KR20200103164A - A method of classificating semantic relations between sentences for providing conversation services - Google Patents

Abstract

The present invention relates to a sentence semantic relation classification method for providing a chat service. The sentence semantic relation classification method for providing a chat service comprises the following steps of: receiving training document data to be trained; performing word vector conversion processing of mapping words of training documents onto a vector space; performing sentence vector conversion processing corresponding to sentences of the training documents by using word vectors; performing semantic relation classification training between sentences according to the degree of relations between sentence vectors; generating a sentence-based classifier in which semantic relations are trained according to the semantic relation classification training; and performing intention classification processing of a user input document based on the generated classifier. Therefore, the performance of the classifier is improved and semantic relations are easily analyzed.

Description

A method of classifying sentence semantic relations for providing conversation services {A METHOD OF CLASSIFICATING SEMANTIC RELATIONS BETWEEN SENTENCES FOR PROVIDING CONVERSATION SERVICES}

The present invention relates to a sentence semantic relationship classification method for providing a conversation service.

Artificial Intelligence is revolutionizing the way business, organizational operations, lifestyles, and communication are done. Various informatization projects are in progress to provide the best service to the rapidly changing modern culture lifestyle and various constantly changing customer requirements.

Among them, with the recent rapid development of big data and deep learning related technologies, artificial intelligence technologies that are applied in real life in specific fields have been implemented, and analysis of specific data and information in various fields specialized to individuals are integrated and provided. It is also being applied to intelligent personal services.

In order to provide various user services such as artificial intelligence and data analysis services as described above, data mining for a very large amount of text and other language-based information is required, and for this purpose, machine learning is mainly used.

Machine learning of such a document is mainly used in a classification method using a document vector representing only sentences as a vector value embedding words in the current document.

However, the existing classification method has a problem in that it is difficult to grasp the meaning relationship between sentences. This problem consequently shows a limit to the performance of the classifier.

For example, if there is a class called'Europe' and a class called'Soccer' with different subjects, and there is a content related to'European football' in the document data of'Europe', the sentence related to'Europe' class and'Soccer' It can be said that the sentences of the and'football' class have a semantic relationship, but the current classifier does not process any information related to this.

In other words, the existing methods learn and classify data for each sentence class, so even if there is any semantic relationship between sentences of different classes, that is, sentences with different subjects in the document, the current method is The meaning relationship cannot be grasped.

This problem leads to a decrease in class classification for new sentences that are not pre-categorized in conversation services such as machine learning-based chatbots and counseling services, and as a result, it does not lead to normal conversations, and deteriorates service quality.

In addition, since sentence vectors and classification processes are currently integrated, it is difficult to variably apply a classification model even if the domain of a document is changed, and thus classification efficiency is also degraded.

(001) Korean Patent Laid-Open Publication No. 10-2006-0043333 (2006.05.15.) (Name: A system and method for determining the intention of data and responding to data based on the intention)

The present invention was conceived to solve the above-described problem, and provides a dialogue service that infers the relationship between sentences and determines the class by grasping the meaning of a new sentence, thereby improving classifier performance and facilitating semantic relationship interpretation. Its purpose is to provide a method for classifying sentence semantic relations.

The sentence semantic relation classification method according to an embodiment of the present invention for solving the above-described problems is a sentence semantic relation classification method for providing a user dialogue service based on data processing using a semantic relation model between sentences. Receiving document data; Performing a word vector conversion process of mapping words of the learning document onto a vector space; Performing sentence vector conversion processing corresponding to sentences of the learning document by using the word vector; Performing semantic relationship classification learning between sentences according to the degree of relationship between sentence vectors; Generating a sentence-based classifier whose semantic relationship is learned according to the semantic relationship classification learning; And performing an intention classification process of the user input document based on the generated classifier.

Meanwhile, the method according to an embodiment of the present invention for solving the above-described problems may be implemented with a program for executing the method on a computer and a recording medium in which the program is recorded.

In addition, the classification apparatus according to an embodiment of the present invention for solving the above-described problems is a sentence semantic relation classification apparatus for providing a user dialogue service based on data processing using a semantic relation model between sentences,

A word-vector conversion unit that receives a training document from a training database storing training model information generated by performing a training model generation process, and performs a word vector conversion process for mapping words of the training document onto a vector space; A sentence-vector converting unit for performing sentence vector conversion processing corresponding to sentences of the learning document by using the word vector; And performing semantic relation classification learning between sentences according to the degree of relation between sentence vectors, generating a sentence-based classifier whose semantic relation is learned according to the semantic relation classification learning, and generating a user input document based on the generated classifier. It includes a semantic relationship classification unit that performs intention classification processing.

According to an embodiment of the present invention, words of a learning document are mapped onto a vector space, and sentence vector conversion processing corresponding to sentences of the learning document is performed using the word vector, and the relationship between the sentence vectors is Accordingly, by performing semantic relation classification learning between sentences, a sentence-based classifier in which semantic relations are learned according to the semantic relation classification learning may be generated.

Accordingly, according to an embodiment of the present invention, it is possible to provide a sentence-based classifier in which a semantic relationship is learned that can determine a class by inferring a relationship between sentences to grasp the meaning of a new sentence. Through semantic relation-based sentence classification processing, it is possible to provide a sentence semantic relation classification apparatus and an operation method thereof for improving classifier performance and providing a dialogue service that facilitates semantic relation interpretation.

In addition, the present invention provides individual classification processes for each of the word vectorization, sentence vector, and classification process, so that the classification model can be variably applied even if the domain of the document is changed, thereby improving system applicability and classification efficiency. It has the advantage of being.

1 is a conceptual diagram schematically showing an entire system according to an embodiment of the present invention.
2 is a block diagram illustrating a classification apparatus according to an embodiment of the present invention in more detail.
3 is a flow chart for explaining in more detail a method of operating a classification device according to an embodiment of the present invention.
4 is a diagram for describing vector transformation processing according to an embodiment of the present invention.
5 and 6 are diagrams illustrating a classification process and results according to an embodiment of the present invention.
7 is a flowchart illustrating a service providing process according to an embodiment of the present invention.
8 and 9 are diagrams illustrating result data according to application of a convolutional neural network-based sampling and triplet loss function according to an embodiment of the present invention.

The following content merely illustrates the principles of the present invention. Therefore, those skilled in the art can implement the principles of the present invention and invent various devices included in the concept and scope of the present invention, although not clearly described or illustrated herein. In addition, it is understood that all conditional terms and examples listed in this specification are, in principle, expressly intended only for the purpose of making the concept of the present invention understood, and are not limited to the embodiments and states specifically listed as such. Should be.

In addition, it is to be understood that all detailed descriptions listing specific embodiments as well as principles, aspects and embodiments of the present invention are intended to include structural and functional equivalents of these matters. It should also be understood that these equivalents include not only currently known equivalents, but also equivalents to be developed in the future, that is, all devices invented to perform the same function regardless of structure.

Thus, for example, the block diagrams herein are to be understood as representing a conceptual perspective of exemplary circuits embodying the principles of the invention. Similarly, all flowcharts, state transition diagrams, pseudocodes, etc. are understood to represent various processes performed by a computer or processor, whether or not the computer or processor is clearly depicted and that can be represented substantially in a computer-readable medium. Should be.

In addition, the explicit use of terms presented as processor, control, or similar concepts should not be interpreted exclusively by referring to hardware capable of executing software, and without limitation, digital signal processor (DSP) hardware, ROM for storing software. It should be understood to implicitly include (ROM), RAM, and non-volatile memory. Other commonly used hardware may also be included.

The above-described objects, features, and advantages will become more apparent through the following detailed description in connection with the accompanying drawings, whereby those of ordinary skill in the technical field to which the present invention pertains can easily implement the technical idea of the present invention. There will be. In addition, in describing the present invention, when it is determined that a detailed description of known technologies related to the present invention may unnecessarily obscure the subject matter of the present invention, a detailed description thereof will be omitted.

Hereinafter, a preferred embodiment according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a diagram schematically showing an entire system according to an embodiment of the present invention.

Referring to FIG. 1, a system according to an embodiment of the present invention may include a learning device 100, a service providing device 200, and a user terminal 300, and the learning device 100 includes a learning database 180. ) Or may be connected to the learning database 180.

More specifically, the learning device 100 may generate a document learning model by performing machine learning based on the learning document data stored in the learning database 180, and use the generated learning model. , Classification information corresponding to a user input sentence input from the service providing device 200 may be output, or a response sentence corresponding thereto may be provided to the service providing device 200.

In particular, the learning device 100 according to an embodiment of the present invention can be used to provide a conversation service of the service providing device 200, receives learning document data as a learning target, and converts words of the learning document into a vector space. A word vector mapping process is performed, and sentence vector conversion processing corresponding to sentences of the learning document is performed using the word vector, and semantic relation classification between sentences is learned according to the degree of relationship between the sentence vectors. Then, according to the semantic relationship classification learning, a sentence-based classifier in which the semantic relationship is learned may be generated, and intention classification processing of the user input document based on the generated classifier may be performed.

Further, the learning apparatus 100 generates a sentence vector learning model by performing a learning process based on the convolutional neural network between sentence vectors, and generates a triplet loss between sentence vectors corresponding to the sentence vector learning model. (Triplet-loss) function is processed, and according to the result of the triplet loss function processing, a semantic relationship learning model between sentences based on the degree of relationship between sentence vectors can be output.This is the same classification by inferring the relationship between sentences. By making the sentences of the students learn to be close to each other. The output layer of the learning apparatus 100 may include a sentence vector representing a relationship between sentences after all of the relationship inference learning between sentences is completed.

In addition, the service providing apparatus 200 may provide a data processing-based user conversation service using a semantic relationship model between sentences processed as described above to the user terminal 300. The user conversation service uses the learning database 180 generated by the operation of the learning device 100 according to an embodiment of the present invention and a sentence semantic relationship model according to the conversation in response to the input sentence of the user terminal 300. May include services.

For example, the service providing device 200 may provide a chatbot service, a counseling service, and the like to the user terminal 300.

For example, since the user's expression method is always different, if an unclassified sentence is input, there may be a limit on the response. For this, deep learning technology is being used in chatbots, but there is a problem that learning data is significantly less to handle responses to all inquiries, and the data is not classified to use existing counseling data, and it takes a long time for humans to classify. There is a problem to be used.

Accordingly, the learning apparatus 100 according to an embodiment of the present invention has an advantage of being able to perform effective classification processing of similar sentences according to a semantic relationship model between sentences even if sentences that are not classified are input.

Meanwhile, the user terminal 300 and the service providing device 200 may be wired or wirelessly connected through a network, and each user terminal 100 and the service providing device 300 may be connected to an Internet network, a LAN , WAN, Public Switched Telephone Network (PSTN), Public Switched Data Network (PSDN), cable TV network, WIFI, mobile communication network, and other wireless communication networks. In addition, each user terminal 300 and the service providing apparatus 200 may include respective communication modules for communicating through a protocol corresponding to each communication network.

In addition, the user terminal 100 described in the present specification includes a mobile phone, a smart phone, a laptop computer, a terminal for digital broadcasting, a personal digital assistant (PDA), a portable multimedia player (PMP), and a navigation system. Although it may be included, the present invention is not limited thereto, and may be various devices capable of user input and information display.

In such a system, the user terminal 300 may be connected to the service providing device 200 to receive the above-described data-based conversation service.

2 is a block diagram illustrating a learning device according to an embodiment of the present invention in more detail.

Referring to FIG. 2, a learning device 100 according to an embodiment of the present invention may include an input unit 110, a learning unit 120, a classification device 150, and a learning database 180, and a classification device 150 includes a word-vector conversion unit 151, a sentence-vector conversion unit 152, a semantic relation classification unit 153, and a result output unit 154.

First, the input unit 110 receives a learning document for learning from the learning database 180 and delivers it to the learning unit 120 to be used for learning, or the user input document transmitted from the service providing device 200. The result data that has been classified by passing it to the classification device 150 may be processed to be output from the result output unit 154. The result data of the result output unit 154 may be transmitted back to the service providing device 200.

Meanwhile, the learning unit 120 may perform a process of generating a learning model based on the operation of the classification device 150 and learning data input from the input unit 110. The learning model information generated as described above may be stored in the learning database 180.

In particular, the classification apparatus 150 according to an embodiment of the present invention enables semantic relationship classification of sentences by processing such that a semantic relationship learning model between sentences is generated in generating a learning model.

To this end, the classification device 150 includes a word-vector conversion unit 151 that performs word vector conversion processing for mapping words of the learning document onto a vector space, and a sentence of the learning document using the word vector. The sentence-vector converting unit 152 that performs sentence vector conversion processing corresponding to the sentence vectors and the semantic relationship classification learning between sentences according to the degree of the relationship between the sentence vectors, and the semantic relationship according to the semantic relationship classification learning It may include a semantic relationship classifier 153 that generates a classifier based on the learned sentence and performs intention classification processing of a user input document based on the generated classifier.

In addition, the result output unit 154 may be configured with one or more output modules for outputting the classification processing result.

More specifically, the semantic relationship classification unit 153 generates a sentence vector learning model by performing a learning process based on the convolutional neural network (CNN) between the sentence vectors, and corresponds to the sentence vector learning model. A triplet-loss function between one sentence vector is processed, and a semantic relationship learning model between sentences based on the degree of relationship between the sentence vectors is generated according to the result of the triplet loss function processing.

Here, the sentence-vector conversion unit 152, when the word-vector conversion unit 151 calculates word vector conversion values corresponding to words of the learning document, the word vector conversion processing using the word vector Sentence vectors expressed as vectors can be created. Sentence vectors expressed as word vectors as described above may be used as input data of the convolutional neural network.

Here, in order to increase word recognition capacity, the learning document may be predetermined as a document collected from a predetermined information site, for example, a Wikipedia site. Various vocabularies can be accommodated from these documents, and semantic relation information between sentences expressed as word vectors of various vocabularies can be effectively classified.

Meanwhile, the semantic relation classification unit 153 can obtain a desired range within a sentence by adjusting the size of the filter size of the convolutional neural network, and thus, by stacking layers using filters of different sizes, the sentence It can process gradual sentence classification within a range of words, phrases, and sentences.

In addition, the semantic relation classification unit 153 is based on anchor data according to a classification class first classified according to the learning of the convolutional neural network to apply a triplet-loss function, and sentence vectors are It is classified into positive data of the same class and negative data of a different class.

In addition, the semantic relationship classification unit 153 adjusts the positive data of the same class as the anchor data to be located closer to the Euclidean distance from the anchor data, and the distance to the negative data, which is another class, is further located. Balanced data classification can be processed. Such a triplet loss function may be a loss function that facilitates classification between data clusters.

Accordingly, the semantic relation classification unit 153 according to an embodiment of the present invention samples the anchor, positive and negative data for each classification class according to the subject classification corresponding to the sentence vectors, and between each of the sampling data. The loss value between sampling data can be varied until the Euclidean distance value falls within a certain range, and in order to change the loss value, the neural network weight value corresponding to the sentence vector learning model can be adjusted. I can.

In addition, when the Euclidean distance value between the sampling data is within a certain range, the semantic relationship classifier 153 may output a sentence vector learning model according to the currently adjusted weight value as a semantic relationship learning model between sentences.

In addition, the semantic relationship classification unit 153 may generate a sentence intention classifier using the semantic relationship learning model between sentences, and when the user input document is included in a preset classification criterion, classification corresponding to the classification criterion When processing is performed and the user input document is not included in a preset classification criterion, based on a sentence intention classifier using the semantic relationship learning model between sentences, a similar sentence having a similar intention to the user input document is indexed can do.

The result output unit 154 may output classification information corresponding to the indexed similar sentence, and thus may process a similar sentence to be provided to the user even if the sentence is not in the classification criteria.

3 is a flow chart for explaining in more detail a method of operating a learning device according to an embodiment of the present invention.

Referring to FIG. 3, when a learning document is first input through the input unit 110 (S101), the learning apparatus 100 according to an embodiment of the present invention performs word vector conversion processing corresponding to words of the learning document. Do (S109).

Then, the learning device 100 processes sentence vector transformation using word vectors in the learning document (S111).

Thereafter, the learning apparatus 100 performs a semantic relationship classification learning process between sentences according to sentence vector embedding (S113), and generates a classifier using sentences from which the semantic relationship is learned (S115).

Thereafter, the learning device 100 performs intention classification processing of the user input document using the classifier (S117).

4 is a diagram for describing vector transformation processing according to an embodiment of the present invention.

Referring to FIG. 4, data used for an existing classification criterion-based chatbot may be input as learning document data, and 30 sentences written in accordance with an expected scenario that a user can input may be exemplified. As illustrated in FIG. 4, the learning document is separately stored as a label and content, and may be used as an input value of the word vectorization process based on the semantic relation classification unit 153 of the learning unit 120.

According to the vectorization process, the word-vector conversion unit 151 converts words of each sentence into a word vector and maps them into a vector space, and the sentence-vector conversion unit 152 obtains sentence vector data using the word vector data. can do.

Thereafter, the semantic relationship classification unit 153 may input the sentence vector expressed as the word vector as an input value of the sampling module, and by using the sampled vector data, the above-described convolutional neural network generation and triplet loss function are applied. It can be processed to represent the meaning relationship between sentences.

For example, as shown in FIG. 4, the embedding vector data clustered by sampling may be generated, and the data in the vector space clustered as described above is an embedding vector for function processing by the semantic relation classifier 153 It may be used as a data input value, and the semantic relationship classifier 153 may generate a semantic relationship learning model between sentences by learning the relationship of each sentence using the clustered sentences.

5 and 6 are diagrams illustrating a classification process and results according to an embodiment of the present invention.

As shown in FIG. 5, when the classifier 153 according to the semantic relationship learning model between sentences is generated, the classification device 150 includes a word-vector conversion unit 151 and a sentence-vector conversion unit, as described above. A conversation service based on 152 and the classifier 153 can be provided.

In particular, since the word-vector conversion unit 151, the sentence-vector conversion unit 152, and the classifier 153 according to an embodiment of the present invention can be independently generated, future technology development and domain change are made. Even if it is lost, it can be easily configured to change the classification model, and can be processed so that it can be independently connected with other processes. Accordingly, the classification process according to the embodiment of the present invention can be configured by easily combining with separate service processes.

Sample data used as input data may be data that the user inputs on the chatbot, and may be sentences that are not actually learned and are not defined in the existing classification. Typically, as disclosed at the top of FIG. 6, in the case of the existing classification method, the intention of the sentence may not be understood.

On the other hand, referring to FIG. 5, even if unlearned sample data is input, the classification device 150 can check which sentence has a semantic relationship, and the classification result corresponding to each word and sentence is a predicted value. This can indicate the closest in vector space to the sentence with the highest label 10.

Accordingly, the classification device 150 can output sentences having the label 10 with the highest predicted value through the result output unit 154, which has a meaning as shown in the lower part of FIG. Allows various and rich answers to be made.

7 is a flowchart illustrating a service providing process according to an embodiment of the present invention.

Referring to FIG. 7, the learning apparatus 100 according to an embodiment of the present invention performs sentence learning processing based on an existing rule (existing rule) (S201), but when a user-defined sentence is input through the input unit 110 ( S203), it is determined whether the sentence is based on the rule (S205).

First, in the case of a rule-based sentence, the learning device 100 may respond to the service providing device 200 to output rule-based response content (S207).

However, if the sentence is not a rule-based sentence, the learning device 100 may transmit it to the classification device 150 and process a similar sentence to be inferred through sentence semantic relationship inference (S209), and correspond to the inferred similar sentence. A response may be made to the service providing device 200 so that the response content is output (S211).

As described above, the existing rule-based chatbot service is designed to only respond to user inquiries according to the written scenario, so if an unsaved sentence comes in, the chatbot service does not understand the problem.

In the case of using the learning device 100 capable of classifying sentence semantic relations according to an exemplary embodiment of the present invention, even if a sentence that is not learned or classified and stored is received, it is determined which sentence and semantic relationship are similar through the learned sentence semantic relation classifier. Therefore, you can respond according to the relevant content.

8 and 9 are diagrams illustrating result data according to application of a convolutional neural network-based sampling and triplet loss function according to an embodiment of the present invention.

First, referring to FIG. 8, FIG. 8 is a diagram for explaining an example of a process of sampling each anchor, negative, and positive data to learn a convolutional neural network.

First, referring to Equation 1 below, a convolutional neural network variable for applying the triplet loss function may be set as follows.

Here, assuming that C is a set of all classes present in data, P representing the number of positive data may represent the data set size of an arbitrary class included in C. Also, since Anchor is also the same class as P, it can represent one of the classes that C has, such as P. N, which expresses the number of negative data, may represent the number of data of a class belonging to C that is different from P and has a smaller number of data sets than P.

And, the triplet loss function according to an embodiment of the present invention is to implement a classifier by matching the number of data samples of positive and anchor by N, that is, equal to negative.

For example, referring to FIG. 8, in the drawing, class may indicate the classification of sentences having the same subject. Each S represents a sentence as a vector of word embedding, and the number represents the sentence identification number. As shown in FIG. 8, there may be a total of 3 classes in the data, class A may have a total of 100 sentences, class B may have a total of 120 sentences, and class C may have a total of 30 sentences.

Here, if the positive sentence set corresponding to a specific sentence classification is a sentence included in class A, the negative sentence set has less data than the number of data of class A among the other classes, class B or class C, not class A. It can contain statements of the class Class C.

Accordingly, there is a difference in the number of data in the negative and positive sets, the distance to the anchor can be calculated, and for accurate classification, a balance according to the distance can be adjusted. The criterion for determining the number of sample data for distance calculation may be based on the class with the smallest number of sentences among the classes, and thus 30, which is the number of data in class C, may be the criterion for the number of sample data.

Accordingly, the semantic relationship classification unit 153 may apply the triplet loss function to the sample data for sentence embedding to learn the relationship between the sentences, and learn the semantic relationship of each sentence according to distance calculation and adjustment. It can be processed to generate a learning model that infers. Equation 2 below shows the triplet loss function application equation.

는 anchor 의 임베딩 값,

는 positive 의 임베딩 값,

는 negative 의 임베딩 값이라고 할 수 있다. In Equation 2, f is an embedding function, a is achor, p is positive, and n is negative. therefore

Is the embedding value of anchor,

Is the embedding value of positive,

Can be said to be a negative embedding value.

이 만족하도록 anchor 와 positive 사이의 거리는 작아지게 하고 negative 사이의 거리는 커지게 조절할 수 있다.

는

와

가 0 이 되지 않게 하기 위한 즉 두 개의 값이 너무 가깝지 않게 하는 margin 을 조절하는 하이퍼파라미터이다.And, the semantic relation classification unit 153, using the Euclidean distance, the difference between the anchor and the positive distance and the anchor and the negative distance

To satisfy this, the distance between the anchor and the positive can be reduced and the distance between the negative can be adjusted to be larger.

Is

Wow

This is a hyperparameter that adjusts the margin of not being 0, that is, the two values are not too close.

9 shows a change in size of input data and result data from a sampling process for applying a triplet loss function to a process of adjusting a loss within a certain range according to a convolutional neural network (CNN) and a triplet loss function.

In this way, the semantic relation classifier 153 uses a loss function using a triplet loss to make the loss of data of the same class, Anchor and Positive, smaller to make the distance closer, and increase the loss between the other class, Anchor and Negative. You can adjust the weight of the convolutional neural network to create it. Accordingly, the semantic relationship classifier 153 may generate a vector model classifier in which sentence semantic relationships for several classes are learned.

Meanwhile, the above-described method according to various embodiments of the present invention may be implemented as a program and provided to each server or devices while being stored in various non-transitory computer readable media. Accordingly, the user terminal 100 can access the server or device and download the program.

The non-transitory readable medium refers to a medium that stores data semi-permanently and can be read by a device, not a medium that stores data for a short moment, such as a register, cache, or memory. Specifically, the above-described various applications or programs may be stored and provided in a non-transitory readable medium such as a CD, DVD, hard disk, Blu-ray disk, USB, memory card, ROM, or the like.

In addition, although the preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above, and the technical field to which the present invention belongs without departing from the gist of the present invention claimed in the claims. In addition, various modifications are possible by those of ordinary skill in the art, and these modifications should not be individually understood from the technical spirit or prospect of the present invention.

Claims (7)

As a sentence semantic relationship classification method for receiving a user voice as a user input document through an input unit and providing a user conversation service based on data processing using a semantic relationship model between sentences,
Receiving learning document data to be learned;
Performing a word vector transformation process of mapping words of the learning document onto a vector space;
Performing a sentence vector conversion process corresponding to sentences of the learning document by using the word vector;
Performing semantic relationship classification learning between sentences according to the degree of relationship between sentence vectors;
Generating a sentence-based classifier whose semantic relationship is learned according to the semantic relationship classification learning; And
Including; performing intention classification processing of the user input document based on the generated classifier
A method of classifying sentence semantic relations to provide dialogue service. The method of claim 1,
Determine whether the user input document is a rule-based sentence according to an existing rule-based sentence learning method, and output rule-based response contents in the case of a rule-based sentence.
A method of classifying sentence semantic relations to provide dialogue service. The method of claim 2,
If the user input document is not a rule-based sentence, inferring a similar sentence based on the classification processing information, and outputting a response to the inferred similar sentence
A method of classifying sentence semantic relations to provide dialogue service. The method of claim 1,
The step of performing semantic relationship classification learning between sentences,
Generating a sentence vector learning model by performing a learning process based on a convolutional neural network between sentence vectors; And
Processing a triplet-loss function between sentence vectors corresponding to the sentence vector learning model; And
And outputting a semantic relationship learning model between sentences based on the degree of relationship between sentence vectors according to the triplet loss function processing result.
A method of classifying sentence semantic relations to provide dialogue service. The method of claim 4,
Processing the triplet loss function,
Sampling anchor, positive and negative data for each classification class according to subject classification corresponding to the sentence vectors; And
Comprising the step of varying the loss (loss) value between the sampling data until the Euclidean distance value between each of the sampled data is within a predetermined range
A method of classifying sentence semantic relations to provide dialogue service. The method of claim 5,
Processing a neural network weight value corresponding to the sentence vector learning model to change the loss value;
If the Euclidean distance between the sampling data is within a certain range, outputting a sentence vector learning model according to the currently adjusted weight value as a semantic relationship learning model between sentences;
A method of classifying sentence semantic relations to provide dialogue service. The method of claim 6,
The step of performing the intention classification process of the user input document,
If the user input document is included in a preset classification standard, performing classification processing corresponding to the classification standard; And
If the user input document is not included in a preset classification criterion, indexing a similar sentence having a similar intention to the user input document based on a sentence intention classifier using the semantic relationship learning model between sentences; And
Including the step of outputting classification information corresponding to the indexed similar sentence
A method of classifying sentence semantic relations to provide dialogue service.

Images