KR20190051574A - Device and method for providing nationality information of user name using neural networks - Google Patents

Abstract

The present invention relates to a device for providing nationality information of a name by using an artificial neural network, which is capable of improving learning and classification performance in a process of estimating and providing the nationality information, and a method thereof. According to one embodiment of the present invention, the device for providing nationality information based on an artificial neural network comprises: a memory storing a nationality information providing program; a data input/output module performing data input/output; and a processor. The nationality information providing program is executed by the processor to estimate the nationality of a person based on a name of the person inputted through the data input/output module. The nationality information providing program outputs the nationality with the highest probability to classify the inputted name of the person thereinto as an output value based on a recurrent neural network-long short term memory (RNN-LSTM)-based learning model. Moreover, the nationality information providing program performs a unigram embedding process of decomposing and vectorizing every character included in the name of the person in accordance with input order, a bigram embedding process of decomposing and vectorizing every two characters included in the name of a person in accordance with the input order, and a trigram embedding process of decomposing and vectorizing every three characters included in the name of the person in accordance with the input order, transfers a processing result to an LSTM memory of the RNN-LSTM, combines vector values outputted from each LSTM memory, and transfers the combined vector value to a hidden layer.

Description

TECHNICAL FIELD [0001] The present invention relates to a device and method for providing nationality information on a name using an artificial neural network,

The present invention relates to a national classification method and apparatus capable of predicting the nationality of a person from a person's name using a neural network structure.

Recently, various kinds of Internet services are provided through applications such as websites and smart terminals, and companies are actively trying to provide optimal services for each user through big data and the like. Particularly, if the user's nationality can be predicted only by the name without requesting additional information to the user, a customized service more suitable for individual users can be provided.

In general, the name of a person has a pattern that is frequently used for that country, so it is possible to identify or predict a rough nationality by name alone. For example, suffixes such as "- sky" and "- v (v)" frequently appear in the case of Russians, while "mann", "- urge" "And so on, the approximate nationality can be predicted based on this.

As a major factor for predicting this nationality, the first is that the language of the country is the most important factor to consider, the second is that the geographical location can provide useful information, and the third is that the history of each country has a crucial role It is possible to do.

There are various considerations, and it is difficult to predict the nationality by its name alone. In the present invention, the technical difficulty is solved through a new platform based on a recurrent neural network.

Chinese Patent Application No. 106600283 (Method and system for identifying the nationalities as well as method for determining transaction risk)

An embodiment of the present invention is to solve the problems of the prior art described above and it is an object of the present invention to provide a national information providing apparatus and method capable of improving learning and classification performance in the process of predicting and providing name nationality information based on a circular neural network .

It should be understood, however, that the technical scope of the present invention is not limited to the above-described technical problems, and other technical problems may exist.

According to an aspect of the present invention, there is provided an apparatus for providing nationality information based on an artificial neural network, including a memory storing a national information providing program, a data input / output module for performing data input / . At this time, the nationality information providing program is driven by the processor, and predicts the nationality of the subject based on the name of the subject inputted through the data input / output module. The national information providing program is a Recurrent Neural Network-Long (RNN-LSTM) A short-term memory), which provides the nationality having the highest probability of classification of the name of the input subject, as an output value, and a unit gram, which decomposes the name of the subject in one- (Unicam) embedding process, a bigram embedding process in which two characters are decomposed and vectorized according to an input sequence, and a trigram embedding process in which three characters are decomposed and vectorized according to an input sequence, -LSTM memory, and then combines the vector values output from each LSTM memory to generate And transmits them to all layers.

According to another aspect of the present invention, there is provided a method of providing nationality information based on an artificial neural network, comprising: receiving name data to be classified; Embedding the received name data into a character; Transferring the character-embedded data to an LSTM memory of a RNN-LSTM (Recurrent Neural Network-Long Short Term Memory); Combining vector values output from the LSTM memory and transmitting the vector values to a hidden layer; And predicting nationality information on the name data according to the probability value calculated through the hidden layer. At this time, the step of embedding the characters may include a unigram embedding process of decomposing and naming the names of the subjects by one character according to the input order, a biram embedding process of decomposing and vectorizing the two characters according to the input sequence ), And trigram embedding in which three characters are decomposed and vectorized according to the input order.

According to any one of the above-mentioned objects of the present invention, the performance of an artificial neural network that predicts nationality from a person's name can be further improved. In particular, learning and classification performance can be improved by performing character embedding such as unigram embedding, bi-gram embedding, and trigram embedding on a name, and inputting the characters into the RNN-LSTM structured neural network.

1 is a configuration diagram of a nationality information providing apparatus according to an embodiment of the present invention.
2 is a view for explaining a configuration of an artificial neural network model according to an embodiment of the present invention.
3 is a flowchart illustrating a method of providing nationality information based on a name according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, which will be readily apparent to those skilled in the art. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In order to clearly illustrate the present invention, parts not related to the description in the drawings are omitted, and like parts are denoted by similar reference numerals throughout the specification. In the following description with reference to the drawings, the same reference numerals will be used to designate the same names, and the reference numerals are merely for convenience of description, and the concepts, features, and functions Or the effect is not limited to interpretation.

Throughout the specification, when a part is referred to as being "connected" to another part, it includes not only "directly connected" but also "electrically connected" with another part in between . Also, when a component is referred to as " comprising ", it is understood that it may include other components as well as other components, But do not preclude the presence or addition of a feature, a number, a step, an operation, an element, a component, or a combination thereof.

Herein, the term " part " or " module " means a unit realized by hardware or software, a unit realized by using both, and a unit realized by using two or more hardware Or two or more units may be realized by one hardware.

1 is a configuration diagram of a nationality information providing apparatus according to an embodiment of the present invention.

The apparatus 100 for providing nationality information shown in FIG. 1 includes a data input / output module 110, a memory 120, and a processor 130.

The data input / output module 110 may perform a function of receiving or transmitting data through a communication module. The data input / output module 110 may be a device including hardware and software necessary to transmit and receive signals such as a control signal or a data signal through a wired or wireless connection with other network devices.

The memory 120 stores a program for predicting the nationality of the user from the name information of the user. The program is implemented based on an artificial neural network that collects and learns user name information and performs classification processing for predicting the nationality information of the inputted user name, and is executed by the processor 130, Function and so on.

This memory 120 is collectively referred to as a non-volatile storage device that keeps stored information even when power is not supplied, or a volatile storage device that requires power to maintain stored information.

2 is a view for explaining a configuration of an artificial neural network model according to an embodiment of the present invention.

The present invention provides a nationality confirmation service based on a known RNN-LSTM (Recurrent Neural Network-Long Short Term Memory). RNN is known as a neural network structure that can be used to share previous information, and is suitable for learning time series data. However, if the time gap between each data becomes longer, The RNN-LSTM technique has been developed to improve this problem. The RNN-LSTM is known to solve the problem of vanishing or exploding a gradient in existing RNN techniques and to better convey the error value in the backpropagation process. LSTM is mainly used to learn the representation of long-lived, clock-chronologically connected data. It is replaced by a memory block consisting of three gates: Input Gate, Output Gate and Forget Gate. . The LSTM is known to be able to learn the representation of data longer than the existing RNN by utilizing memory c and state h. The details of the LSTM block and its operation are the same as those known in the prior art, Is omitted.

In the present invention, there is a technical feature that character embedding is performed with respect to the input name, and the character is input to each LSTM module.

In other words, unigram embedding 202 for decomposing and delivering the input names one by one according to the input order, Bigram embedding 204 for decomposing and delivering two characters according to the input order, And trigram embedding 206 for decomposing and delivering three characters in accordance with the number of characters.

Given an input name X = [x_0, x_1, ..., x_ (n-1)], x_k means each letter of the name. Where x_k may be a unigram, a biagram or a trigram. For example, when the input data is "James", X can be expressed as [J, a, m, e, s] according to the unigram embedding process, es], and X can be expressed as [Jam, ame, mes] according to the trigram embedding process. Thus, in order to enhance the expressiveness of a uniagram, a biagram or a trigram, it is possible to express the vector with a lower dimension, which is called character embedding or distributed representation. In the present invention, the input name data is decomposed into a unigram, a biagram, and a trigram, respectively, and unigram embedding, a biagram embedding, and a trigram embedding processing for expressing the data as a vector are performed. Each of the vectors embedded in the character is processed in three states (h) via a separate LSTM. In the present invention, each state h is concatenated (208) into one long vector, Through the maxma layer, a probability value (Equation 1) in which the input data belongs to the classification of a specific country is calculated.

On the other hand, classification models that classify names according to nationality are learned to maximize the conditional probability according to the following equation.

[Equation 1]

는 해당 이름의 국적(또는 인종) 정보이며, D( )는 θ라는 학습 파라메터를 갖고 있는 분류 모델이며, g( )는 특징(feature)을 추출하기 위한 함수이다. 이때, 국적은 127개 정도의 경우의 수가 존재하지만 인종은 13 개 정도로 그 수가 현저히 줄어들기 때문에, 국적을 정확하게 분류하는 것이 더욱 어려운 문제에 해당한다. here

Is a nationality (or race) information of the name, D () is a classification model with a learning parameter θ, and g () is a function for extracting features. At this time, the number of nationalities is about 127, but since the number of races is about 13, the number of nationalities is more difficult to classify correctly.

More specifically, according to the RNN scheme, the hidden state (h) can be defined as Equation (2).

&Quot; (2) "

는 l 번째 레이어에서의 시간 구간 t 에서의 RNN의 히든 상태를 나타내는 것이다.

,

및

는 학습해야할 파라미터를 나타내고,

는 기본 시그모이드(sigmoid) 함수를 나타낸다. 또한, 각 국적

클래스에 대한 확은은 아래 수학식 3과 같이 정의될 수 있다.

Represents the hidden state of the RNN at the time interval t in the lth layer.

,

And

Indicates a parameter to be learned,

Represents a basic sigmoid function. In addition,

The magnitude of the class can be defined as shown in Equation 3 below.

&Quot; (3) "

와

는 훈련가능한 파라미터들이다. 이러한 RNN 기반의 학습모델에 앞서 소개한 LSTM 모델을 병합하면 다음 수학식 4와 같이 구현될 수 있다.At this time,

Wow

Are trainable parameters. If the LSTM model introduced before the RNN-based learning model is merged, it can be implemented as the following Equation (4).

&Quot; (4) "

와

는 훈련가능한 파라미터들이고,

는 행렬 요소별(element wise) 곱을 나타낸다. 또한, 장기 메모리(Long Term memory)는 입력 게이트(input gate) i 와 이전 셀

이 저장된

로부터 복원될 수 있고, 포겟 게이트 f 에 업데이트될 수 있다.

Wow

Are trainable parameters,

Represents an element wise product. In addition, a long term memory includes an input gate i,

This stored

And can be updated to forget gate f.

Finally, the probability obtained through the above procedure is used to calculate the negative log likelihood (NLL) through the cross entropy technique. Learning the classification model to minimize this NLL It is possible to learn a model that maximizes the probability.

를 통해 해당 확률이 최대화도록 한다.For example, the cross entropy loss function < RTI ID = 0.0 >

So that the probability is maximized.

3 is a flowchart illustrating a method of providing nationality information based on a name according to an embodiment of the present invention.

First, name information for confirming nationality information is received as input data (S310). The input data is received and stored through the data input / output module 110 of the national information providing apparatus 100 of FIG.

Next, character embedding is performed on the input name data (S320).

In the present invention, unigram embedding is performed by decomposing and vectorizing one character according to the input order, Bigram embedding processing for decomposing and vectorizing two characters according to the input order, and decomposing the three characters according to the input order And performs trigram embedding processing for vectorization.

Next, the character-embedded data is input to the LSTM, and the values output from the LSTM are concatenated to generate a single vector (S330).

The generated vector is transmitted to the hidden printer, and a probability value indicating which country is most suitable is calculated based on the generated vector (S340).

As described above, character embedding is performed in several stages, so that the RNN-LSTM learning model can be designed more precisely, and the accuracy of the service for estimating the nationality by only the name of a person can be further improved.

The apparatus and method for providing nationality information according to the embodiments of the present invention described above can be implemented in the form of a recording medium including instructions executable by a computer such as a program module executed by a computer. Such a recording medium includes a computer-readable medium, which may be any available medium that can be accessed by a computer, and includes both volatile and non-volatile media, both removable and non-removable media. The computer-readable medium may also include computer storage media, which may be volatile and non-volatile, implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data , Both removable and non-removable media.

It will be understood by those skilled in the art that the foregoing description of the present invention is for illustrative purposes only and that those of ordinary skill in the art can readily understand that various changes and modifications may be made without departing from the spirit or essential characteristics of the present invention. will be. It is therefore to be understood that the above-described embodiments are illustrative in all aspects and not restrictive. For example, each component described as a single entity may be distributed and implemented, and components described as being distributed may also be implemented in a combined form.

Furthermore, while the methods and systems of the present invention have been described in terms of specific embodiments, some or all of those elements or operations may be implemented using a computer system having a general purpose hardware architecture.

It is intended that the present invention covers the modifications and variations of this invention provided they come within the scope of the appended claims and their equivalents. .

100: Nationality information providing apparatus 110: Data input / output module
120: memory 30: processor

Claims (4)

An apparatus for providing nationality information based on an artificial neural network,
A memory in which the nationality information providing program is stored,
A data input / output module for performing data input / output and
≪ / RTI >
Wherein the nationality information providing program is driven by the processor and predicts the nationality of the subject based on the name of the subject inputted through the data input / output module, wherein the nationality information providing program is an RNN-LSTM (Recurrent Neural Network-Long Term memory is based on a learning model based on a term memory,
A unigram embedding process for decomposing and naming the names of the subjects one by one according to the input order, a bigram embedding process for decomposing and vectorizing the two characters according to the input sequence, and a three- Tristimulation embedding processing for decomposing and vectorizing the LSTM memories, and transferring the vector values output from the LSTM memories to the hidden layer after combining the vector values outputted from the LSTM memories. The method according to claim 1,
The processor combines the value output from the LSTM memory after the uniagram embedding process, the value output from the LSTM memory after the bi-gram embedding process, and the value output from the LSTM memory after the trigram embedding process And the hidden layer calculates a conditional probability that the vector value is classified into a predetermined nationality based on a classification model for classifying names according to nationality, and the calculated conditional probability is a maximum And the learning operation is performed so that the learning operation is performed. A method for providing nationality information based on an artificial neural network,
Receiving name data to be classified;
Embedding the received name data into a character;
Transferring the character-embedded data to an LSTM memory of a RNN-LSTM (Recurrent Neural Network-Long Short Term Memory);
Combining vector values output from the LSTM memory and transmitting the vector values to a hidden layer; And
And predicting nationality information on the name data according to a probability value calculated through the hidden layer,
Wherein the character embedding comprises:
A unigram embedding process for decomposing and naming the names of the subjects one by one according to the input order, a bigram embedding process for decomposing and vectorizing the two characters according to the input sequence, and a three- And performing trigram embedding processing for decomposing and vectorizing the localized information. The method of claim 3,
And performing a learning operation such that the calculated probability value is maximized.

Images