions of the exclusion criteria for rule-based judgment, the prompt for artificial intelligence semantic judgment, the threshold, and the behavior portfolio policy, and accordingly, the correspondence between the decision history and the change in policy may be tracked.

[0076] FIG. 2 is a flowchart illustrating a method for determining automatic interaction with content targets related to an embodiment of the present invention.

[0077] The method illustrated in FIG. 2 can be performed by the content target automatic interaction determination device (100) described with reference to FIG. 1. For convenience of explanation, each step is described below as being performed by the content target automatic interaction determination device (100), but each step can be performed by the corresponding component described in FIG. 1.

[0078] Referring to FIG. 2, the content target automatic interaction determination device (100) can collect a plurality of candidate content items from a content source (S110).

[0079] The content target automatic interaction determination device (100) can collect candidate content items from one content source and can also collect candidate content items from multiple different content sources.

[0080] The content target automatic interaction determination device (100) can convert each of the collected candidate content items into a common schema including one or more of a content identifier, a writer identifier, a body, a time of creation, a reaction indicator, and a writer influence indicator.

[0081] The content target automatic interaction determination device (100) can exclude some of the candidate content items by applying rule-based determination to the candidate content items (S120).

[0082] The above rule-based determination can exclude candidate content items based on whether they meet pre-set exclusion conditions, and the exclusion conditions may include one or more of financial contexts, promotional phrases, phrases encouraging participation, non-target languages, translation indications, prohibited topics, content items older than a reference time, content items below a reference response indicator, and content items of the operating account itself. The content target automatic interaction determination device (100) can generate exclusion reasons corresponding to each exclusion as standardized codes.

[0083] The content target automatic interaction determination device (100) can obtain an artificial intelligence semantic determination result for candidate content items that are not excluded by the rule-based determination and can select interaction target candidates using the artificial intelligence semantic determination result (S130).

[0084] The above artificial intelligence semantic judgment result may include a judgment result regarding the semantic relevance between the body of a candidate content item and a topic of interest, and may take the form of structured data including whether action is possible and the reason for the judgment. As the above rule-based judgment is applied prior to the above artificial intelligence semantic judgment, the number of candidate content items requiring the acquisition of the artificial intelligence semantic judgment result may be reduced. The content target automatic interaction determination device (100) may directly generate and acquire the above artificial intelligence semantic judgment result, or may receive and acquire it from an external judgment means.

[0085] The content target automatic interaction determination device (100) can determine one or more of the behaviors among a plurality of behavior types for the selected interaction target candidates (S140).

[0086] The above multiple types of behavior may include one or more of writing a response, displaying a reaction, author relationship behavior, requesting approval, and skipping.

[0087] In one embodiment, the content target automatic interaction determination device (100) can determine the behavior based on one or more of a target score, a risk score, a behavior budget, and a behavior portfolio policy.

[0088] The content target automatic interaction determination device (100) can store a decision history including the result and reason for the rule-based judgment, the result of the artificial intelligence semantic judgment, and the determined action (S150).

[0089] In one embodiment, the decision history may further include one or more of the reasons for exclusion for candidate content items excluded by the rule-based judgment, the version of the policy applied to the screening and the decision of the action, and the result of the execution of the action.

[0090] The content target automatic interaction determination device (100) can aggregate the stored decision history to generate a coordination signal for a policy applied to one or more of the selection of the interaction target candidate and the decision of the action (S160).

[0091] The above adjustment signal may include one or more adjustment signals for the exclusion criteria of the rule-based judgment, the prompt used for the AI ​​semantic judgment, the author influence criteria, and the behavior portfolio policy. The above adjustment signal may be applied automatically or may be applied after human approval.

[0092] The adjustment signal generated in step S160 may be reflected in a policy applied to one or more of the rule-based judgment and the decision of the action subsequently performed. Accordingly, the content target automatic interaction decision device (100) can adjust the policy applied to the selection of interaction targets and the decision of action through iterative execution.

[0093] Meanwhile, the order of each step shown in FIG. 2 is exemplary, and within the scope that does not contradict the spirit of the present invention, the order of some steps may be changed, two or more steps may be merged, or any one step may be divided into multiple steps. However, the exclusion by rule-based judgment (S120) may be performed prior to the acquisition of the artificial intelligence semantic judgment result (S130).

[0094] FIG. 3 is a flowchart illustrating the process of determining candidate content items and deciding actions related to an embodiment of the present invention. The process illustrated in FIG. 3 represents a detailed embodiment corresponding to exclusion (S120) or decision of action (S140) by rule-based determination in FIG. 2, and can be performed by a content target automatic interaction determination device (100).

[0095] Referring to FIG. 3, the content target automatic interaction determination device (100) can collect candidate content items from a content source (S210).

[0096] The content target automatic interaction determination device (100) can collect candidate content items based on search terms, topics, or interest descriptions, and can convert the collected candidate content items into a common schema.

[0097] The content target automatic interaction determination device (100) can exclude some of the candidate content items by applying rule-based determination to the candidate content items (S220).

[0098] The content target automatic interaction determination device (100) can exclude candidate content items that should be clearly excluded from the subject of artificial intelligence semantic determination by excluding candidate content items that meet pre-set exclusion conditions prior to obtaining the artificial intelligence semantic determination result. The exclusion conditions may take the form of comparing a specific attribute value of a candidate content item with a threshold value, or may take the form of determining whether a candidate content item belongs to a pre-set exclusion target. For example, the exclusion conditions may include a condition determining whether the creation time of a candidate content item is older than a reference time or whether a reaction indicator falls short of a reference reaction indicator by comparison with a threshold value, and may include a condition determining whether the body of a candidate content item corresponds to a prohibited topic, promotional phrase, or participation-inducing phrase by comparison with a pre-set exclusion target. The content target automatic interaction determination device (100) can generate exclusion reasons corresponding to each exclusion as standardized codes.

[0099] The content target automatic interaction determination device (100) can determine duplicate contact conditions for the same content item or the same author based on past interaction history (S230).

[0100] The content target automatic interaction determination device (100) may exclude candidate content items corresponding to the duplicate contact condition, and may restrict some of the plurality of behavior types for candidate content items corresponding to the duplicate contact condition and determine the behavior from among the remaining behavior types. For example, if a response is written for the same author within a reference time, the content target automatic interaction determination device (100) may restrict the writing of a response for the corresponding candidate content item and allow the display of a response or monitoring to be performed.

[0101] The content target automatic interaction determination device (100) can obtain an artificial intelligence semantic determination result for candidate content items that are not excluded by the rule-based determination and the duplicate contact condition (S240).

[0102] The above artificial intelligence semantic determination result may take the form of structured data including whether an action is possible and the reason for the determination. The above artificial intelligence semantic determination may be performed by a determination means implemented by a generative model, a classification model, an embedding similarity model, or a combination thereof, and the content target automatic interaction determination device (100) may obtain the result from the said determination means.

[0103] The content target automatic interaction determination device (100) can verify one or more of the author influence indicator and the reaction indicator for the above-mentioned unexcluded candidate content items, and calculate a target score and a risk score based on the verification result and the above-mentioned artificial intelligence semantic determination result (S250).

[0104] The above author influence indicator may include one or more of the number of followers, number of subscribers, average number of reactions, authentication status, and community reputation score. The above target score may be calculated from one or more of the content item's recency, reaction indicator, author influence, semantic relevance, and recent contact interval, and the above risk score may be calculated from one or more of the proximity to prohibited topics, promotional nature, whether it is a sensitive topic, possibility of repeated contact, and possibility of policy violation. An example of the formulas for calculating the above target score and the above risk score will be described later with reference to FIG. 4.

[0105] The content target automatic interaction determination device (100) can select interaction target candidates using the target score and the risk score (S260).

[0106] The content target automatic interaction determination device (100) can calculate the priority of candidate content items based on the target score and the risk score, and select interaction target candidates according to the calculated priority.

[0107] The content target automatic interaction determination device (100) can determine one or more of the behaviors among a plurality of behavior types for the selected interaction target candidates (S270).

[0108] The content target automatic interaction determination device (100) can determine the action based on one or more of the target score, the risk score, the action budget, and the action portfolio policy. In one embodiment, the content target automatic interaction determination device (100) can determine that a reaction display is performed instead of writing a response based on one or more of the action budget and the action portfolio policy. Additionally, the content target automatic interaction determination device (100) can determine that a writer relationship action is performed based on the relationship action budget after the success of writing a response, and can ensure that not all writing responses leads to the same subsequent action.

[0109] The content target automatic interaction determination device (100) can execute the determined action (S280). The content target automatic interaction determination device (100) can transmit the action to a human approval queue if the action corresponds to a risk criterion (S281), transmit the action to a delay queue if the action corresponds to an uncertainty criterion (S282), and execute the action on the target content source if the action does not correspond to either the risk criterion or the uncertainty criterion (S283).

[0110] The content target automatic interaction determination device (100) can store one or more of the judgment result, reason for judgment, selected action, and execution result of each step as a decision history (S290).

[0111] The above decision history may include step-by-step pass or exclusion results, reasons for exclusion, or reasons for selection generated in steps S220 to S283, and the content target automatic interaction determination device (100) may record in the decision history which candidate content item was passed or excluded at which step and for what reason.

[0112] Meanwhile, the order of exclusion by rule-based judgment (S220), determination of duplicate contact conditions (S230), acquisition of artificial intelligence semantic judgment results (S240), and influence verification and score calculation (S250) as illustrated in FIG. 3 is exemplary, and the content target automatic interaction determination device (100) may change some of the above order considering cost or processing efficiency. However, the exclusion by rule-based judgment (S220) may be performed prior to the acquisition of artificial intelligence semantic judgment results (S240).

[0113] FIG. 4 is a diagram showing a formula for calculating a target score and a risk score related to an embodiment of the present invention. The formula shown in FIG. 4 is an example of a formula that can be used in the influence verification and score calculation (S250) of FIG. 3, and does not limit the content target automatic interaction determination device (100) to a specific formula.

[0114] The content target automatic interaction determination device (100) can calculate a target score for a candidate content item. The target score may be a value representing the priority that the candidate content item has as a target for interaction, and may be calculated by combining a plurality of elements related to the candidate content item. The plurality of elements may include one or more of the semantic appropriateness, reaction indicator, author influence, recency, policy risk, and contact frequency of the candidate content item.

[0115] In one embodiment, the target score can be calculated by the following formula.

[0116] target_score = relevance_score + normalized_engagement + author_reach_weight + recency_weight - policy_risk_penalty - contact_frequency_penalty

[0117] In the above calculation formula, relevance_score may be a value corresponding to the semantic relevance between the body of the candidate content item and the topic of interest, normalized_engagement may be a value obtained by normalizing the response metric of the candidate content item, and author_reach_weight may be a weight corresponding to the author's influence. recency_weight may be a weight corresponding to the recency of the candidate content item, policy_risk_penalty may be a penalty value corresponding to the policy risk of the candidate content item, and contact_frequency_penalty may be a penalty value corresponding to the frequency of contact with the same content item or the same author.

[0118] According to the above calculation formula, a target score for a candidate content item with one or more of semantic appropriateness, response indicators, author influence, and recency can be calculated as high, and a target score for a candidate content item with high policy risk or contact frequency can be calculated as low. Accordingly, the content target automatic interaction determination device (100) can assign a high priority to a candidate content item suitable as a target for interaction.

[0119] The content target automatic interaction determination device (100) can calculate a risk score for a candidate content item. The risk score may be a value representing the degree of risk associated with an interaction with the candidate content item, and may be calculated from one or more of proximity to a prohibited topic, promotional nature, whether it is a sensitive topic, possibility of repeated contact, and possibility of policy violation. In one embodiment, the risk score may be used as a deduction factor in calculating the target score, and may be calculated separately from the target score and used to determine the action described below.

[0120] Each element constituting the target score and the risk score may be assigned a weight, and the weight may be a fixed value or may be set differently depending on the campaign stage or type of behavior. Each element and the weight are exemplary, and the content target automatic interaction determination device (100) may calculate the target score and the risk score by excluding some of the elements or adding other elements.

[0121] In one embodiment, the elements constituting the target score and the risk score or the weights may be adjusted by an adjustment signal described below. Accordingly, the content target automatic interaction determination device (100) may adjust the calculation criteria for the target score and the risk score based on a stored decision history.

[0122] FIG. 5 is a diagram showing a formula for calculating a control signal related to an embodiment of the present invention. The formula shown in FIG. 5 is an example of a formula that can be used by a content target automatic interaction determination device (100) to aggregate a decision history and generate a control signal, but the content target automatic interaction determination device (100) is not limited to a specific formula.

[0123] The content target automatic interaction determination device (100) can set an analysis window. The analysis window is an aggregation interval that serves as a unit for calculating adjustment signals, and may correspond to any one of a certain period, a certain number of candidate content items, or a interval in which a specific policy version is maintained. The content target automatic interaction determination device (100) can aggregate the decision history within the analysis window to calculate one or more of the number of candidate content items, the step-by-step exclusion rate, the pass rate of AI semantic judgment, the conversion rate to human approval, the execution rate by behavior type, and the execution result by behavior type.

[0124] The content target automatic interaction determination device (100) can calculate an adjustment score based on the aggregate result. In one embodiment, one or more of the following may be used to calculate the adjustment score: the frequency of occurrence by exclusion reason code, the ratio of differences between automatic judgment and human approval results, performance after execution by behavior type, the pass rate of the author influence criterion or response criterion, and the semantic cluster of reasons for failure of artificial intelligence semantic judgment.

[0125] In one embodiment, the adjustment score can be calculated by the following formula.

[0126] adjustment_score = reason_frequency_weight + approval_reversal_weight + outcome_delta_weight - policy_risk_weight - sample_uncertainty_weight

[0127] In the above calculation formula, reason_frequency_weight may be a value corresponding to the frequency of occurrence for each exclusion reason, approval_reversal_weight may be a value corresponding to the ratio of differences between automatic judgment and human approval results, and outcome_delta_weight may be a value corresponding to the change in performance after execution for each behavior type. policy_risk_weight may be a deduction value corresponding to policy risk, and sample_uncertainty_weight may be a deduction value corresponding to the uncertainty arising from the small number of decision histories used for aggregation.

[0128] According to the above calculation formula, if one or more of the frequency of occurrence of specific exclusion reasons, the ratio of difference from human approval results, or changes in performance after execution are large, the adjustment score may be calculated high, and if the policy risk is high or the number of decision history records used for aggregation is small, the adjustment score may be calculated low. Accordingly, the content target automatic interaction determination device (100) can generate an adjustment signal when sufficient grounds are secured and the risk is low.

[0129] The content target automatic interaction determination device (100) can generate an adjustment signal when the adjustment score is above a threshold. In one embodiment, the content target automatic interaction determination device (100) can generate an adjustment signal that relaxes the exclusion criteria corresponding to the exclusion reason when the frequency of occurrence of a specific exclusion reason is above a threshold and the human approval result for the excluded candidate content item differs from the rule-based judgment. Here, relaxing the exclusion criteria may mean changing the exclusion criteria corresponding to the exclusion reason so that the range of candidate content items excluded by the rule-based judgment is reduced. In one embodiment, changing the exclusion criteria may include one or more of changing the threshold of the exclusion condition corresponding to the exclusion reason in a direction such that the candidate content item is not excluded, or removing a specific condition from the exclusion target corresponding to the exclusion reason. Additionally, the content target automatic interaction determination device (100) can generate an adjustment signal that strengthens the prompt used for the explanation or judgment of the topic of interest when the performance of the candidate content item that passed the artificial intelligence semantic judgment is low, and can generate an adjustment signal that raises the author influence standard when the performance of a specific author influence range is low. Furthermore, the content target automatic interaction determination device (100) can generate an adjustment signal that expands the range where automatic execution is allowed when a candidate content item moved to the human approval queue is repeatedly approved, and can generate an adjustment signal that strengthens the criteria requiring approval when a risk candidate content item causes a negative reaction after execution.

[0130] In one embodiment, the content target automatic interaction determination device (100) may store the adjustment signal as a policy change candidate including one or more of an applied target policy, a previous policy value, a proposed policy value, the number of decision history that served as the basis, an expected effect, and a risk level. The adjustment signal may be applied automatically or may be applied after human approval.

[0131] The elements constituting the adjustment score and the weights assigned to the elements are exemplary, and the content target automatic interaction determination device (100) may calculate the adjustment score by excluding some of the elements or adding other elements.

[0132] FIG. 6 is a diagram showing the configuration of a decision history record related to an embodiment of the present invention. The decision history record illustrated in FIG. 6 is an example of a decision history that a content target automatic interaction determination device (100) can store for one candidate content item, and the configuration of the decision history record is not limited to the illustrated form.

[0133] The content target automatic interaction determination device (100) can generate and store a decision history record for each candidate content item. The decision history record may include one or more of information for identifying a candidate content item, information for identifying a policy applied to the candidate content item, a step-by-step judgment result for the candidate content item, and a score calculated for the candidate content item.

[0134] Information for identifying the above candidate content items may include one or more of a content identifier and an author identifier. Referring to FIG. 6, the content identifier may be represented as "post_id" and the author identifier as "author_id".

[0135] Information for identifying the policy applied to the above candidate content item may include one or more versions of the policy applied to the rule-based decision and the policy applied to the AI ​​semantic decision. Referring to FIG. 6, the version of the policy applied to the rule-based decision may be expressed as "policy_version", and the version of the policy applied to the AI ​​semantic decision may be expressed as "prompt_version". By the information for identifying the policy, it can be determined under which policy the decision history record was created, and accordingly, the correspondence between the decision history and the policy change can be traced.

[0136] The above-mentioned step-by-step judgment results may include one or more of the stage name, judgment result, and judgment reason for each judgment stage in which the candidate content item has passed. Referring to FIG. 6, the above-mentioned step-by-step judgment results may be expressed as "stage_results" and may include the judgment result and judgment reason for each stage corresponding to rule-based judgment, the stage corresponding to judgment of duplicate contact conditions, the stage corresponding to AI semantic judgment, and the stage corresponding to decision of action. For example, the stage corresponding to AI semantic judgment may include a result indicating that interaction is possible and the judgment reason, and the stage corresponding to decision of action may include a selected action and the reason for selection.

[0137] The score calculated for the above candidate content item may include one or more of a target score and a risk score. Referring to FIG. 6, the target score may be expressed as "target_score" and the risk score as "risk_score". Additionally, the decision history record may include information indicating the finally selected action, and referring to FIG. 6, the selected action may be expressed as "selected_action".

[0138] In one embodiment, the decision history record may further include the execution result of an action selected for the candidate content item. The execution result may include one or more of the following: whether the execution of the action was successful, the reason for the failure of the execution, and a waiting state resulting from transfer to a human approval queue or a delay queue. As the execution result is included in the decision history record, the content target automatic interaction determination device (100) may use the execution result by action type to calculate a coordination signal.

[0139] The above decision history record may further include reasons for exclusion for candidate content items excluded by the above rule-based judgment. Accordingly, the content target automatic interaction determination device (100) may store information regarding candidate content items for which a selected action was executed, as well as information regarding candidate content items for which interaction was not performed due to exclusion, and may use the information regarding excluded candidate content items for adjusting subsequent policies.

[0140] The names and structures of each field of the decision history record illustrated in FIG. 6 are exemplary, and the content target automatic interaction determination device (100) may configure the decision history record by excluding some of the fields or adding other fields.

[0141] FIG. 7 is a diagram showing the hardware configuration of a content target automatic interaction determination device related to one embodiment of the present invention.

[0142] Referring to FIG. 7, the content target automatic interaction determination device (100) may be implemented as a computing device (900) including one or more processors (910) and memory (920) in hardware. The collection unit (110), judgment unit (120), action determination unit (130), execution unit (140), history storage unit (150), and policy adjustment unit (160) described above with reference to FIG. 1 are configurations distinguished by function and do not necessarily mean physically separated components, but may be implemented in correspondence with functions performed by one or more processors (910).

[0143] The processor (910) can perform the function of each of the above components by executing one or more instructions stored in memory (920). The processor (910) may include one or more of a central processing unit, a graphics processing unit, or equivalent computational means, and may be implemented as a single processor or as a plurality of processors that are interconnected. The plurality of processors may be physically placed in a single device or distributed across a plurality of devices connected via a network.

[0144] The memory (920) can store instructions executed by the processor (910). Additionally, the memory (920) can store one or more of candidate content items, candidate content items converted into a common schema, decision history, and policies. The memory (920) may include one or more of volatile memory and non-volatile memory, and the history storage unit (150) that stores the decision history may correspond to at least a part of the memory (920) or to a separate storage means connected to the processor (910).

[0145] In one embodiment, the computing device (900) may further include communication means. The communication means may be used to receive candidate content items from a content source, to execute a determined action on a target content source, or to receive an artificial intelligence semantic determination result from an external determination means.

[0146] The function of each component performed by the processor (910) may be implemented in the form of instructions and stored in a computer-readable recording medium. The computer-readable recording medium may store instructions that enable the function of each component to be performed when executed by one or more processors, and may include one or more of magnetic media, optical media, semiconductor memory, or equivalent storage means.

[0147] As described above, according to one embodiment of the present invention, by first applying rule-based judgment to candidate content items to exclude some of them and obtaining AI semantic judgment results only for the candidate content items that were not excluded, it is possible to determine the thematic relevance of content items that are difficult to judge by keyword matching alone, while reducing the consumption of computational resources for AI semantic judgment. Furthermore, by removing unsuitable content items that should be clearly excluded at the rule stage prior to AI judgment, interaction with unsuitable content items can be reduced and the quality of content item selection can be improved.

[0148] According to one embodiment of the present invention, the selection of interaction targets and the determination of behavior types are combined in a single decision process, and by determining one or more of a plurality of behavior types based on a behavior budget and a behavior portfolio policy, it is possible to prevent the interaction pattern from becoming monotonous due to the same behavior being performed for all candidate content items. Furthermore, by storing as a decision history which candidate content item was passed or excluded at which stage and for what reason, and which behavior was selected for what reason, it is possible to explain retrospectively not only the selection of interaction targets but also the reasons for selecting behavior types.

[0149] According to one embodiment of the present invention, a decision history including the exclusion reason for excluded candidate content items and the version of the applied policy is aggregated, and an adjustment signal is generated to adjust one or more of the exclusion criteria for rule-based judgment, AI semantic judgment, and policies applied to action decision based on an adjustment score calculated therefrom, thereby enabling the use of information regarding excluded candidate content items as well as the results of the performed interaction for the adjustment of subsequent policies. Accordingly, an interaction decision structure in which the criteria for selecting interaction targets and determining actions are continuously corrected through iterative execution, rather than being limited to one-time automatic execution, can be provided in a manner independent of a specific platform.

[0150] Meanwhile, the method according to the present invention is written as a program executable on a computer and can be implemented on various recording media such as magnetic storage media, optical reading media, and digital storage media.

[0151] Implementations of the various technologies described herein may be implemented in digital electronic circuits, computer hardware, firmware, software, or a combination thereof. Such implementations may be implemented as computer programs stored on a computer-readable storage medium to control the operation of a data processing device. Such implementations may be implemented as computer program products, i.e., information carriers, such as computer programs stored on a machine-readable storage device or a computer-readable non-transient storage medium, for processing by the operation of a data processing device, e.g., a programmable processor, a computer, or a plurality of computers, or for controlling such operation. Computer programs such as the computer program(s) described above may be written in any form of programming language including compiled or interpreted languages, and may be deployed in any form including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. Computer programs may be deployed to be processed on one computer or a plurality of computers at a single site, or distributed across a plurality of sites and interconnected by a communication network.

[0152] Processors suitable for processing computer programs include, for example, both general-purpose and special-purpose microprocessors, and any one or more processors of any type of digital computer. Generally, the processor will receive instructions and data from read-only memory or random access memory, or both. The elements of the computer may include at least one processor that executes instructions and one or more memory devices that store instructions and data. Generally, the computer may include one or more mass storage devices that store data, for example, magnetic, magneto-optical disks, or optical disks, or may be combined to receive data from these, transmit data to these, or both. Information carriers suitable for embodying computer program instructions and data include, for example, semiconductor memory devices, magnetic media such as hard disks, floppy disks, and magnetic tapes; optical media such as CD-ROMs (Compact Disk Read Only Memory) and DVDs (Digital Video Disks); magneto-optical media such as floptical disks; ROMs (Read Only Memory); RAMs (Random Access Memory); flash memory; EPROMs (Erasable Programmable ROM); EEPROMs (Electrically Erasable Programmable ROM); etc. Processors and memory may be supplemented by or included in special-purpose logic circuit organizations.

[0153] Although this specification contains details of a number of specific embodiments, they should not be understood as limiting the scope of any invention or claimables, but rather as descriptions of features that may be characteristic of a specific embodiment of a specific invention. Specific features described in this specification in the context of individual embodiments may be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented in multiple embodiments individually or in any appropriate sub-combination. Furthermore, while features may operate in a specific combination and be described as initially claimed, one or more features from the claimed combination may be excluded from the combination in some cases, and the claimed combination may be changed to a sub-combination or a variation of the sub-combination.

[0154] Likewise, although operations are depicted in the drawings in a specific order, this should not be understood as requiring that such operations be performed in that specific or sequential order depicted to obtain a desirable result, or that all depicted operations must be performed. In certain cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of the various device components of the above-described embodiments should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and devices can generally be integrated together into a single software product or packaged into multiple software products.

[0155] Meanwhile, the embodiments of the present invention disclosed in this specification and drawings are merely specific examples provided to aid understanding and are not intended to limit the scope of the present invention. It is obvious to those skilled in the art that other variations based on the technical concept of the present invention are possible in addition to the embodiments disclosed herein.

[0156] [Explanation of the symbol]

[0157] 100: Automatic interaction determination device for content targets

[0158] 110: Collection Department

[0159] 120: Judgment panel

[0160] 130: Action Decision Unit

[0161] 140: Execution section

[0162] 150: History Storage

[0163] 160: Policy Coordination Department

[0164] 900: Computing device

[0165] 910: Processor

[0166] 920: Memory

Claims

Claim 1 A content target automatic interaction determination device comprising: a collection unit that collects a plurality of candidate content items from a content source; a determination unit that applies a rule-based judgment to the candidate content items to exclude some of the candidate content items and obtains an artificial intelligence semantic judgment result for the candidate content items not excluded by the rule-based judgment, and selects interaction target candidates using the artificial intelligence semantic judgment result; an action determination unit that determines one or more actions among a plurality of action types for the selected interaction target candidates; a history storage unit that stores a decision history including the result and reason for the rule-based judgment, the artificial intelligence semantic judgment result, and the determined action; and a policy adjustment unit that aggregates the decision history stored in the history storage unit and generates an adjustment signal for a policy applied to one or more of the selection by the determination unit and the action decision by the action determination unit. Claim 2 A content target automatic interaction determination device according to claim 1, wherein the history storage unit stores the decision history, which further includes the reason for exclusion for candidate content items excluded by the rule-based judgment, the judgment of the judgment unit, and the version of the policy applied to the action decision of the action decision unit. Claim 3 In paragraph 2, the policy adjustment unit generates an adjustment signal that changes the exclusion criteria corresponding to the exclusion reason so that the range of candidate content items excluded by the rule-based judgment is reduced when the frequency of occurrence of the exclusion reason is above a standard and the human approval result for the excluded candidate content item differs from the rule-based judgment. Claim 4 delete Claim 5 A content target automatic interaction determination device according to claim 1, wherein the action determination unit determines the action among writing a response, displaying a reaction, author relationship action, requesting approval, and skipping, and determines to perform displaying a reaction instead of writing a response based on at least one of an action budget and an action portfolio policy. Claim 6 A content target automatic interaction determination device, wherein the determination unit converts each of the candidate content items into a common schema including one or more of a content identifier, an author identifier, a text, a time of creation, a reaction indicator, and an author influence indicator. Claim 7 A content target automatic interaction determination device according to claim 1, wherein the determination unit determines duplicate contact conditions for the same content item or the same author based on past interaction history, excludes candidate content items corresponding to the duplicate contact conditions from the interaction target candidates, or restricts some of the plurality of behavior types for candidate content items corresponding to the duplicate contact conditions and determines the behavior among the remaining behavior types. Claim 8 A content target automatic interaction determination device according to claim 1, wherein the determination unit verifies at least one of a creator influence indicator and a reaction indicator for candidate content items not excluded by the rule-based determination, calculates a target score and a risk score based on the verification result and the artificial intelligence semantic determination result, and selects the interaction target candidates using the target score and the risk score. Claim 9 The content target automatic interaction determination device according to claim 1 further comprises an execution unit that transmits the action determined by the action determination unit to a human approval queue when the action corresponds to a risk criterion, transmits the action to a delay queue when the action corresponds to an uncertainty criterion, and executes the action on a target content source when the action does not correspond to either the risk criterion or the uncertainty criterion. Claim 10 In claim 1, the content target automatic interaction determination device further includes an execution unit that executes the action determined by the action determination unit, and the history storage unit stores the decision history that further includes the result of executing the action by the execution unit. Claim 11 A method for automatically determining interaction with content, performed by a computing device including a processor and memory, comprising: a step in which the computing device collects a plurality of candidate content items from a content source; a step in which the computing device applies a rule-based judgment to the candidate content items to exclude some of the candidate content items; a step in which the computing device obtains an artificial intelligence semantic judgment result for candidate content items not excluded by the rule-based judgment and selects interaction target candidates using the artificial intelligence semantic judgment result; a step in which the computing device determines at least one action among a plurality of action types for the selected interaction target candidates; a step in which the computing device stores a decision history including the result and reason for the rule-based judgment, the artificial intelligence semantic judgment result, and the determined action; and a step in which the computing device aggregates the stored decision history to generate a coordination signal for a policy applied to one or more of the selection of interaction target candidates and the determination of the action. Claim 12 A method for automatically determining content target interactions according to claim 11, wherein the step of storing the decision history further includes the reason for exclusion for candidate content items excluded by the rule-based judgment, and the version of the policy applied to the screening and the decision of the action. Claim 13 A method for automatically determining content target interactions according to claim 12, wherein the step of generating the adjustment signal comprises the step of generating the adjustment signal to change the exclusion criteria corresponding to the exclusion reason so as to reduce the range of candidate content items excluded by the rule-based judgment when the frequency of occurrence of the exclusion reason is greater than or equal to a standard and the human approval result for the excluded candidate content item differs from the rule-based judgment. Claim 14 delete Claim 15 A method for determining automatic interaction with content according to claim 11, wherein the step of determining the action comprises determining the action among writing a response, displaying a reaction, author relationship behavior, requesting approval, and skipping, and determining to perform displaying a reaction instead of writing a response based on at least one of an action budget and an action portfolio policy. Claim 16 In claim 11, the automatic interaction determination method for content targets further comprises the step of the computing device converting each of the candidate content items into a common schema including one or more of a content identifier, an author identifier, a text, a time of creation, a reaction indicator, and an author influence indicator. Claim 17 In claim 11, the automatic interaction determination method for content targets further comprises the step of the computing device determining duplicate contact conditions for the same content item or the same author based on past interaction history, and the step of selecting interaction target candidates comprises the step of excluding candidate content items corresponding to the duplicate contact conditions from the interaction target candidates, or restricting some of the plurality of behavior types for candidate content items corresponding to the duplicate contact conditions and determining the behavior among the remaining behavior types. Claim 18 A method for automatically determining interaction for content targets according to claim 11, wherein the step of selecting the interaction target candidates comprises: verifying at least one of the author influence indicator and the reaction indicator for candidate content items not excluded by the rule-based judgment; calculating a target score and a risk score based on the verification result and the artificial intelligence semantic judgment result; and selecting the interaction target candidates using the target score and the risk score. Claim 19 In claim 11, the automatic interaction determination method for a content target further comprises: a step of transmitting the action to a human approval queue if the determined action corresponds to a risk criterion; a step of transmitting the action to a delay queue if the action corresponds to an uncertainty criterion; and a step of executing the action on a target content source if the action does not correspond to either the risk criterion or the uncertainty criterion. Claim 20 In claim 11, the automatic interaction determination method for content targets further comprises the step of the computing device executing the determined action, and the step of storing the decision history further comprises the step of storing the decision history including the result of executing the action.

Citation Information

Patent Citations

  • Method for recommending based on context-awareness and apparatus thereof

    KR1020190064042A

  • Apparatus for selecting data to train artificial intelligence and method for controlling the same

    KR1020240103948A

  • Fact-consistency-based insight-oriented automatic article generation and publishing system

    KR1020250176156A

  • System providing user customized social network service marketing information based on automatically generated content based artificial intelligence and operating method thereof

    KR102953166B1