KR101985900B1 - A method and computer program for inferring metadata of a text contents creator - Google Patents

Abstract

One embodiment of the present invention relates to a method for inferring metadata of a text content creator. The method may comprise the steps of: receiving text content; dividing the text content into one or more morphemes and determining a morpheme vector from the divided morphemes based on a morpheme-vector converter; dividing the text content into one or more syllables and determining a syllable vector from the divided syllables based on a syllable-vector converter; and determining a metadata vector corresponding to metadata of a creator based on the morpheme vector, the syllable vector, and a creator recognizer.

Description

A METHOD AND COMPUTER PROGRAM FOR INFERRING METADATA OF A TEXT CONTENTS CREATOR}

Embodiments of the present invention relate to a method and a computer program for inferring meta information of a text content creator, and more particularly to a method for inferring meta information of a content creator based on morphemes and syllables of the text content.

Due to the rapid development of information and communication technology, various types of terminals such as mobile communication terminals and personal computers have been implemented to perform various functions.

In particular, these terminals are implemented to use various functions such as generating secondary information from the received information in the conventional function of simply receiving and displaying information. In such an environment, the amount of content posted on the web is increasing exponentially, and thus, it is difficult to grasp the attributes of individual content (for example, the age range and gender of the author).

The present invention seeks to infer the meta information of the author based on the morphological features of the text content.

In addition, the present invention is to infer the meta information of the author based on the syllable characteristics of the text content, that is, the dialogue (or tone) of the text content, and to infer the meta information of the author with higher accuracy.

In addition, the present invention intends to infer the meta information of the author in consideration of both the morphological features and the syllable features of the text contents.

According to an embodiment of the present invention, a method for inferring meta information of a text content creator includes: receiving text content; Dividing the text content into one or more morphemes and determining a morpheme vector from the segmented one or more morphemes based on a morpheme-vector converter; Dividing the text content into one or more syllables and determining a syllable vector from the divided one or more syllables based on a syllable-vector converter; And determining a meta information vector corresponding to the meta information of the writer based on the morpheme vector, the syllable vector, and the writer recognizer. The morpheme-vector converter is a data set representing a correlation between a plurality of morphemes and a plurality of morpheme vectors, and the syllable-vector converter is a data set representing a correlation between a plurality of syllables and a plurality of syllable vectors. May be a data set representing a correlation between a plurality of morpheme vectors, a plurality of syllable vectors, and a plurality of meta information vectors.

The method of inferring meta information of the author may further include generating one or more divided contents by dividing the text contents into predetermined units after receiving the text contents. The determining of the morpheme vector may include determining a morpheme vector for each of the one or more divided contents, and determining the syllable vector for determining a syllable vector for each of the one or more divided contents, and determining the meta information vector. The determining may determine the meta information vector corresponding to the meta information of the creator based on the morpheme vector of each of the one or more pieces of divided content, the syllable vector of each of the one or more pieces of divided content, and the author recognizer.

The method of inferring the meta information of the creator may include determining meta information of the creator for each of the one or more pieces of divided content based on the meta information vector for each of the one or more pieces of divided contents after determining the meta information vector; And determining the meta information of the text content creator based on the meta information of the creator for each of the one or more pieces of divided content. In this case, the predetermined unit may be a sentence unit.

The method of inferring meta information of the creator may further include determining meta information of the text content creator based on the determined meta information vector after determining the meta information vector.

The meta information of the author of the text content may include at least one of the age group of the author, the gender of the author, a region associated with the author, the political propensity of the author, the educational background of the author, and whether the author is married.

The determining of the meta information of the author may include determining the age group of the text content creator in the form of a probability that the creator belongs to a plurality of candidate age groups, and in the form of a probability that the creator is a male and a probability that the creator is a female. The gender of the text content creator can be determined.

The method of inferring meta information of the creator may include: training the morpheme-vector converter based on a plurality of first training data including a first test morpheme and a first morpheme vector corresponding to the first test morpheme; Training the syllable-vector converter based on a plurality of second training data including a first test syllable and a first syllable vector corresponding to the first test syllable; And training the creator recognizer based on a third learning data including a second morpheme vector, a second syllable vector, and a meta information vector corresponding to the second morpheme vector and the second syllable vector. Can be.

The determining of the meta information vector may include generating a content vector by merging the morpheme vector and the syllable vector; And determining the meta information vector based on the content vector and the creator recognizer.

Other aspects, features, and advantages other than those described above will become apparent from the following drawings, claims, and detailed description of the invention.

According to the embodiments of the present invention, the meta information of the author may be inferred based on the morpheme feature of the text content.

In addition, the meta information of the author may be inferred based on the syllable characteristics of the text content, that is, the dialogue (or tone) of the text content, and the meta information of the author may be inferred with higher accuracy.

In addition, the meta information of the author can be inferred by considering both the morphological features and the syllable features of the text contents.

1 schematically illustrates a content creator meta information inference system according to an embodiment of the present invention.
2 schematically illustrates a configuration of an author meta information inference apparatus according to an embodiment of the present invention.
3A to 3C are diagrams for describing a method of generating or learning a morpheme-vector converter, a syllable-vector converter, and a writer recognizer by a writer meta information inference apparatus according to an embodiment of the present invention.
FIG. 4 is a diagram for describing a method of dividing text content including a plurality of sentences into divided content by a controller according to an exemplary embodiment of the present invention.
FIG. 5 is a diagram for describing a method of inferring meta information of an author from text content (or fragmented content) by a control unit according to an embodiment of the present invention.
6 is a flowchart illustrating a method of inferring meta information of a text content creator by an apparatus for inferring meta information of an author according to an embodiment of the present invention.
7 is an example of a screen displayed on the display unit of the user terminal according to an embodiment of the present invention.

As the invention allows for various changes and numerous embodiments, particular embodiments will be illustrated in the drawings and described in detail in the written description. Effects and features of the present invention, and methods of achieving them will be apparent with reference to the embodiments described below in detail together with the drawings. However, the present invention is not limited to the embodiments disclosed below but may be implemented in various forms.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings, and the same or corresponding components will be denoted by the same reference numerals, and redundant description thereof will be omitted. .

In the following embodiments, the terms first, second, etc. are used for the purpose of distinguishing one component from other components rather than a restrictive meaning. In the following examples, the singular forms "a", "an" and "the" include plural forms unless the context clearly indicates otherwise. In the following examples, the terms including or having have meant that there is a feature or component described in the specification and does not preclude the possibility of adding one or more other features or components. In the drawings, components may be exaggerated or reduced in size for convenience of description. For example, the size and shape of each component shown in the drawings are arbitrarily shown for convenience of description, and thus the present invention is not necessarily limited to the illustrated.

1 schematically illustrates a content creator meta information inference system according to an embodiment of the present invention.

Referring to FIG. 1, a content creator meta information inference system according to an embodiment of the present invention may include a server 100, a user terminal 200, an external device 300, and a communication network 400 connecting them. .

In the content creator meta information inference system according to an embodiment of the present invention, the server 100 receives text content from the user terminal 200 and / or the external device 300, and infers meta information of the creator of the received content. can do. In addition, the content creator meta-information inference system, the server 100 may obtain the content that knows the meta information of the author in advance from the user terminal 200 and / or the external device 300, and can learn the author recognizer based on this. . A more detailed description will be described later.

In the present invention, 'the author's meta information' may refer to personal information or personal information about the author such as the age group, gender, residence area, political orientation, etc. of the creator.

In the present invention, 'text content' may mean various contents including text in at least a portion of the content. For example, the text content may mean content including only text. In this case, the server 100 according to an embodiment of the present invention may infer meta information of the author by analyzing the full text content. The text content may mean content that further includes content such as an image or an image in addition to the text. In this case, the server 100 according to an embodiment of the present invention may separate only the text from the content and infer meta information of the content creator from the content. In the following description, the text content includes only text for convenience of explanation.

In the present invention, the user terminal 200 may refer to various devices capable of transmitting and receiving the above-described text content with the server 100. In this case, the terminal may be a personal computer 202 or a portable terminal 201. In FIG. 1, the portable terminal 201 is illustrated as a smart phone, but the inventive concept is not limited thereto.

Meanwhile, the user terminal 200 may include display means for displaying content and input means for obtaining a user input for such content. In this case, the input means and the display means may be configured in various ways. For example, the input means may include a keyboard, a mouse, a trackball, a microphone, a button, a touch panel, and the like, but is not limited thereto.

In the present invention, the external device 300 may refer to various devices for transmitting and receiving data through the server 100 and / or the user terminal 200 and the communication network 400.

According to an embodiment of the present invention, the external device 300 may be a device that provides training data for learning the creator identifier provided in the server 100. For example, the external device 300 is a server that provides content (for example, a newspaper article or a comment on an article) and meta information of the creator of the content (for example, meta information of a reporter who wrote the article or meta information of a user who created a comment). Can be. The external device 300 may be singular or plural.

According to another embodiment of the present invention, the external device 300 transmits identification information on text content to be provided to the user terminal 200 by the external device 300 itself to the server 100, and from the server 100. It may be a device for receiving the meta information of the text content creator. For example, the external device 300 may be a server of a service subject that wants to provide a service for providing meta information on an article for which users want to know the meta information of the author. However, the above two cases are exemplary and the spirit of the present invention is not limited thereto.

In the present invention, the communication network 400 serves to connect the server 100, the user terminal 200 and the external device 300. For example, the communication network 400 provides a connection path for transmitting and receiving packet data after the user terminal 200 accesses the server 100. The communication network 400 may be, for example, wired networks such as local area networks (LANs), wide area networks (WANs), metropolitan area networks (MANs), integrated service digital networks (ISDNs), wireless LANs, CDMA, Bluetooth, satellite communications, and the like. Although it may encompass a wireless network, the scope of the present invention is not limited thereto.

In the present invention, the server 100 may receive text content from the user terminal 200 and / or the external device 300 and infer meta information of the content creator. In addition, the server 100 may obtain one or more learning contents labeled with meta information from the user terminal 200 and / or the external device 300, and train the creator recognizer based thereon. To this end, the server 100 may include a creator meta information inference apparatus as shown in FIG. 2.

2 schematically illustrates a configuration of an author meta information inference apparatus 110 according to an embodiment of the present invention.

2, the creator meta information inference apparatus 110 according to the present embodiment may include a communication unit 111, a control unit 112, and a memory 113. In addition, although not shown in the drawing, the creator meta information inference apparatus 110 according to the present embodiment may further include an input / output unit, a program storage unit, and the like.

The communication unit 111 is necessary for the creator meta information inference apparatus 110 to transmit and receive a signal such as a control signal or a data signal through a wired or wireless connection with another network device such as the user terminal 200 and / or the external device 300. It may be a device including hardware and software.

The controller 112 may include all kinds of devices capable of processing data, such as a processor. Here, 'processor' may refer to a data processing apparatus embedded in hardware having, for example, a physically structured circuit for performing a function represented by code or instructions included in a program. As an example of the data processing unit embedded in the hardware, a microprocessor, a central processing unit (CPU), a processor core, a multiprocessor, and an application-specific integrated (ASIC) Circuits, field programmable gate arrays (FPGAs), and the like, but may include a processing device, but the scope of the present invention is not limited thereto.

The memory 113 performs a function of temporarily or permanently storing data processed by the creator meta information inference apparatus 110. The memory 113 may include a magnetic storage medium or a flash storage medium, but the scope of the present invention is not limited thereto.

Hereinafter, the author meta information inference apparatus 110 will be described on the premise that the server 100 is provided. However, the author meta information inference apparatus 110 may be provided separately from the server 100 according to role allocation.

Meanwhile, as described above, the server 100, that is, the author meta information inference apparatus 110 receives text content from which the meta information of the author is to be inferred from the user terminal 200 and / or the external device 300, and is received. Meta information of the author of the text content can be inferred. In addition, the creator meta information inference apparatus 110 may obtain one or more learning contents labeled with metadata of the creator from the user terminal 200 and / or the external device 300, and train the creator recognizer based on the learning content.

Hereinafter, the author meta information inference apparatus 110 will first describe how to create and / or learn an author recognizer from one or more learning contents, and the creator of the received text content based on the created and / or learned author recognizer. We will explain later how to deduce the meta information of.

In the present invention, the 'writer recognizer' may be a data set representing a correlation between a plurality of morpheme vectors, a plurality of syllable vectors, and a plurality of meta information vectors. That is, the author recognizer may indicate a correspondence relationship between 'plural morpheme vectors and plural syllable vectors' and 'plural meta information vectors'. In this case, the morpheme vector may be a vector generated based on the morpheme analysis of the text content to infer the meta information of the author. In addition, the syllable vector may be a vector generated based on the syllable analysis of the text content to infer the meta information of the author. Meanwhile, the meta information vector is a vector corresponding to the meta information of the creator of the text content and may be a vector including information on various items (eg, age group, gender, residential area, political orientation, etc.).

The 'writer recognizer' may be generated by machine learning based on a plurality of learning data. In this case, the learning data may be received from the above-described external device 300. Of course, the learning data may be stored in the memory 113 of the meta-information inference apparatus 110.

Meanwhile, the creator recognizer may be a data set as described above. For example, the author recognizer may be a data set composed of a plurality of numbers, such as a matrix. However, this is merely exemplary and the spirit of the present invention is not limited thereto.

Similar to the author recognizer described above, the 'morpheme-vector converter' may be a data set representing a correlation between a plurality of stems and a plurality of stem vectors. Also, the syllable-vector converter may be a data set representing a correlation between a plurality of syllables and a plurality of syllable vectors.

Both the morpheme-vector converter and the syllable-vector converter described above may be generated by machine learning based on a plurality of learning data similarly to the author recognizer. For example, a morpheme-vector converter may be generated based on learning data including a morpheme and a morpheme vector corresponding to the morpheme. Similarly, in the case of a syllable-vector converter, the syllable-vector converter may be generated based on learning data including a syllable and a syllable vector corresponding to the syllable.

3A to 3C illustrate that the author meta information inference apparatus 110 generates or learns a morpheme-vector converter 520, a syllable-vector converter 620, and a creator recognizer 720 according to an embodiment of the present invention. It is a figure for demonstrating a method.

First, referring to FIG. 3A, the controller 112 of the author meta information inference apparatus 110 according to an embodiment of the present invention may be configured to the first test morpheme and the first test morpheme for the generation of the morpheme-vector converter 520. The plurality of first training data 500 including the corresponding first morpheme vector may be received and / or obtained. In other words, the controller 112 according to an embodiment of the present invention may acquire learning data including a morpheme and a morpheme vector corresponding to the morpheme. For example, the first learning data 510 may include a morpheme and a morpheme vector 511 corresponding to the morpheme. In this case, the number of morphemes corresponding to any one morpheme vector may be singular or plural.

Subsequently, the controller 112 according to an embodiment of the present invention may train the morpheme-vector converter 520 based on the received and / or obtained first training data 500.

As described above, the morpheme-vector converter 520 may be a correlation between a plurality of morphemes and a plurality of morpheme vectors generated by machine learning, that is, mapping information of both.

Accordingly, the controller 112 according to an embodiment of the present invention may set the data set of the morpheme-vector converter 520 so that the morpheme of the first training data 500 and the morpheme vector corresponding to the morpheme may be mapped to each other. The morpheme-to-vector converter 520 can be trained in such a manner as to update a.

For example, when the data set of the morpheme-vector converter 520 is composed of a plurality of coefficients, the controller 112 corresponds to a morpheme of the first training data 500 and a morpheme vector corresponding to the morpheme. The coefficients constituting the morpheme-vector converter 520 may be adjusted accordingly. In this case, the controller 112 may obtain the morpheme-vector converter 520 having improved accuracy by updating the coefficients based on the plurality of training data.

Accordingly, the present invention can infer the meta information of the author based on the morpheme feature of the text content.

Referring to FIG. 3B, the control unit 112 according to an embodiment of the present invention includes a plurality of first syllable vectors and first syllable vectors corresponding to the first test syllables to generate the syllable-vector converter 620. May receive and / or obtain the second training data 600. In other words, the controller 112 according to an embodiment of the present invention may acquire the second learning data 600 including the syllable and the syllable vector corresponding to the syllable. For example, the first learning data 610 may include syllable vectors and syllable vectors 611 corresponding to syllables. In this case, the number of syllables corresponding to any one syllable vector may be singular or plural.

Subsequently, the controller 112 according to an exemplary embodiment may train the syllable-vector converter 620 based on the received and / or obtained second training data 600.

As described above, the syllable-vector converter 620 may be a correlation between a plurality of syllables and a plurality of syllable vectors generated by machine learning, that is, mapping information of both.

Accordingly, the controller 112 according to an embodiment of the present invention may set the data set of the syllable-vector converter 620 so that the syllable vector of the second training data 600 and the syllable vector corresponding to the syllable are mapped to each other. The syllable-vector converter 620 may be trained in such a manner as to update a.

For example, when the data set of the syllable-vector converter 620 is composed of a plurality of coefficients, the controller 112 maps a syllable vector of the second training data 600 and a syllable vector corresponding to the syllable. Coefficients constituting the syllable-vector converter 620 may be appropriately adjusted. In this case, the controller 112 may obtain the syllable-vector converter 620 having more accuracy by updating the coefficients based on the plurality of learning data.

As a result, the present invention can infer the meta information of the author based on the syllable characteristics of the text content, that is, the dialogue (or tone) of the text content, and can infer the meta information of the author with higher accuracy. have.

Meanwhile, referring to FIG. 3C, the control unit 112 according to an embodiment of the present invention corresponds to the second morpheme vector, the second syllable vector, the second morpheme vector, and the second syllable vector to generate the creator recognizer 720. The plurality of third training data 700 including the meta information vector may be received and / or obtained. In other words, the controller 112 according to an embodiment of the present invention may acquire training data including a morpheme vector and a syllable vector, and a meta information vector corresponding to the morpheme vector and the syllable vector.

For example, the first training data 710 may include a morpheme vector Vm2, a syllable vector Vs2, and a meta information vector 711 corresponding thereto.

Subsequently, the controller 112 according to an embodiment of the present invention may train the creator recognizer 720 based on the received and / or obtained third learning data 700.

As described above, the author recognizer 720 may be a correlation between a plurality of morpheme vectors, a plurality of syllable vectors, and a plurality of meta information vectors generated by a machine learning technique, that is, mapping information of both.

Accordingly, the controller 112 according to an embodiment of the present invention updates the data set of the author recognizer 720 so that the stemming vector, the syllable vector, and the meta information vector of the training data can be mapped to each other (mapping). The creator recognizer 720 can be learned.

For example, when the data set of the creator recognizer 720 is composed of a plurality of coefficients, the controller 112 may correspond to (map) the morpheme vector and syllable vector of the training data and the corresponding meta information vector. Coefficients constituting the author recognizer 720 may be appropriately adjusted. In this case, the controller 112 may obtain the creator recognizer 720 having improved accuracy by updating the coefficients based on the plurality of learning data.

Thus, the present invention can infer the meta information of the author in consideration of both the morphological features and the syllable features of the text content.

By the above-described process, the controller 112 according to an embodiment of the present invention is a morpheme-vector converter 520, a syllable-vector converter 620, and an author recognizer 720 for deducing meta information of an author from text contents. ) And / or learn.

Hereinafter, a method of inferring meta information of a text content creator based on the morpheme-vector converter 520, the syllable-vector converter 620, and the author recognizer 720 generated and / or learned by the above-described process will be described. .

The controller 112 according to an embodiment of the present invention may receive the text content from which the user wants to infer the meta information of the creator from the user terminal 200 and / or the external device 300. In this case, as described above, the text content may include various types of content (for example, an image or an image) in addition to the text content.

Meanwhile, the text content may include only one sentence or may include a plurality of sentences. If the text content includes a plurality of sentences, the controller 112 according to an embodiment of the present invention may infer meta information of the author by dividing the text content into a plurality of contents.

FIG. 4 is a diagram for describing a method of dividing text content 800 including a plurality of sentences into divided contents 810, 820, and 830 by the control unit 112 according to an exemplary embodiment.

For convenience of explanation, it is assumed that the text content 800 includes a plurality of sentences as described above. In this case, the controller 112 according to an embodiment of the present invention may divide the text content 800 into predetermined units to generate one or more divided contents 810, 820, and 830. In this case, the predetermined unit may be a sentence unit, a paragraph unit, or a subject unit. However, this is merely exemplary and the spirit of the present invention is not limited thereto.

Hereinafter, for convenience of description, a description will be given based on a method of inferring the meta information of the creator for one divided content 810.

FIG. 5 is a diagram for describing a method of the controller 112 inferring meta information of an author from text content 810 (or divided content) according to an embodiment of the present invention.

As described above, the controller 112 according to an embodiment of the present invention may receive or acquire text content, and generate one or more divided contents by dividing the text content into predetermined units as necessary.

Subsequently, the controller 112 according to an embodiment of the present invention may divide the text content 810 into one or more morphemes 811. For example, when the text content is 'Give it for the peace of Korea', the controller 112 may divide the content into one or more morphemes such as 'South Korea', 'Peace', 'For the sake of', 'Give it'. . The controller 112 according to an embodiment of the present invention may determine the morpheme vector 821 from one or more morphemes 811 divided based on the morpheme-vector converter 520.

As described above, the morpheme-vector converter 520 may be a correlation between a plurality of morphemes and a plurality of morpheme vectors generated by machine learning, that is, mapping information of both.

Accordingly, the controller 112 may input one or more morphemes 811 into the morpheme-vector converter 520, and as a result, obtain a morpheme vector 821 corresponding to the one or more morphemes 811.

Thus, the present invention can infer the meta information of the author based on the morphological characteristics of the text content.

The controller 112 according to an embodiment of the present invention may divide the text content 810 into one or more syllables 812. For example, as in the above example, if the text content is 'Come for the peace of Korea', the control unit 112 converts the content into one or more syllables such as 'large', 'han', 'min', 'guk', and the like. Can be divided The controller 112 according to an embodiment of the present invention may determine the syllable vector 822 from the divided syllables 812 based on the syllable-vector converter 620.

As described above, the syllable-vector converter 620 may be a correlation between a plurality of syllables and a plurality of syllable vectors generated by machine learning, that is, mapping information of both.

Accordingly, the controller 112 may input one or more syllables 812 to the syllable-vector converter 620, and as a result, obtain a syllable vector 822 corresponding to the one or more syllables 812.

As a result, the present invention can infer the meta information of the author based on the syllable characteristics of the text content, that is, the dialogue (or tone) of the text content, and can infer the meta information of the author with higher accuracy.

The control unit 112 according to an embodiment of the present invention is based on the morpheme vector 821, the syllable vector 822, and the author recognizer 720 determined by the above-described process. 830 may be determined.

Similar to the morpheme-vector converter 520 and syllable-vector converter 620 described above, the author recognizer 720 also includes a plurality of morpheme vectors, a plurality of syllable vectors, and a plurality of meta generated by machine learning techniques. Correlation of information vectors, that is, mapping information of both.

Accordingly, the controller 112 inputs the morpheme vector 821 and the syllable vector 822 to the author recognizer 720, and as a result, the meta information vector 830 corresponding to the morpheme vector 821 and the syllable vector 822. Can be obtained.

Meanwhile, the controller 112 according to an embodiment of the present invention may generate a content vector by merging a morpheme vector and a syllable vector, and determine the meta information vector 830 based on the generated content vector and the author recognizer 720. have. In this case, merging a vector may mean creating a new vector corresponding to the sum of the number of dimensions of the two vectors, and generating a vector of a new dimension (dimensions less than or equal to the sum of the number of dimensions of the two vectors) through a predetermined operation. It might mean doing.

The control unit 112 according to an embodiment of the present invention performs the same series of processes (determining a morpheme vector, a syllable vector, and a meta-information vector) for the divided content generated by dividing the text content into predetermined units. Can be done. In other words, the controller 112 may determine a morpheme vector for each of the one or more pieces of divided content. In addition, the controller 112 may determine a syllable vector for each of the one or more pieces of divided content. In addition, the controller 112 may determine the meta information vector corresponding to the meta information of the creator based on the morpheme vector of each of the one or more pieces of divided content, the syllable vector of each of the one or more pieces of divided content, and the author recognizer. As such, the controller 112 may determine the meta information vector for each divided content. On the other hand, the method of determining the meta information of the author for the entire content from this will be described later.

The controller 112 according to an embodiment of the present invention may determine the meta information 840 of the creator of the text content 810 based on the meta information vector 830 determined by the above-described process. In this case, the meta information of the content creator may include at least one of the age group of the creator, the gender of the creator, a region related to the creator, the political propensity of the creator, the educational background of the creator, and whether the creator is married. Of course, the above-described items are exemplary and may be used as the meta information of the present invention as long as the items can be used as meta information of a person.

The control unit 112 according to an embodiment of the present invention may determine the meta information with respect to each item of the meta information in the form of probability of a plurality of options for each item. For example, the controller 112 may determine the age group of the text content creator in the form of a probability that the creator belongs to a plurality of candidate age groups (for example, teenagers, 20s, 30s, 40s, 50s or more). Similarly, the controller 112 may determine the gender of the creator in the form of a probability that the creator is a male and a probability that the creator is a female. However, the above-mentioned age group and gender items are exemplary, and the spirit of the present invention is not limited thereto.

On the other hand, when the meta-information vector is determined for each of the divided contents, the control unit 112 according to an embodiment of the present invention may determine the meta information about each of the divided contents.

At this time, the control unit 112 according to an embodiment of the present invention may determine the meta information for the full text content by merging the meta information for each divided content. For example, when the probability that the creator belongs to the plurality of candidate ages is determined for each of the divided contents, the controller 112 may determine the total as the sum of the probabilities for each age group (for example, the sum of probabilities belonging to the teenager, the sum of probabilities belonging to the 20s, etc.). Meta information about the text content may be determined. However, this is merely illustrative and the spirit of the present invention is not limited thereto.

Accordingly, the present invention can infer the meta information of the author in consideration of both the morpheme features and the syllable features of the text content.

FIG. 6 is a flowchart illustrating a method of inferring meta information of a text content creator by the creator meta information inference apparatus 110 according to an embodiment of the present invention. Hereinafter, descriptions of contents overlapping with those described in FIGS. 1 to 5 will be omitted.

Author meta-information inference apparatus 110 according to an embodiment of the present invention may generate or learn a morpheme-vector converter, syllable-vector converter, and author recognizer (S61).

In the present invention, the 'writer recognizer' may be a data set representing a correlation between a plurality of morpheme vectors, a plurality of syllable vectors, and a plurality of meta information vectors. That is, the author recognizer may indicate a correspondence relationship between 'plural morpheme vectors and plural syllable vectors' and 'plural meta information vectors'. In this case, the morpheme vector may be a vector generated based on the morpheme analysis of the text content to infer the meta information of the author. The syllable vector may be a vector generated based on the syllable analysis of text content to infer meta information of the author. Meanwhile, the meta information vector is a vector corresponding to the meta information of the creator of the text content and may be a vector including information on various items (eg, age group, gender, residential area, etc.).

The 'writer recognizer' may be generated by machine learning based on a plurality of learning data. In this case, the learning data may be received from the above-described external device 300. Of course, the learning data may be stored in the memory 113 of the meta-information inference apparatus 110.

Meanwhile, the creator recognizer may be a data set as described above. For example, the author recognizer may be a data set composed of a plurality of numbers, such as a matrix. However, this is merely exemplary and the spirit of the present invention is not limited thereto.

Similar to the author recognizer described above, the 'morpheme-vector converter' may be a data set representing a correlation between a plurality of stems and a plurality of stem vectors. Also, the syllable-vector converter may be a data set representing a correlation between a plurality of syllables and a plurality of syllable vectors.

Both the morpheme-vector converter and the syllable-vector converter described above may be generated by machine learning based on a plurality of learning data similarly to the author recognizer. For example, a morpheme-vector converter may be generated based on learning data including a morpheme and a morpheme vector corresponding to the morpheme. Similarly, in the case of a syllable-vector converter, the syllable-vector converter may be generated based on learning data including a syllable and a syllable vector corresponding to the syllable.

3A to 3C, the meta information inference apparatus 110 generates / learns the morpheme-vector converter 520, the syllable-vector converter 620, and the author recognizer 720.

First, referring to FIG. 3A, the author meta information inference apparatus 110 of the author meta information inference apparatus 110 according to an embodiment of the present invention may include a first test morpheme and a first test morpheme for generating the morpheme-vector converter 520. A plurality of first training data 500 including a first morpheme vector corresponding to one test morpheme may be received and / or obtained. In other words, the creator meta information inference apparatus 110 according to an embodiment of the present invention may acquire learning data including a morpheme and a morpheme vector corresponding to the morpheme. For example, the first learning data 510 may include a morpheme and a morpheme vector 511 corresponding to the morpheme. In this case, the number of morphemes corresponding to any one morpheme vector may be singular or plural.

Subsequently, the apparatus for inferring meta information 110 according to an embodiment of the present invention may train the morpheme-vector converter 520 based on the received and / or obtained first training data 500.

As described above, the morpheme-vector converter 520 may be a correlation between a plurality of morphemes and a plurality of morpheme vectors generated by machine learning, that is, mapping information of both.

Accordingly, the author meta information inference apparatus 110 according to the embodiment of the present invention may correspond to a morpheme of the first training data 500 and a morpheme vector corresponding to the morpheme, so that the morpheme-vector converter 520 can be mapped. The morpheme-vector converter 520 can be trained in such a manner as to update the data set.

For example, when the data set of the morpheme-vector converter 520 is composed of a plurality of coefficients, the creator meta information inference apparatus 110 may determine a morpheme of the first training data 500 and a morpheme vector corresponding to the morpheme. The coefficients constituting the morpheme-vector converter 520 may be appropriately adjusted to correspond to each other (map). In this case, the creator meta information inference apparatus 110 may obtain the morpheme-vector converter 520 having improved accuracy by updating the coefficients based on the plurality of learning data.

Accordingly, the present invention can infer the meta information of the author based on the morpheme feature of the text content.

Referring to FIG. 3B, the creator meta information inference apparatus 110 according to an embodiment of the present invention may include a first test syllable and a first syllable vector corresponding to the first test syllable for generating the syllable-vector converter 620. A plurality of second learning data 600 including a may be received and / or obtained. In other words, the author meta information inference apparatus 110 according to an embodiment of the present invention may acquire the second training data 600 including the syllable and the syllable vector corresponding to the syllable. For example, the first learning data 610 may include syllable vectors and syllable vectors 611 corresponding to syllables. In this case, the number of syllables corresponding to any one syllable vector may be singular or plural.

Subsequently, the apparatus for inferring meta information 110 according to an embodiment of the present invention may train the syllable-vector converter 620 based on the received and / or obtained second training data 600.

As described above, the syllable-vector converter 620 may be a correlation between a plurality of syllables and a plurality of syllable vectors generated by machine learning, that is, mapping information of both.

Accordingly, the apparatus for inferring meta information 110 according to an exemplary embodiment of the present invention is a syllable-vector converter 620 so that the syllable vector of the second training data 600 and the syllable vector corresponding to the syllable can be mapped to each other. The syllable-vector converter 620 may be trained by updating a data set of the ().

For example, when the data set of the syllable-vector converter 620 is composed of a plurality of coefficients, the creator meta information inference apparatus 110 may generate a syllable vector corresponding to the syllable of the second training data 600 and the syllable. Coefficients constituting the syllable-vector converter 620 may be appropriately adjusted to correspond to each other (mapping). In this case, the creator meta information inference apparatus 110 may obtain the syllable-vector converter 620 having improved accuracy by updating coefficients based on the plurality of learning data.

As a result, the present invention can infer the meta information of the author based on the syllable characteristics of the text content, that is, the dialogue (or tone) of the text content, and can infer the meta information of the author with higher accuracy. have.

Meanwhile, referring to FIG. 3C, the author meta information inference apparatus 110 according to an embodiment of the present invention may generate a second stem vector, a second syllable vector, a second syllable vector, and a second stem vector to generate the author recognizer 720. A plurality of third training data 700 including meta information vectors corresponding to syllable vectors may be received and / or obtained. In other words, the creator meta information inference apparatus 110 according to an embodiment of the present invention may acquire learning data including a morpheme vector and a syllable vector, and meta information vectors corresponding to the morpheme vector and the syllable vector.

For example, the first training data 710 may include a morpheme vector Vm2, a syllable vector Vs2, and a meta information vector 711 corresponding thereto.

Subsequently, the creator meta information inference apparatus 110 according to an embodiment of the present invention may train the creator recognizer 720 based on the received and / or obtained third training data 700.

As described above, the author recognizer 720 may be a correlation between a plurality of morpheme vectors, a plurality of syllable vectors, and a plurality of meta information vectors generated by a machine learning technique, that is, mapping information of both.

Accordingly, the author meta information inference apparatus 110 according to an embodiment of the present invention updates the data set of the author recognizer 720 so as to correspond (map) the morpheme vector, the syllable vector, and the meta information vector of the training data. The author recognizer 720 can be trained in a manner of updating.

For example, when the data set of the author recognizer 720 is composed of a plurality of coefficients, the creator meta information inference apparatus 110 corresponds to a mapping of the morpheme vector and syllable vector of the training data and the corresponding meta information vector (mapping). Coefficients constituting the author recognizer 720 may be appropriately adjusted. In this case, the creator meta information inference apparatus 110 may obtain the creator recognizer 720 having improved accuracy by updating coefficients based on the plurality of learning data.

Thus, the present invention can infer the meta information of the author in consideration of both the morphological features and the syllable features of the text content.

By the above-described process, the author meta information inference apparatus 110 according to the embodiment of the present invention is a morpheme-vector converter 520, a syllable-vector converter 620, and a syllable-vector converter for inferring meta information of an author from text contents. The author recognizer 720 may be generated and / or trained.

Author meta information inference apparatus 110 according to an embodiment of the present invention may receive the text content to infer the meta information of the author (S62). Content (eg, images, images, etc.).

Meanwhile, the text content may include only one sentence or may include a plurality of sentences. When the text content includes a plurality of sentences, the author meta information inference apparatus 110 according to an embodiment of the present invention may infer meta information of the author by dividing the content into a plurality of contents.

Hereinafter, with reference to FIG. 5 again, step S63 to step S66 will be described.

The author meta information inference apparatus 110 according to an embodiment of the present invention may divide the text content 810 into one or more morphemes 811, and determine a morpheme vector 821 from the divided one or more morphemes 811. (S63)

For example, the author meta information reasoning device 110, if the text content is 'give for the peace of the Republic of Korea', the content in one or more morphemes, such as 'Korea,' 'peace', 'for' The morpheme vector 821 can be determined based on the division.

As described above, the morpheme-vector converter 520 may be a correlation between a plurality of morphemes and a plurality of morpheme vectors generated by machine learning, that is, mapping information of both.

Accordingly, the creator meta information inference apparatus 110 may input one or more morphemes 811 into the morpheme-vector converter 520, and as a result, obtain a morpheme vector 821 corresponding to the one or more morphemes 811. .

Thus, the present invention can infer the meta information of the author based on the morphological characteristics of the text content.

Author meta information inference apparatus 110 according to an embodiment of the present invention may divide the text content 810 into one or more syllables 812, and determine the syllable vector 822 from the divided one or more syllables 812. (S64)

For example, as in the above example, when the text content is "Give it for the peace of Korea," the author meta information inference apparatus 110 displays the content such as "large", "han", "min", "country", and the like. The syllable vector 822 may be determined based on the division into one or more syllables.

As described above, the syllable-vector converter 620 may be a correlation between a plurality of syllables and a plurality of syllable vectors generated by machine learning, that is, mapping information of both.

Accordingly, the creator meta information inference apparatus 110 may input one or more syllables 812 to the syllable-vector converter 620, and as a result, may obtain a syllable vector 822 corresponding to the one or more syllables 812. .

As a result, the present invention can infer the meta information of the author based on the syllable characteristics of the text content, that is, the dialogue (or tone) of the text content, and can infer the meta information of the author with higher accuracy.

Author meta-information inference apparatus 110 according to an embodiment of the present invention corresponds to the meta information of the author based on the morpheme vector 821, syllable vector 822, and author recognizer 720 determined by the above-described process. The meta information vector 830 may be determined (S65).

Similar to the morpheme-vector converter 520 and syllable-vector converter 620 described above, the author recognizer 720 also includes a plurality of morpheme vectors, a plurality of syllable vectors, and a plurality of meta generated by machine learning techniques. Correlation of information vectors, that is, mapping information of both.

Accordingly, the author meta information inference apparatus 110 inputs the morpheme vector 821 and the syllable vector 822 to the author recognizer 720, and as a result, the meta information corresponding to the morpheme vector 821 and the syllable vector 822. Vector 830 may be obtained. Meanwhile, the apparatus for inferring meta information 110 according to an embodiment of the present invention generates a content vector by merging a morpheme vector and a syllable vector, and based on the generated content vector and the author recognizer 720, the meta information vector 830. You can also decide). In this case, merging a vector may mean creating a new vector corresponding to the sum of the number of dimensions of the two vectors, and generating a vector of a new dimension (dimensions less than or equal to the sum of the number of dimensions of the two vectors) through a predetermined operation. It might mean doing.

The creator meta information inference apparatus 110 according to an embodiment of the present invention may determine the aforementioned series of processes (morpheme vector, syllable vector, and meta information vector) for the divided content generated by dividing the text content into predetermined units. Process) can be performed in the same way. In other words, the author meta information inference apparatus 110 may determine a morpheme vector for each of the one or more pieces of divided content. In addition, the creator meta information inference apparatus 110 may determine a syllable vector for each of the one or more pieces of divided content. In addition, the creator meta information inference apparatus 110 may determine a meta information vector corresponding to the creator's meta information based on a morpheme vector of each of the one or more pieces of divided content, a syllable vector of each of the one or more pieces of divided content, and an author recognizer. As such, the creator meta information inference apparatus 110 may determine the meta information vector for each divided content. On the other hand, the method of determining the meta information of the author for the entire content from this will be described later.

The creator meta information inference apparatus 110 according to an embodiment of the present invention may determine the meta information 840 of the creator of the text content 810 based on the meta information vector 830 determined by the above-described process. In this case, the meta information of the content creator may include at least one of the age group of the creator, the gender of the creator, a region related to the creator, the political propensity of the creator, the educational background of the creator, and whether the creator is married. Of course, the above-described items are exemplary and may be used as the meta information of the present invention as long as the items can be used as meta information of a person.

The creator meta information inference apparatus 110 according to an embodiment of the present invention may determine meta information for each item of the meta information in the form of probability of a plurality of options for each item. For example, the author meta information inference apparatus 110 may determine the age group of the text content creator in the form of a probability that the creator belongs to a plurality of candidate age groups (for example, teenagers, 20s, 30s, 40s, 50s or more). Similarly, the author meta information inference apparatus 110 may determine the gender of the creator in the form of a probability that the author is a male and a probability that the author is a female. However, the above-mentioned age group and gender items are exemplary, and the spirit of the present invention is not limited thereto.

Meanwhile, when the meta information vector is determined for each of the divided contents, the creator meta information inference apparatus 110 according to an embodiment of the present invention may determine the meta information about the respective divided contents.

In this case, the author meta information inference apparatus 110 according to an embodiment of the present invention may determine the meta information for the full text content by merging the meta information for each divided content. For example, when the probability that an author belongs to a plurality of candidate ages is determined for each of the divided contents, the author meta information inference apparatus 110 may add the sum of the probabilities for each age group (for example, the sum of probabilities belonging to teenagers and the sum of probabilities belonging to 20s). Etc.) may determine meta information about the entire text content. However, this is merely illustrative and the spirit of the present invention is not limited thereto.

Accordingly, the present invention can infer the meta information of the author in consideration of both the morpheme features and the syllable features of the text content.

7 is an example of a screen 900 displayed on the display unit of the user terminal 200 according to an embodiment of the present invention.

For convenience of explanation, the server 100 provides a service for providing the meta information of the creator for the text content input by the user through the user terminal 200, and the user may access the server 100 through the user terminal 200. Assume that you have sent a phrase as shown by).

Assuming that the above-described, the server 100 according to an embodiment of the present invention can provide the meta-information of the inferred content creator to the user terminal 200 as the screen 900.

In this case, the screen 900 displays an area 910 in which the content of the content input by the user is displayed, an area 920 in which the gender of the inferred author is displayed in the form of probability, and an age group of the inferred creator is displayed in the form of probability. Region 930 may be included.

Embodiments according to the present invention described above may be implemented in the form of a computer program that can be executed through various components on a computer, such a computer program may be recorded on a computer readable medium. In this case, the medium may be a program executable by a computer. Examples of the medium include magnetic media such as hard disks, floppy disks and magnetic tape, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, And ROM, RAM, flash memory, and the like, configured to store program instructions.

On the other hand, the computer program may be specially designed and configured for the present invention, or may be known and available to those skilled in the computer software field. Examples of computer programs may include not only machine code generated by a compiler, but also high-level language code executable by a computer using an interpreter or the like.

Particular implementations described in the present invention are embodiments and do not limit the scope of the present invention in any way. For brevity of description, descriptions of conventional electronic configurations, control systems, software, and other functional aspects of the systems may be omitted. In addition, the connection or connection members of the lines between the components shown in the drawings by way of example shows a functional connection and / or physical or circuit connections, in the actual device replaceable or additional various functional connections, physical It may be represented as a connection, or circuit connections. In addition, unless specifically mentioned, such as "essential", "important" may not be a necessary component for the application of the present invention.

Therefore, the spirit of the present invention should not be limited to the above-described embodiments, and the scope of the spirit of the present invention is defined not only in the claims below, but also in the ranges equivalent to or equivalent to the claims. Will belong to.

100: server
110: meta information inference device
111: communication unit
112: control unit
113: memory
200: user terminal
300: external device
400: network

Claims (10)

A method of inferring meta information of an author by dividing text contents into predetermined units by a meta information inference device,
Receiving text content;
Dividing the text content into predetermined units to generate one or more divided contents;
Dividing each of the one or more fragmented contents into one or more morphemes and determining a morpheme vector for each of the one or more fragmented contents from the segmented one or more morphemes based on a morpheme-vector converter;
Dividing each of the one or more divided contents into one or more syllables, and determining a syllable vector for each of the one or more divided contents from the divided one or more syllables based on a syllable-vector converter; And
Determining a meta information vector corresponding to the meta information of the author based on a morpheme vector of each of the one or more pieces of divided content, a syllable vector of each of the one or more pieces of divided content, and an author recognizer;
The morpheme-vector converter
A data set representing a correlation between a plurality of stems and a plurality of stem vectors,
The syllable-vector converter
A data set representing a correlation between a plurality of syllables and a plurality of syllable vectors,
The creator recognizer
A method of inferring meta information of an author, which is a data set representing a correlation between a plurality of morpheme vectors, a plurality of syllable vectors, and a plurality of meta information vectors. delete delete delete The method of claim 1
The method of inferring the meta information of the author
After determining the meta information vector
And determining the meta information of the text content creator based on the determined meta information vector. The method of claim 5
The meta information of the text content creator
And at least one of an age range of the author, a gender of the author, a region associated with the author, a political propensity of the author, an education of the author, and whether the author is married. The method of claim 6
Determining the meta information of the creator
Determine the age group of the text content creator in the form of a probability that the author belongs to a plurality of candidate age groups,
And determining the gender of the text content creator in the form of a probability that the author is a male and a probability that the author is a female. The method of claim 1
The method of inferring the meta information of the author
Training the morpheme-vector converter based on a plurality of first training data comprising a first test morpheme and a first morpheme vector corresponding to the first test morpheme;
Training the syllable-vector converter based on a plurality of second training data including a first test syllable and a first syllable vector corresponding to the first test syllable; And
Training the creator recognizer based on a second morpheme vector, a second syllable vector, and third training data including a meta information vector corresponding to the second morpheme vector and the second syllable vector; How to infer the author's meta information. delete A computer program stored in a computer readable recording medium for carrying out the method of any one of claims 1 and 5 to 8 using a computer.

Images