System for providing matching between adoption applicants and companion animals based on artificial intelligence

Abstract

According to one embodiment of the present invention, a user input receiving unit may receive user input, and a preference interpretation unit may interpret survey responses or natural language data included in the user input to calculate user preference distribution parameters by attribute, hard constraints, and importance weights by attribute. An attribute extraction unit may extract quantitative and categorical attributes of animals from animal-related data and calculate reliability to generate animal attribute information. A matching engine may compare the user preference distribution parameters with the animal attribute information to determine hard constraints, calculate similarity by attribute, score soft conditions by reflecting the importance weights by attribute, and calculate a final matching score based on the soft condition score and hard constraints. A recommendation providing unit may select recommended target animals based on the final matching score, output them to a user interface, and update the user preference distribution parameters and importance weights by collecting user behavior logs and feedback information.

Description

Technology Field

[0001] The present invention relates to an artificial intelligence-based system that interprets user input to calculate preference distribution parameters, constraints, and weights for each attribute, compares them with the attributes of animals to calculate a final matching score, and then provides a recommendation result. Background Technology

[0002] As social interest in pet adoption grows, adoption management services utilizing information and communication technology and online platforms are expanding. In particular, technologies that recommend suitable candidates by reflecting user characteristics and needs are being researched across various industries, and the introduction of such customized matching services is becoming increasingly active in the pet sector as well.

[0003] The development of artificial intelligence and data analysis technologies provides a foundation for structuring and utilizing vast amounts of information, presenting the potential for the evolution of services that consider the mutual compatibility between humans and pets.

[0004] Existing technologies related to pet adoption have primarily relied on simple searches based on basic attributes or conditional filtering methods. For example, matching services based on limited criteria such as an animal's age, gender, and size often failed to adequately reflect the user's living environment or care conditions. Furthermore, there were issues where it was difficult to effectively interpret and incorporate incomplete shelter records or ambiguous user preferences into the matching process.

[0005] Due to these limitations, the need for improvement has been continuously raised, as there have been cases where adoption was annulled due to poor suitability in real life after adoption.

[0006] Therefore, there is a growing need for technology capable of deriving precise matching results by accurately reflecting users' diverse environmental conditions and preferences while also considering the individual tendencies of animals. The problem to be solved

[0007] The problem that the present invention aims to solve is to provide a system that calculates preference distribution parameters, constraints, and weights for each attribute from user input, compares them with the quantitative and categorical attributes of an animal to derive a final matching score, and enables precise evaluation of the suitability between a user and an animal through a platform equipped with said system.

[0008] Another task is to provide a reliable matching service on the platform by selecting multiple recommended target animals based on the above system and providing the results through a user interface.

[0009] The problems that the present invention aims to solve are not limited to those mentioned above. means of solving the problem

[0010] According to one embodiment of the present invention, a user input receiving unit may receive user input, and a preference interpretation unit may interpret survey responses or natural language data included in the user input to calculate user preference distribution parameters by attribute, hard constraints, and importance weights by attribute. An attribute extraction unit may extract quantitative and categorical attributes of animals from animal-related data and calculate reliability to generate animal attribute information. A matching engine may compare the user preference distribution parameters with the animal attribute information to determine hard constraints, calculate similarity by attribute, score soft conditions by reflecting the importance weights by attribute, and calculate a final matching score based on the soft condition score and hard constraints. A recommendation providing unit may select recommended target animals based on the final matching score, output them to a user interface, and update the user preference distribution parameters and importance weights by collecting user behavior logs and feedback information.

[0011] According to one embodiment of the present invention, when calculating user preference distribution parameters by attribute, the preference analysis unit may normalize quantitative attributes to model them as a beta distribution and convert categorical attributes into probability values ​​to model them as a Dirichlet distribution.

[0012] According to one embodiment of the present invention, the preference analysis unit can extract essential conditions from user input as hard constraints using user preference distribution parameters for each attribute. Additionally, the preference analysis unit can calculate importance weights for each attribute by evaluating the relative importance of each attribute.

[0013] According to one embodiment of the present invention, a matching engine can determine whether a hard constraint has been violated by comparing user preference distribution parameters for each attribute with animal attribute information. The matching engine can calculate a first attribute-specific similarity based on probability distribution values ​​or distance functions for the quantitative attributes of the animal, and can calculate a second attribute-specific similarity based on probability for the categorical attributes of the animal. It can calculate a soft condition score by reflecting attribute-specific importance weights in the first and second attribute-specific similarities, and calculate a final matching score by combining the soft condition score with penalty and bonus items for violation of hard constraints.

[0014] According to one embodiment of the present invention, a recommendation provider may select a plurality of recommended target animals based on a final matching score. For each selected animal, the recommendation provider may output a recommendation result to a user interface that includes the final matching score, basis information by attribute, whether hard constraints are satisfied, and explainable reasons. Effects of the invention

[0015] According to the present invention, the suitability between a user and an animal can be precisely evaluated by calculating a final matching score by comparing attribute-specific preference distribution parameters, constraints, and attribute-specific importance weights derived from user input with the quantitative and categorical attributes of the animal.

[0016] Accordingly, since calculations reflecting probability distributions and weights are performed during the matching process, going beyond simple attribute comparison, the reliability of recommendations is enhanced and the presentation of unnecessary candidates is reduced. Furthermore, the recommendation provider can deliver the final matching score, supporting information for each attribute, whether hard constraints were satisfied, and explainable reasons to the user interface, thereby improving the ease of interpreting results and enhancing the user experience.

[0017] In addition, according to the present invention, user behavior logs and feedback information can be collected to update preference distribution parameters and importance weights by attribute, thereby improving the accuracy of personalized recommendations over time. This enables stable adoption matching on the platform and allows for the provision of a recommendation service that comprehensively reflects user needs and animal characteristics. Brief explanation of the drawing

[0018] FIG. 1 is a block diagram illustrating the overall configuration of an artificial intelligence-based adoption hopeful-pet matching system according to one embodiment of the present disclosure. FIG. 2 is a block diagram illustrating the configuration of a user input receiving unit and a preference interpreting unit according to one embodiment of the present disclosure. FIG. 3 is a block diagram illustrating the configuration of an attribute extraction unit according to one embodiment of the present disclosure. FIG. 4 is a flowchart illustrating the operation procedure of a matching engine according to one embodiment of the present disclosure. FIG. 5 is a diagram illustrating the operation procedure of a recommendation providing unit according to one embodiment of the present disclosure. Specific details for implementing the invention

[0019] Hereinafter, various embodiments of the present invention will be described in detail with reference to the attached drawings so that those skilled in the art can easily implement the present invention. The present invention may be embodied in various different forms and is not limited to the embodiments described herein.

[0020] To clearly explain the present invention, parts unrelated to the explanation have been omitted, and the same reference numerals are assigned to identical or similar components throughout the specification. Accordingly, the reference numerals described above may also be used in other drawings.

[0021] Furthermore, the size and thickness of each component shown in the drawings are depicted arbitrarily for convenience of explanation, and thus the present invention is not necessarily limited to what is illustrated. Thickness may be exaggerated in the drawings to clearly represent various layers and regions.

[0022] Furthermore, the expression "identical" in the explanation may mean "substantially identical." In other words, it may be an identicality to the extent that a person with ordinary knowledge would accept it as identical. Other expressions may also be those in which "substantially" has been omitted.

[0023] Furthermore, when a part in the description is described as 'including' a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. As used in this specification, 'part' refers to a unit that processes at least one function or operation, and may, for example, mean software, FPGA, or hardware components. The function provided by the 'part' may be performed separately by multiple components or integrated with other additional components. The 'part' in this specification is not necessarily limited to software or hardware, and may be configured to reside in an addressable storage medium or configured to run one or more processors. Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[0025] FIG. 1 is a block diagram illustrating the overall configuration of an artificial intelligence-based adoption hopeful-pet matching system according to one embodiment of the present disclosure.

[0026] Referring to FIG. 1, the artificial intelligence-based adoption hopeful-pet matching system (1) may be composed of multiple components and may be composed of a user input receiving unit (10), a preference interpretation unit (20), an attribute extraction unit (30), a matching engine (40), a recommendation providing unit (50), and a user interface (60).

[0027] Each component (10, 20, 30, 40, 50, 60) can operate independently and can also be interconnected to perform the overall function of evaluating the suitability between the user and the animal and providing results.

[0028] Meanwhile, the artificial intelligence-based adoption hopeful-pet matching system (1) according to the present embodiment may omit or add some of the components shown in FIG. 1, which means that various modifications and changes are possible within the scope of the technical concept of the present invention.

[0029] The user input receiving unit (10) can receive survey responses or natural language data input provided by the user (adoption hopeful).

[0030] Survey responses may include the user's lifestyle / environment (activity level, living space, frequency of going out, family composition, current presence of pets, ...), experience / attitude (pet experience, mindset / willingness to care, ...), preference (animal size, gender, age group, ...), constraints (rental regulations, budget range, working hours, ...), etc., and natural language input may be collected as freely written sentence data. The user input receiving unit (10) may convert the user input into a digital data format and transmit it to the preference interpretation unit (20).

[0031] The preference analysis unit (20) can analyze data transmitted from the user input receiving unit (10) to calculate user preference distribution parameters for each attribute. At this time, the user's quantitative attributes are numerical data such as activity level, living space size, and budget level, which can be modeled as a beta distribution after undergoing a normalization process, and the user's categorical attributes are classification data such as the size, gender, and breed type of the preferred animal, which can be converted into probability values ​​and modeled as a Dirichlet distribution.

[0032] Additionally, the preference interpretation unit (20) can identify the essential conditions included in the user input and determine them as hard constraints, and evaluate the relative importance of each attribute to calculate the importance weight for each attribute, and the parameters and conditions calculated in this way can be used as basic information for calculating the final matching score in the matching engine (40) thereafter.

[0033] The attribute extraction unit (30) can extract quantitative attributes and categorical attributes of animals from animal-related data.

[0034] The quantitative attributes of an animal may include items expressed as quantitative or qualitative scores between 0 and 5, such as activity level, sensitivity, sociability, and separation anxiety. The categorical attributes of an animal may consist of size (S / M / L), sex, age range, neutering status, history, specifics, etc., and may be converted into classification data for processing.

[0035] The attribute extraction unit (30) can also collect unstructured text data such as shelter records or observation notes, and this text data can be converted into structured information using a large-scale language model. Through this, additional context regarding the characteristics of the animal individual can be extracted, and animal attribute information combined with quantitative and categorical attributes can be generated.

[0036] The matching engine (40) can compare the user attribute distribution parameter provided by the preference interpretation unit (20) with the animal attribute information provided by the attribute extraction unit (30).

[0037] The matching engine (40) can first determine whether hard constraints are satisfied, and for the quantitative attributes of the animal, it can calculate the attribute-specific similarity (hereinafter, first attribute-specific similarity) with the user preference distribution based on probability distribution values ​​or distance functions.

[0038] In addition, the matching engine (40) can perform probability-based calculations on the categorical attributes of animals to calculate the attribute-specific similarity (hereinafter, second attribute-specific similarity) with the user's categorical preference.

[0039] The matching engine (40) can calculate a soft condition score by reflecting the importance weight of each attribute to the similarity of the first and second attributes, and can calculate a final matching score by combining this with a penalty based on whether the hard constraint is violated and a bonus item based on positive factors.

[0040] The recommendation provider (50) can select multiple recommended target animals based on the final matching score calculated by the matching engine (40).

[0041] The recommendation provider (50) can generate recommendation results for each selected animal, including a final matching score, basis information by attribute, whether hard constraints are satisfied, and explainable reasons. The generated recommendation results can be provided to the user through a user interface (60), and visual and textual information can be output together to make it easy for the user to understand.

[0042] The user interface (60) can display recommendation results transmitted from the recommendation provider (50). For example, it can list recommended target animals according to the final matching score ranking and provide key attributes and explainable reasons for each animal. Additionally, the user interface (60) can be configured to allow the user to input feedback or leave a behavior log.

[0043] The recommendation providing unit (50) can collect user behavior logs and feedback information to update the user preference distribution parameters and attribute-specific importance weights calculated by the preference interpretation unit (20). This update process can improve the accuracy of user-customized recommendations over time and continuously improve the matching performance of the entire system.

[0045] FIG. 2 is a block diagram illustrating the configuration of a user input receiving unit and a preference interpreting unit according to one embodiment of the present disclosure.

[0046] The user input receiving unit (10) can collect various forms of user input provided by the user and convert them into digitized data, and the preference interpretation unit (20) can interpret the user input to calculate user preference distribution parameters by attribute, hard constraints, and importance weights by attribute.

[0047] The user input receiving unit (10) can collect both structured input in the form of a survey and free-form natural language input.

[0048] Survey input may include attributes related to lifestyle and environment (activity level, living space, frequency of going out, family composition, current ownership of pets, etc.), attributes related to experience and attitude (experience with pets, willingness and attitude toward care, etc.), preference attributes (size, gender, age group preference, etc.), and constraints (rental regulations, budget range, working hours, etc.).

[0049] For example, a user may respond as "living space=medium-sized apartment, activity level=medium, family=living with children, budget=300,000 won or less per month," and this data can subsequently be classified into quantitative and categorical attributes.

[0050] Natural language input can be provided as free-form sentences such as “I want a dog that gets along well with my child,” “Large dogs are difficult,” and “I would like a personality that is active but not overly sensitive,” and the user input receiving unit (10) can convert this into digital data through a text preprocessing process.

[0051] The preference interpretation unit (20) can analyze the data transmitted from the user input receiving unit (10) and calculate user preference distribution parameters classified into quantitative attributes and categorical attributes.

[0052] Quantitative attributes of users refer to numerical data such as activity level, living space size, and budget level, which can be modeled using a Beta distribution after being normalized to the [0,1] range.

[0053] For example, if activity levels are converted into values ​​such as "low (0.2)", "medium (0.5)", and "high (0.8)", and preference tendencies are concentrated at the median value, the α and β parameters of the Beta distribution can be adjusted to allocate a high probability mass to the median region.

[0054] Users' categorical attributes refer to classification data such as size (S / M / L), preferred breed, gender, and age group, which can be modeled using the Dirichlet distribution.

[0055] For example, if a user inputs "Medium is the most preferred, but small is acceptable, and large is not," the parameters of the Dirichlet distribution can be set as M=5, S=2, and L=0.5 to assign probabilities in the order of "Medium > Small > Large".

[0056] The preference interpretation unit (20) can also extract essential conditions from user input and determine them as hard constraints.

[0057] For example, the input "Large dogs not allowed" can be set as a hard constraint on the size attribute, and the input "Must be neutered" can be set as a constraint on the neutering status attribute. Such hard constraints are reflected as conditions that must be satisfied at the matching engine stage.

[0058] In addition, the preference interpretation unit (20) can evaluate the relative importance of each attribute and calculate the importance weight of each attribute.

[0059] For example, if the user inputs "Social skills are very important," a high weight may be assigned to the social skills attribute, and if the user inputs "Gender is not important," a low weight may be reflected for that attribute. The calculated weights can be normalized so that the sum of all attributes is 1.

[0060] Ultimately, the user input receiving unit (10) and preference interpretation unit (20) illustrated in FIG. 2 can reflect both survey-based data and natural language data to precisely model the user's quantitative and categorical preferences into a probability distribution, and provide basic data for calculating the final matching score in the matching engine (40) along with hard constraints and importance weights for each attribute.

[0062] FIG. 3 is a block diagram illustrating the configuration of an attribute extraction unit according to one embodiment of the present disclosure.

[0063] The attribute extraction unit (30) can collect animal-related data to extract quantitative attributes and categorical attributes, and can also generate animal attribute information by calculating the reliability of each attribute value.

[0064] In this process, the attribute extraction unit (30) can preprocess the input data and convert it into a database format so that it can be directly utilized by the matching engine (40) thereafter.

[0065] The attribute extraction unit (30) can extract quantitative attributes of the animal.

[0066] Quantitative attributes include items such as activity level, sensitivity, sociability, and separation anxiety, each of which can be converted into a numerical value or score between 0 and 5 based on observations or past records at the shelter.

[0067] For example, an active individual can be recorded as Activity Level = 5, and an individual with no separation anxiety at all can be expressed as Separation Anxiety = 0. These quantitative attributes can be consistently managed according to standardized scales.

[0068] The attribute extraction unit (30) can also extract categorical attributes of animals.

[0069] Categorical attributes can consist of items such as size (S / M / L), gender, age range, neutering status, history, and specifics, and can be converted into a classified data format for processing.

[0070] For example, information such as "medium, male, 3 years old, neutered, abandoned" can be classified into a set of categorical attributes. In this case, each categorical attribute is organized by a clear code system and can be consistently utilized in matching calculations.

[0071] The attribute extraction unit (30) may include a function for structuring unstructured text data.

[0072] Free-form data, such as shelter records or observation notes, can be analyzed using large-scale language models, and personality traits, peculiar behaviors, and health-related notes can be automatically extracted from the records.

[0073] For example, the record "friendly to people but sensitive to unfamiliar entities" can be converted into quantitative attributes such as sociality=4 and sensitivity=3.

[0074] In addition, the attribute extraction unit (30) can calculate the reliability of each attribute.

[0075] For example, recently updated records or data confirmed by repeated verification by an observer may be assigned high reliability, while old or uncertain records may be assigned low reliability. The calculated reliability value can be reflected in the weighting adjustment for each attribute in the matching engine (40) to increase the stability of the result.

[0076] The attribute extraction unit (30) can finally generate comprehensive animal attribute information including quantitative attributes, categorical attributes, attributes structured from unstructured data, and reliability for each attribute.

[0077] This combined animal attribute information can be transmitted to a matching engine (40) and used as target data to be compared with user preference distribution parameters.

[0078] As such, the attribute extraction unit (30) illustrated in FIG. 3 can perform the function of generating animal attribute information by extracting quantitative attributes and categorical attributes from animal-related data and calculating reliability, and can provide key basic data to enable precise comparison with user preferences during the matching process.

[0080] FIG. 4 is a flowchart illustrating the operation procedure of a matching engine according to one embodiment of the present disclosure.

[0081] The matching engine (40) can perform the function of calculating a final matching score by combining user input and animal attribute data, and can perform comparison and calculation step by step through the step-by-step procedure described below.

[0082] In step (S10), the matching engine (40) can receive user preference distribution parameters, hard constraints, and attribute-specific importance weights from the preference analysis unit (20).

[0083] At the same time, the matching engine (40) can receive animal attribute information including quantitative attributes, categorical attributes, and reliability from the attribute extraction unit (30).

[0084] The data obtained in this step (S10) can be used as basic data for all subsequent operations, and data integrity and consistency must be ensured.

[0085] In step (S20), the input user hard constraints and animal attribute information can be compared to determine whether the condition is satisfied.

[0086] For example, if a user enters "Large dogs not allowed," individuals with a size attribute of Large (L) may be immediately excluded. As another example, if there is a condition "Must be neutered," animals that have not been neutered may be automatically excluded. These hard constraints must be strictly adhered to, and violations may be reflected as penalties during the subsequent score calculation process.

[0087] In step (S30), the similarity between the animal's quantitative attributes and the user preference distribution for each first attribute can be calculated. The quantitative attributes may consist of numerically expressed items such as activity level, sociability, sensitivity, and separation anxiety. At this time, the similarity can be calculated as [Equation 1].

[0088]

[0089] Here, vk represents the attribute value of the animal, Pk(vk) represents the probability density of the corresponding value in the user Beta distribution, and μk and σk represent the mean and variance of the distribution, respectively. For example, if the user's activity preference is concentrated in the "Medium (0.5)" and the activity level of a specific animal is rated as 0.6, a high similarity can be calculated.

[0090] In step (S40), the similarity between the animal's categorical attributes and the user's categorical preferences can be calculated for each second attribute. The categorical attributes are expressed as classification data such as size (S / M / L), gender, age range, and breed, and the similarity can be calculated as [Equation 2].

[0091]

[0092] Here, vk is the animal category value, and Pk(c=vk) can correspond to the probability of the corresponding category assigned by the user through the Dirichlet distribution.

[0093] For example, if a user inputs preferences such as "Medium=0.7, Small=0.3, Large=0.0", medium-sized entities can obtain high similarity. On the other hand, large entities are evaluated as having a similarity of 0 and can effectively be excluded from recommendations.

[0094] In step (S50), the similarity between the first and second attributes calculated in steps (S30) and (S40) can be combined with the importance weights for each attribute to calculate a soft condition score.

[0095] The importance weight for each attribute is a value calculated by the preference interpretation unit (20), and if the user responds that "sociality is very important," a high weight may be assigned to the sociality attribute. Accordingly, the sociality score can contribute significantly to the overall soft condition score, and the gender attribute, which was responded as not important, may be reflected with a low weight.

[0096] In step (S60), a final matching score can be calculated by combining the soft condition score, the satisfaction of hard constraints, and bonus items. At this time, the final score can be defined as [Equation 3].

[0097]

[0098] Here, wk is the importance weight for each attribute, j represents a penalty coefficient applied when a hard constraint is violated, and γ represents a bonus weight given to socially beneficial factors such as long-term care animals or animals with specific diseases.

[0099] For example, if an animal meets most of the user's criteria and is classified as a long-term care animal, bonus points are added, which can further improve the final matching score.

[0100] The calculated final matching score is transmitted to the recommendation provider (50) and can serve as a criterion for selecting multiple recommended candidate animals. In this way, the operation procedure of the matching engine (40) illustrated in FIG. 4 can precisely evaluate the suitability between the user and the pet by integrally reflecting the user input-based probability distribution, animal attribute information, constraints, and bonus elements.

[0102] FIG. 5 is a diagram illustrating the operation procedure of a recommendation providing unit according to one embodiment of the present disclosure.

[0103] The recommendation provider (50) can receive the final matching score transmitted from the matching engine (40) and select multiple recommended target animals.

[0104] The recommendation provider (50) can sort animals in order of highest final matching score or filter candidates that meet a preset threshold to determine the final recommendation target.

[0105] At this time, since the matching score for each animal is a value that comprehensively reflects the suitability between user preference and animal attributes, the recommendation provider (50) can select a reliable candidate based on this.

[0106] The recommendation provider (50) can generate recommendation results for each selected animal that include not only the final matching score but also the basis for similarity by attribute, whether hard constraints are satisfied, and explainable reasons.

[0107] For example, if a user inputs "prefers medium-sized dogs, social skills are important, large dogs are not allowed," the recommendation provider (50) can provide explanatory information about a specific animal with the basis that "social skills are high at the 4 / 5 level, and size = medium, satisfying the condition."

[0108] These explainable recommendation results are not simple score-based suggestions, but allow users to intuitively verify the rationality of the results.

[0109] The recommendation provider (50) can output the generated recommendation results to the user interface (60). The output method can be implemented in various ways, such as a ranking table based on the final matching score, a summary information card by attribute, or an indication of whether conditions are met.

[0110] In addition, the recommendation provision unit (50) can configure the results so that the user can intuitively check visual and textual information, thereby supporting the user (adoption hopeful) in making a quick and accurate decision.

[0111] The recommendation provider (50) may also include a function to collect feedback and behavior logs from the user. If the user favorites a specific animal, marks it as 'unsuitable', or leaves additional comments, the recommendation provider (50) may record the information and reflect it in the system's database.

[0112] Subsequently, this data is transmitted to the preference interpretation unit (20) and the matching engine (40) and can be used to update user preference distribution parameters and importance weights by attribute.

[0113] As such, the recommendation providing unit (50) illustrated in FIG. 5 can go beyond simply outputting a final score and can integrally perform a series of processes including the selection of multiple candidate animals, the generation of explainable recommendation results, the output of a user interface, and the collection of user logs and feedback.

[0114] Through this, the recommendation provider (50) can provide a stable and reliable adoption matching service on the platform and play a key role in improving the accuracy of the system (1) and user satisfaction over time.

[0116] The drawings and detailed description of the invention referenced so far are merely exemplary of the invention and are used only for the purpose of explaining the invention, not to limit the meaning or the scope of the invention as defined in the claims. Therefore, those skilled in the art will understand that various modifications and equivalent alternative embodiments are possible therefrom. Accordingly, the true technical scope of protection of the invention should be determined by the technical spirit of the appended claims.

[0117] The embodiments described above may be implemented as hardware components, software components, and / or combinations of hardware components and software components. For example, the devices, methods, and components described in the embodiments may be implemented using one or more general-purpose computers or special-purpose computers, such as, for example, a processor, a controller, an Arithmetic Logic Unit (ALU), a Digital Signal Processor (DSP), a microcomputer, a Field Programmable Gate Array (FPGA), a Programmable Logic Unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions.

[0118] The processing unit may execute an operating system and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For convenience of understanding, the processing unit may be described as being used as a single unit, but a person of ordinary skill in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements.

[0119] For example, the processing unit may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as a parallel processor, are also possible. Software may include a computer program, code, instructions, or a combination of one or more of these, and may configure the processing unit to operate as desired or command the processing unit independently or collectively.

[0120] Software and / or data may be embodied in any type of machine, component, physical device, virtual equipment, computer storage medium, or device so as to be interpreted by a processing device or to provide instructions or data to a processing device. Software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0121] The method according to the embodiment 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 medium may be those specifically designed and configured for the embodiment, or they may be those known and available to those skilled in the art of computer software.

[0122] Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions 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 hardware devices described above may be configured to operate as one or more software modules to perform the operation of the embodiments, and vice versa.

[0123] Although the embodiments have been described above with reference to limited examples and drawings, those skilled in the art can make various modifications and variations from the description above. For example, suitable results may be achieved even if the described techniques are performed in a different order than described, and / or if the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents. Therefore, other implementations, other embodiments, and equivalents to the claims are also included within the scope of the claims set forth below. Explanation of the symbols

[0124] 1: AI-based adoption hopeful-pet matching system 10: User input receiver 20: Preference Interpretation Section 30: Attribute Extraction Section 40: Matching Engine 50: Recommendation Provider 60: User Interface

Claims

Claim 1 A user input receiving unit that receives user input composed of natural language data; and a preference interpretation unit that interprets the natural language data using a pre-trained artificial intelligence model to calculate user preference distribution parameters in the form of probability distributions defined for each attribute, hard constraints, and importance weights for each attribute; An attribute extraction unit that extracts quantitative and categorical attributes of animals from animal-related data and calculates reliability to generate animal attribute information; a matching engine that compares the user preference distribution parameter with the animal attribute information, determines hard constraints, calculates similarity by attribute, scores soft conditions by reflecting importance weights by attribute, and calculates a final matching score by reflecting hard constraints in the soft condition score; and a recommendation providing unit that selects a recommended target animal according to the final matching score, outputs it to a user interface, and collects user behavior logs and feedback information to update the user preference distribution parameter and the importance weights by attribute; wherein the matching engine calculates similarity by attribute based on the corresponding user preference distribution parameter by attribute and attribute information included in the animal attribute information, determines whether the hard constraint is satisfied separately from the calculation of similarity by attribute, applies a penalty to the soft condition score when the hard constraint is violated, or reflects a bonus to the soft condition score when satisfied to calculate the final matching score, and the preference interpretation unit interprets the natural language data and, for quantitative attributes, Beta An AI-based adoption hopeful-pet matching system that calculates user preference distribution parameters modeled as a distribution, and for categorical attributes, calculates user preference distribution parameters modeled as a Dirichlet distribution, wherein the hard constraints are essential conditions included in the user input for animal adoption. Claim 2 delete Claim 3 In claim 1, the preference analysis unit evaluates the relative importance of each attribute using the user preference distribution parameter for each attribute, and calculates the importance weight for each attribute based on the evaluated relative importance of each attribute, in an artificial intelligence-based adoption hopeful-pet matching system. Claim 4 An AI-based adoption hopeful-pet matching system according to claim 1, wherein the matching engine calculates a first attribute-specific similarity based on a probability distribution value or a distance function for the quantitative attributes of the animal, calculates a second attribute-specific similarity based on probability for the categorical attributes of the animal, and calculates a soft condition score by reflecting the attribute-specific importance weights to the first and second attribute-specific similarities. Claim 5 An AI-based adoption hopeful-pet matching system according to claim 1, wherein the recommendation providing unit selects a plurality of recommended target animals based on the final matching score, and outputs a recommendation result to the user interface for each selected recommended target animal, including the final matching score, basis information by attribute, whether the hard constraint is satisfied, and an explainable reason.

Images