KR101083804B1 - Ontology Based Method And System For Searching Similar Civil Petition - Google Patents

Abstract

The present invention comprises the steps of extracting the main word from the search data received the search data for the new complaint; Generating a query instance expressing the new complaint in the form of a triple by an ontology; Among the complaint instances expressing the inclusion relationship of existing complaints and existing complaints in triple form by ontology using the similar judgment rule that compares the main words and terms. Retrieving a complaint instance for an existing complaint similar to a new complaint; And generating a result instance expressing a similar relationship between the new complaint and the existing complaint of the retrieved complaint instance. The present invention relates to a method for retrieving similar complaints based on ontology and a system implementing the same.

According to the present invention, when a new complaint is input, an existing complaint similar to the new complaint and additional data related thereto are automatically searched, thereby enabling a quick and easy process for the new complaint.

Complaint, ontology, search

Description

Ontology Based Method And System For Searching Similar Civil Petition}

The present invention relates to a method and system for retrieving an existing complaint similar to a new complaint, and specifically, extracts the main word of the new complaint and generates a query instance expressed in triple form by the ontology for the new complaint and the subject. The present invention relates to a method and system for retrieving an existing complaint similar to a new complaint by determining whether the complaint instance and the query instance satisfy the similar judgment rule.

Complaint means a request for a resident of a region to make a request to an administrative body of that region. The affairs of such complaints are those concerning applications for permits, authorizations, licenses, registrations, etc., or appeals, complaints, recommendations, and inquiries. In the conventional civil affairs, in order to deal with such complaints, the complaint recipient reads the name of the complaint or the content of the complaint, classifies it, passes it to the person in charge of the complaint, and the person in charge of the complaint has to read the complaint name or the content of the complaint again and process it.

However, for the easy handling of these complaints, the civil service officer reads all the contents of the civil complaints, similar to the new civil complaints, and then selects a new keyword or the subject of the complaint and inputs the search term of the civil search program. After reading the contents of the various complaints entered into the window and reading them, the similar complaints had to be selected, but this was very cumbersome and required a double time and effort.

In addition, it was difficult for the complainant to know the additional information related to his complaint, such as the processing period, fees, processing flows, and related laws of the complaint, and it was difficult to predict the outcome of the complaint if the complainant was not an expert. Therefore, in order to anticipate additional information or processing results of a complaint, the complainant inputs a word or key sentence that is considered to be a keyword or a subject of the complaint in the search term input box of the complaint search program and is also searched for as a result. After reading the contents of various complaints, it was necessary to select similar complaints, but this was not easy for ordinary complaints.

In addition, in the case of the existing simple keyword matching method, only those complaints that include the search term entered in the search box input are searched. Even though similar complaints were searched, only the contents of the complaints were searched, so it was difficult to know the processing period, the processing flow, the person in charge, the related laws, etc.

In order to solve the above problems, an object of the present invention is to provide a method and system for searching for existing complaints similar to new complaints based on the ontology based on the similar decision rule.

An ontology-based similar complaint search system according to an embodiment of the present invention for achieving the above technical problem is an extraction unit for receiving the search data for the new complaints to extract the main word from the search data; A generation unit for generating a query instance expressing that the new complaint includes the main word in a triple form by an ontology; An instance storage unit for storing the existing civil complaints and the civil complaint instances expressing the inclusion relationship of the terms included in the existing civil complaints in triple form by the ontology; And searching for an instance of an existing complaint similar to the search data by using a similar decision rule that compares the main term and the term to determine the similarity between the new complaint and the existing complaint. It includes; search unit for generating a result instance representing the similarity of the complaint.

Ontology-based similar complaint search method according to an embodiment of the present invention for achieving the above technical problem is the step of extracting the main word from the search data received the search data for the new complaint; Generating a query instance expressing that the new complaint includes the main word in a triple form by ontology; Expressing the inclusion relationship between the existing complaints and terms included in the existing complaints in triple form by the ontology by using the similar judgment rule to compare the main term and the terms to determine the similarity between the new complaints and the existing complaints Retrieving a complaint instance for an existing complaint similar to the new complaint among the complaint instances; And generating a result instance representing a similar relationship between the new complaint and an existing complaint of the retrieved complaint instance.

In addition, the following provides a computer-readable recording medium having recorded thereon a program for executing a similar complaint search method based on the ontology described above.

According to the present invention, existing complaints similar to the new complaints to be filed are automatically searched, thereby simplifying the processing process of the civil affairs, making the process of civil affairs quicker and easier, and the civil affairs of the civil affairs without the professional knowledge of the civil affairs. Similarly, the complaints and their processing results are automatically retrieved.

In addition, since similar complaints are searched based on the ontology technology, a person seeking to search for similar complaints has an advantage that similar complaints related to the search term data input by the ontology inference method are searched without necessarily inputting an exact keyword.

Embodiments of the present invention are provided to more fully explain the present invention to those skilled in the art, the following examples can be modified in many different forms, the scope of the invention It is not limited to the following example. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the inventive concept to those skilled in the art.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. The term "and / or" includes any and all combinations of one or more of the listed items. It includes. Also, as used herein, “comprise” and / or “comprising” does not exclude the presence or addition of the steps, operations, components, and / or groups thereof mentioned.

Although terms such as first and second are used herein to describe various steps, operations and / or components, it is obvious that these steps, operations and / or components should not be limited by these terms. These terms are only used to distinguish one step, operation or component from another step, operation or component. Accordingly, the first step, operation or component described below may refer to the second step, operation or component without departing from the teachings of the present invention.

Prior to describing the embodiments of the present invention, terms and concepts to be used below are briefly described.

The semantic web is a machine, or computer, that can process information about resources (web documents, various files, services, etc.) and the relationships between them in a distributed environment such as the Internet. It is a framework and technology that expresses it in the form of ontology and allows an automated machine to process it. Here, ontology is a model that abstracts and shares what people think about things, and is a technology that has been formalized and explicitly defined types of concepts or usage constraints. In the field of computer science, ontology is a data model that represents a specific domain and is defined as a set of formal vocabularies describing the concepts and relationships between them. In particular, ontology is a tool that can implement the semantic web. It is used as a tool that can semantically connect knowledge concepts, and it enables the computer to process and process the concept of things in the form of a database. give.

The semantic web can be expressed as a semantic markup language based on XML (Extensible Markup Language). In the semantic web, the concept is expressed in triple form of subject, predicate, and object, and each subject, predicate, and object are represented as a Uniform Resource Identifier (URI) of XML. Can be. Currently, standard languages describing semantic web ontology include RDF, RDF-Schema, OWL, SW RL, and ISO proposed TopicMaps.

Hereinafter, with reference to the accompanying drawings, it will be described in detail an embodiment of the present invention. The same reference numerals are used for the same elements in the drawings for ease of explanation and clarity, and duplicate descriptions of portions that are not related to the description or the same elements are omitted.

1 is a block diagram showing the configuration of a similar complaint search system based on the ontology according to an embodiment of the present invention. In FIG. 1, the similar complaint search system 1 exchanges data with the service providing interface 2, and the similar complaint search system 1 includes an extraction unit 10, a generation unit 20, a search unit 30, and a result. It includes a processor 40, an instance storage unit 50, the inference unit 60 and the inference rule storage unit 70.

The service providing interface 2 is an interface in charge of exchanging data between the outside and the similar complaint search system 1. The service providing interface 2 forwards the new complaint to the similar complaint search system 1 when an external search service program or the like is requested to search for an existing complaint similar to the new complaint and / or to provide additional data related thereto. .

To this end, the service providing interface 2 transmits the search data for the new complaint to the similar complaint search system 1. In addition, the service providing interface 2 receives an existing complaint and / or additional data related thereto that are determined to be similar to the new complaint from the similar complaint system 1 and provides the same to an external search service program.

The new complaint refers to a new complaint that the user of the present invention intends to raise, and the existing complaint is a previously created and filed complaint that is expressed in triple form by an ontology related to the instance storage unit 50 of the similar complaint search system 1. The instance to be created is already created and stored.

The extractor 10 receives the search data for the new complaint and extracts the main word of the new complaint from the search data. The search data refers to data entered by a user of the present invention about a new complaint, such as a search term input window, to search for an existing complaint similar to a new complaint among previously filed complaints.

The search data can be the name or terminology of the complaint, the subject of the complaint, or a sentence in natural language for the new complaint. Search data can also be the entire content of a new complaint.

As the main word that can be extracted by the extractor 10, candidate words that can be the main word are determined, and the extractor 10 searches for and extracts whether a word matching the candidates is included in the search data. The extractor 10 may extract one or a plurality of such subject words with respect to one search data, and a method such as morpheme analysis may be used to extract the subject words.

For example, when one of the candidate words is "floating dust", the extraction unit 10 searches whether the search data includes the word "floating dust", and when the search data includes "floating dust", " "Floating dust" will be extracted.

Also, the extractor 10 may normalize the irregular words included in the search data through a matching method and convert the extracted normalized words into subject words. For example, if the candidate word of the subject is "floating dust" and the search data includes "floating dust", "floating dust", or "air dust", the extraction unit 10 includes the words included in the search data. The "floating dust" can be extracted as the main word by matching them with "floating dust".

The generation unit 20 generates a query instance when the extraction unit 10 extracts the main word. The query instance refers to an instance expressing in a triple form by the ontology that the new complaint includes the main word extracted by the extractor 10.

For example, suppose the class of a complaint is represented by d, its main class is w, and each instance is followed by a number. In this case, if the new complaint is d1 and the main word extracted by the extractor 10 is w1, the query instance generated by the generator 20 may be expressed as HasWord (d1, w1). This indicates that the subject of the complaint called d1 is w1. Here the subject will be d1, the object will be w1 and the predicate will be HasWord.

Applying this to the example of "floating dust", if the subject of the new complaint named d1 is "floating dust," the generation unit 20 will generate a query instance called HasWord (d1, floating dust). The generation unit 20 may store the generated query instance in the instance storage unit 50.

The search unit 30 searches whether the instance of the civil complaint similar to the new complaint is stored in the instance storage unit 50 using the similar decision rule. In addition, the search unit 30 generates a result instance when a complaint instance for an existing complaint similar to the new complaint is searched.

Complaint instance is an instance expressing the relationship between the existing complaints about the administrative agency and terms contained in the existing complaints in triple form by ontology, and its basic form is the same as the query instance. For example, if the existing complaint is d2 and d2 contains the term w1, the complaint instance would be HasWord (d2, w1). These complaint instances are previously stored in the instance storage unit 50. The terms included in these complaint instances are preferably selected to represent existing complaints.

The result instance is an instance expressing the similar relationship between the new complaint and the existing complaint of the retrieved complaint instance in a triple form by the ontology. For example, if it is determined that a new complaint called d1 and an existing complaint called d2 are similar, the resulting instance may be expressed as SimPet (d1, d2).

The similarity judgment rule refers to a rule that determines the similarity between a new complaint and an existing complaint by comparing the terms included in the query instance with the terms included in the complaint instance. Such a similar decision rule may be expressed as first-order logic based on the relationship between the main word and the term. Such judging rules include one or more rules.

The rule of similarity may include a rule that an existing complaint that contains the same terms as the subject is similar to a new complaint. This rule is expressed by the following first logical expression:

HasWord (d1, w) ^ HasWord (d2, w) => SimPet (d1, d2)

This means that if the main word of the new complaint d1 is w and the term contained in the existing complaint d2 is w and the terms and terms are the same, the new complaint d1 and the existing complaint d2 are similar. For example, if the query instance is HasWord (Complaint 1, Floating Dust) and a Complaint instance called HasWord (Complaint 2, Floating Dust) is searched, the search unit 30 generates a result instance called SimPet (Complaint 1, Complaint 2). The search unit 30 may store the generated result instance in the instance storage unit 50.

The similarity determination rule may include the rule that existing complaints that contain terms that are subject to or similar to the subject are similar to the new complaints. For this purpose, the instance storage unit 50 includes the term instance.

The term instance is an instance that expresses the relationship between the main word and the term in a triple form by the ontology. An example of a term ontology that defines the relationship between atmospheric-environmental terms with respect to this term instance is shown in FIG. 2.

If the subject w1 is a subword of the term w2, the term instance will be expressed as SubClassOf (w1, w2). In addition, when the term w2 is a subword of the main word w1, the term instance will be expressed as SubClassOf (w2, w1). If the subject w1 is similar to the term w2, the term instance will be expressed as EquivalentClassOf (w1, w2). This subject w1 and the term w2 are one of the words shown in the term ontology of FIG.

As a specific example, in FIG. 2, since the processing apparatus is a lower word of the environmental facility, the term instance related to this may be SubClassOf (processing device, environmental facility).

In addition, although not shown in FIG. 2, in relation to the above-described example, since the floating dust is a lower word of dust, the term instance will be SubClassOf (floating dust, dust) in this case. In addition, since scattering dust is a sub word of floating dust, the term instance may be SubClassOf (floating dust, floating dust). In another example, a diesel vehicle is a subordinate of a car, so the term instance would be SubClassOf.

The rule that existing complaints, including terms that are related to the subject, are similar to the new complaints, is expressed by the following first logical expressions.

HasWord (d1, w1) ^ HasWord (d2, w2) ^ SubClassOf (w1, w2) => SimPet (d1, d2)

HasWord (d1, w1) ^ HasWord (d2, w2) ^ SubClassOf (w2, w1) => SimPet (d1, d2)

The rule that the existing complaint d2, which includes the term w2, which is similar to the subject w1, is similar to the new complaint d1, is expressed by the following first logical expression.

HasWord (d1, w1) ^ HasWord (d2, w2) ^ EquivalentClassOf (w1, w2) => SimPet (d1, d2)

Examples of applying the above specific examples to these first order logical expressions are as follows.

HasWord (Complaint 1, Complaint) ^ HasWord (Complaint 2, Completion) ^ SubClassOf (Processing Device, Completion) => SimPet

HasWord (Complaint 1, Shattering Dust) ^ HasWord (Crime 2, Floating Dust) ^ SubClassOf (Scattered Dust, Floating Dust) => SimPet

HasWord (Complaint 1, Car) ^ HasWord (Complaint 2, Light Car) ^ SubClassOf (Light Vehicle, Car) => SimPet (Complaint 1, Complaint 2)

HasWord (Complaint 1, floating dust) ^ HasWord (Complaint 2, fine dust) ^ EquivalentClassOf (Floating dust, fine dust) => SimPet (Complaint 1, Complaint 2)

In this case, the search unit 30 searches for an instance of an existing complaint similar to the new complaint using the term instance and the similar decision rule stored in the instance storage unit 50, and then generates a result instance according to the search result.

The Markov Logic Network (MLN) may be applied to the similar complaint search system of the present invention. The Markov Logic Network was proposed by Pedro Domingos and Matthew Richardson of the University of Washington in a paper.

In the Markov logic network, a term refers to an object, which may be a constant, variable, or function. And predicates in the Markov logic network represent relationships between objects.

In the Markov logic network M _{L, C} , L is a set {(F _i , w _i )} and C is a finite set of constants C = {c ₁ , c ₂ , ..., c _n }. In L, F _i is the first-order logical expression. w _i is the weight of the first-order logical expression. And each node in the graph of Markov logic network, M _L, is when the _C 1, false when true that one of the constant C is substituted into the first logical expression F _i of the L has a value of zero.

Since the Markov logic network itself is not a subject of the present invention and is already known, a more detailed description of the Markov logic network itself will be omitted and the following will explain how to apply the Markov logic network to the ontology model of the present invention.

There may be several ways to represent the ontology model of the present invention in a Markov logic network. One of them is that one class constituting the ontology corresponds to the predicate of the Markov logic network, and the predicate in the ontology predicates another Markov logic network predicate between instances belonging to the class. To see). In this case, each new and existing complaint is a set of statements that indicate which instances are true for the predicates and instances of the ontology.

The second way is to view a complaint as a set of terms rather than a set of propositions. In other words, each class or instance of the ontology corresponds to a term. And the predicates in the ontology become predicates between the words of the Markov logic network.

According to the second method, the first-order logics of the similar decision rules described above correspond to the first-order logics of the Markov logic network, and they represent one node when instances of ontology are assigned as constants when graphing the Markov logic network. Is achieved.

Accordingly, the similar decision rule may include a plurality of rules to which weights according to the Markov logic network are applied. For example, the plurality of rules that can be expressed as a first order logical expression may be as follows. In the following, d1 denotes a new complaint, d2 denotes an existing complaint, w1 denotes a main term, and w2 denotes a term.

HasWord (d1, w) ^ HasWord (d2, w) => SimPet (d1, d2)

HasWord (d1, w1) ^ HasWord (d2, w2) ^ (w1! = W2) =>! SimPet (d1, d2)

HasWord (d1, w1) ^ HasWord (d2, w2) ^ SubClassOf (w1, w2) => SimPet (d1, d2)

HasWord (d1, w) ^! HasWord (d2, w) =>! SimPet (d1, d2)

The second of the four first-order logics is that if the subject of the new complaint d1 is w1, the term of the existing complaint d2 is w2, and w1 and w2 are different, then d1 and d2 are not similar complaints. And fourth, if the main word of new complaint d1 is w and this topic w is not included in existing complaint d2, then d1 and d2 are not similar complaints.

If the above rules are applied to the similarity judgment of the new and existing complaints without the weighting according to the Markov Logic Network applied to each equation, d1 and d2 will be judged to be similar only if all of the rules are satisfied. If none are satisfied, they will not be considered similar.

For example, suppose the new complaint d1 and the existing complaint d2 are similar complaints. It is assumed that the query instance is HasWord (d1, car) and the complaint instance is HasWord (d2, car).

In this case, the third of the previous four rules is true, but the other three are false. And if the weights in the Markov logic network were not applied to each rule, applying the above four rules at the same time would falsely assume that there are false rules and that d1 and d2 are not similar complaints.

However, in the case where the weight of the Markov logic network is applied to each rule, even if all of the plurality of rules are not satisfied, the search unit 30 may determine that the new complaint and the existing complaint are similar complaints. These weights are given for each rule and indicate how close the rule is to the truth.

In this case, the search unit 30 calculates the weights by applying the civil instance and the query instance to a plurality of rules, and determines that the existing civil complaint of the civil instance is similar to the new civil complaint when the result of the operation satisfies the preset criteria. can do.

The operation performed by the search unit 30 may add up only the weights of the rules whose values are true when the civil instance and the query instance are applied, or may calculate the average by adding the weights. In this case, the preset criterion may be to determine whether the sum or average value of the weights is greater than or equal to a predetermined constant.

For example, when the sum of weights is 2.5 and the preset criterion is for determining whether the sum of weights is greater than 2, the search unit 30 is similar to the standard complaint and the new complaint because the sum of the weights satisfies the preset criteria. Will judge.

As another example, when a civil instance and a query instance are applied to determine whether an average value of weights of rules whose value is true is 1 and the preset criterion is that the average value is 1.5 or more, the search unit does not satisfy the preset criterion. (30) would determine that the standard complaint and the new complaint are not similar.

Such an operation may also be to add weights of rules whose value is true and subtract weights of rules whose value is false. In this case, the preset criterion may be to determine whether the value of the operation is greater than or equal to a predetermined constant.

For example, if a civil instance and a query instance are applied to multiple rules, the sum of the weights of the true ones is 5 and the sum of the weights of the false ones is 4, the value of the operation is 5 minus 1 would. If the preset criterion is for determining whether the value of the operation is greater than or equal to zero, the search unit 30 may determine that the reference complaint is similar to the new complaint. The manner of operation and the preset criteria can be freely selected by the user or designer of the present invention.

In the Markov logic network, such weights may be learned and determined based on a plurality of complaint instances and similarities between existing complaints included in the complaint instances.

This learning can be achieved by applying appropriate local search-based algorithms, such as Monte-Carlo maximum likelihood estimation or limited-memory Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithms. The manner used for learning in such Markov logic networks can be freely determined by the choice of the user or designer of the present invention.

The result processor 40 may output a similar complaint determined to be similar to the new complaint using the result instance and transfer it to the service providing interface 2. In addition, the result processing unit 40 may output the additional data using the result instance and the additional instance and transfer it to the service providing interface 2.

A supplementary instance is an instance that expresses the relationship between supplementary data related to an existing complaint and an existing complaint in a triple form by ontology. Additional instances may also be instances expressing the relationship between the terms included in an existing complaint and additional data. These additional instances are stored in the instance storage unit 50. An example of additional data ontology that defines the relationship between the additional data on the air-environment and the complaint in relation to this additional instance is shown in FIG. 3.

The supplementary data is related to the existing complaint and may be related to the existing complaint, the relevant affairs, the relevant law, the area concerned, the type of complaint concerned, the existing complaint and / or the person in charge. 3, an example of such an additional instance may be as follows when the existing complaint, which is determined to be similar to the new complaint, is Civil Complaint 2.

Related Institutions (Complaint 2, Ministry of Environment)

Related Institutions (Citizenship 2, Seoul Micro Alarm Center)

Related Work (Complaint 2, Prior Notification and Administrative Disposal for Business Sites that Generate Dust)

Related Laws (Citizenship 2, Enforcement Rules of the Air Quality Preservation Act)

Related complaint types (Complaint 2, Fugitive Dust Reporting Business Report)

This means that the related organizations of Civil Complaint 2 are the Ministry of Environment and the Seoul Fine Warning Center, the related work is pre-reporting and administrative disposition of scattering dust generating workplaces, the related laws are the enforcement regulations of the Air Environment Conservation Act, and the related types of complaints are reporting scattering dust generating business.

If the additional instances in the above example are stored in the instance storage unit 50, the result processing unit 40 is the homepage address of Civil Complaint 2 and the Ministry of Environment and the Seoul Fine Warning Center, preliminary report of scattering dust generation workplace, and the Administrative Disposal or Atmospheric Environment Conservation Act. The additional data such as the contents of the enforcement regulations and the complaint form of the scattering dust generating business report are output and transmitted to the service providing interface (2).

The instance storage unit 50 stores a complaint instance, a term instance and a related instance as part or all of a storage device or a database capable of storing various types of information and / or data.

The inference unit 60 generates a term instance by inferring a relationship between terms included in an existing complaint according to a pre-defined inference rule, and creates an additional instance by inferring a relationship between the existing complaint and additional data to generate a term instance and / or The additional instance is stored in the instance storage unit 50. The inference rule relates to the relationship between terms or defines existing relationships or relationships between terms and additional data, and is preferably stored in the inference rule storage unit 70 in advance.

For example, if the inference rule for the term is "floating dust is a sub-word of dust." The inference unit 60 generates the term SubClassOf (floating dust, dust) and stores it in the instance storage unit 50. . In addition, when the reasoning rule for the relation between the term and the additional data is "the dust related organization is the environment department", the inference unit 60 creates an additional instance called the related organization (dust, environment department) and stores it in the instance storage unit 50. Save it.

Such inference rules may be determined, modified, or changed according to a user's or designer's selection of the present invention and stored in the inference rule storage unit 70. In such a case, the inference unit 60 may be one of the existing terms and additional instances. It is possible to create term instances and supplementary instances that meet the inference rules, or delete, modify, or change existing term instances and supplementary instances that do not conform to the inference rules.

4 is a flowchart illustrating a method of searching for similar complaints based on ontology according to an embodiment of the present invention. In the following, duplicate descriptions will be omitted as necessary.

The extraction unit 10 receives the search data for the new complaint and extracts the main word from the search data (S41). The generation unit 20 generates a query instance expressing the new complaint including the main word in a triple form by the ontology (S42).

The search unit 30 determines whether the new complaint of the query instance and the existing complaint of the complaint instance are similar by using a similar decision rule and searches whether the complaint instance for the existing complaint similar to the new complaint exists in the instance storage unit 50 ( S43).

Complaint instance is an instance expressing the relationship between the existing complaint and the term included in the existing complaint in a triple form by the ontology is preferably stored in the instance storage unit 50 in advance.

The similarity judgment rule refers to a rule that determines the similarity between a new complaint and an existing complaint by comparing the terms included in the query instance with the terms included in the complaint instance. Such a similar decision rule may be expressed as first-order logic based on the relationship between the main word and the term.

The rule of similarity may include a rule that an existing complaint that contains the same terms as the subject is similar to a new complaint. Similar rules may also include a rule that existing complaints that contain terms that are superficial and / or similar to the subject are similar to the new complaints.

The search unit 30 further searches the complaint instance for the existing complaint similar to the new complaint using the term instance. In this case, the term instance is an instance expressed in the triple form by the ontology related to the hierarchical and / or similar relationship between the main word and the term.

The Markov logic network may be applied to the search of the present invention. In this case, the similar decision rule includes a plurality of rules to which the weight is applied according to the Markov logic network. The search unit 30 calculates weights by applying the complaint instance and the query instance to the plurality of rules. If the result of the operation satisfies the preset criteria, the search unit 30 determines that the existing complaint of the civil instance is similar to the new complaint.

If a civil instance is found for an existing civil complaint that is determined to be similar to the new civil complaint, the search unit 30 generates a result instance (S44). The resulting instance is an instance that expresses the similarity between new and existing complaints in triple form by ontology.

In this case, the result processor 40 outputs the existing complaint similar to the new complaint and / or additional data associated with the existing complaint using the additional instance and the result instance, and transmits the additional data to the service providing interface 2 (S45). An additional instance is an instance that expresses a relationship between an existing complaint and an additional data associated with the existing complaint.

If the search result of the search unit 30 does not exist in the instance storage unit 50 for an existing complaint similar to the new complaint of the query instance, the result processing unit 40 outputs a result message indicating the absence of the service. The transfer is sent to the providing interface 2 (S46).

The method of the present invention can also be embodied in computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like, which are also implemented in the form of a carrier wave (for example, transmission over the Internet). It also includes. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

The technical scope of the present invention described above through the embodiments is not limited to the above-described embodiments, and various modifications and changes may be made without departing from the spirit and scope of the present invention. It is evident to those who have knowledge. Therefore, such modifications or variations will have to be belong to the scope of the invention described in the claims of the present invention.

1 is a block diagram showing the configuration of a similar complaint search system based on the ontology according to an embodiment of the present invention.

2 is a diagram illustrating a term ontology defining a relationship between terms according to an exemplary embodiment of the present invention.

3 is a block diagram of an additional data ontology defining a relationship between a complaint and additional data according to an embodiment of the present invention.

4 is a flowchart illustrating a method of searching for similar complaints based on the ontology according to an embodiment of the present invention.

Claims (11)

An extraction unit configured to receive search data for a new complaint and extract a main word from the search data; A generation unit for generating a query instance expressing that the new complaint includes the main word in a triple form by an ontology; An instance storage unit for storing the existing civil complaints and the civil complaint instances expressing the inclusion relationship of the terms included in the existing civil complaints in triple form by the ontology; And Search for complaint instances for existing complaints that are similar to the search data by using similarity determination rules to compare the main term and the term to determine the similarity between the new complaints and the existing complaints. Includes; Search unit for generating a result instance representing the similarity of the; The instance storage unit further includes an additional instance representing a relationship between the existing complaint and the additional data associated with the existing complaint. And a result processing unit for outputting additional data associated with the existing civil complaint similar to the new civil complaint using the result instance and the additional instance. The ontology-based similar complaint search system according to claim 1, wherein the similar decision rule includes a rule that an existing complaint including the same term as the main word is similar to the new complaint. The method of claim 1, wherein the instance storage unit further comprises a term instance related to the vertical relationship between the main word and the term, The rule of similarity includes a rule that an existing complaint including terms that are related to the subject is similar to the new complaint. The search unit is a similar complaint search system based on the ontology to search for a complaint instance for the existing complaint similar to the new complaint using the term instance. The method of claim 1, wherein the similar decision rule includes a plurality of rules to which weights according to the Markov logic network are applied. The search unit calculates weights by applying the complaint instance and the query instance to the plurality of rules, and determines that the existing complaint of the complaint instance is similar to the new complaint when the result of the operation satisfies a preset criterion. Similar complaint search system based on ontology. delete Receiving a search data for a new complaint and extracting a main word from the search data; Generating a query instance expressing that the new complaint includes the main word in a triple form by ontology; The inclusion of the term and the term in a triple form by ontology using a similar decision rule that compares the subject word and the term included in the existing complaint, and determines the similarity between the new complaint and the existing complaint. Retrieving a complaint instance for an existing complaint similar to the new complaint among the expressing complaint instances; Generating a result instance representing a similar relationship between the new complaint and an existing complaint of the retrieved complaint instance; And Outputting additional data associated with the existing complaint similar to the new complaint using the additional instance expressing the relationship between the existing complaint and the additional data associated with the existing complaint, and the resultant instance. Way. The method of claim 6, wherein the similarity determination rule includes a rule that an existing complaint including the same term as the main word is similar to the new complaint. The method of claim 6, wherein the rule of similarity includes a rule that an existing complaint including a term that is related to the subject is similar to the new complaint. Retrieving a civil complaint instance for an existing civil complaint similar to the new civil complaint may include searching for a civil complaint instance for an existing civil complaint similar to the new civil complaint using a term instance related to the subject and the relationship between the term and the term. Search method. 7. The method of claim 6, wherein the similar decision rule includes a plurality of rules to which weights according to the Markov logic network are applied. Retrieving a civil complaint instance for an existing civil complaint similar to the new civil complaint may be performed by applying the civil complaint instance and the query instance to the plurality of rules to calculate weights and satisfying a predetermined criterion. An ontology-based similar complaint search method that determines that the existing complaint of the complaint instance is similar to the new complaint. delete A computer-readable recording medium having recorded thereon a program for realizing the method for searching similarity claim according to any one of claims 6 to 9.

Images