KR102479381B1 - System for content clustering based on thesaurus and method therefor - Google Patents

Abstract

The present invention relates to a content clustering system based on synonyms, and more particularly to a content clustering system and method for constructing a database by classifying content elements based on synonyms. According to the present invention, there is an effect of creating a database for expressions of an equal level having similar meanings, and classifying and creating a database for expressions having different categories of expressions that can be hierarchically arranged.

Description

System for content clustering based on thesaurus and method therefor}

The present invention relates to a content clustering system based on synonyms, and more particularly, to a content clustering system and method for constructing a database by classifying content elements based on synonyms.

As the number of users rapidly increases with the development of the Internet, the web service environment is changing in various ways. While conventional web services are statically passive, web services are gradually changing dynamically and actively, and Web 2.0 has been introduced to reflect the trend of web service changes.

Web 2.0 refers to a series of movements developed to induce information sharing and participation among Internet users through the opening of information and to continuously increase the value of created information. In other words, in Web 2.0, netizens can freely participate, create, recreate, and share content based on an open web environment.

In Web 2.0, information is created by users, and information is organized by tags attached by users. Users can easily share this information, so that various resources are interrelated. As such, the Web 2.0 phenomenon has become an essential strategy for all Internet sites, and various techniques are being introduced to successfully implement Web 2.0.

One of these techniques is tagging. Tagging is widely used from web documents such as blogs to multimedia contents such as images and videos, and allows users to easily search and classify content by attaching tags to content they create.

According to the conventional tag clustering apparatus based on related tags disclosed in Patent Publication 2010-0013157, the occurrence frequency of the same tag is extracted in a tag mapping process, and only related tags having a frequency higher than a threshold value are extracted among related tag pairs to generate a tag cluster. Therefore, tags representing synonyms are regarded as the same tag.

Statistical guesses, such as frequency-based text analysis, can theoretically solve this problem very well if there are records of all languages used in human activities, but in reality there are limits to the textual information that can be collected, and there are limits to geography, age, and region. The specific language usage habits of subgroups of language use, which can be divided into areas of interest, occupations, etc., have linguistic diversity that can be missed by this general analysis, which is used in statistical inference methods to predict similarity of expressions. Language databases also have the problem of not being able to use the same one in every situation.

Patent Publication No. 10-2010-0013157 (2010.02.09)

The present invention has been devised to solve the above problems, and a synonym-based content clustering system that defines similar data as various linguistic metadata and comprehensively recognizes query data to semantically classify it in various purposes and situations, and its The purpose is to provide a method.

A content clustering system based on synonyms according to an aspect of the present invention includes a tag extraction unit for extracting a tag related to the same content from among content included in a predetermined population; an arrangement unit for arranging tags in a semantically hierarchical hierarchical structure in the tag extraction process; an anchor unit constituting an anchor connecting semantically adjacent tags in the process of arranging the tags; and a classifier for clustering content by classifying concepts according to connection strength by assigning weights to connection adjacencies of tag concepts for connections through the anchor unit.

Preferably, the tag extraction unit extracts a tag related to the same content from among contents including predefined tagging information in a text format in order to show search results based on a user's text format query.

Preferably, the classification unit stores metadata including tags or descriptions for linguistically explaining the content of the data together with the original data of the content and storing them in conjunction with each other.

Preferably, the arranging unit arranges each concept in a hierarchical structure having a semantic hierarchy in a tag extraction process, and stores inclusion relationships and synonym relationships in a database.

Preferably, the anchor unit configures anchors that horizontally connect semantically adjacent tags in the process of arranging tags to connect adjacent concepts.

Meanwhile, a content clustering method based on synonyms includes the steps of (a) extracting tags related to the same content from among content included in a predetermined population; (b) arranging tags in a hierarchical structure having a semantic hierarchy in the tag extraction process; (c) constructing anchors connecting semantically close tags in the process of arranging the tags; (d) a classification step of clustering content by assigning weights to connection adjacencies of tag concepts for connections through the anchor unit and classifying concepts according to connection strength.

According to the synonym-based content clustering system according to an embodiment of the present invention, similar data can be defined as various linguistic meta data to comprehensively recognize query data and semantically classify it for various purposes and situations. .

1 is a diagram showing the configuration of a content clustering system based on synonyms according to an embodiment of the present invention.
2 is a diagram illustrating text information tagging in a content clustering system based on synonyms according to an embodiment of the present invention.
3 is a diagram showing categories of images displayed on a video.
4 is a diagram illustrating a situation in which a notation on a database is changed even though it is a synonym.
5 is a diagram showing associations between words.
6 is a diagram for explaining a hierarchical arrangement structure and anchor connection of a synonym-based content clustering system according to an embodiment of the present invention.
7 is a flowchart illustrating a content clustering method based on synonyms according to an embodiment of the present invention.

Specific structural or functional descriptions presented in the embodiments of the present invention are merely exemplified for the purpose of explaining embodiments according to the concept of the present invention, and embodiments according to the concept of the present invention may be implemented in various forms. In addition, it should not be construed as being limited to the embodiments described in this specification, but should be understood to include all modifications, equivalents, or substitutes included in the spirit and scope of the present invention.

Meanwhile, in the present invention, terms such as first and/or second may be used to describe various elements, but the elements are not limited to the above terms. The above terms are used only for the purpose of distinguishing one component from other components, for example, within a range not departing from the scope of rights according to the concept of the present invention, a first component may be referred to as a second component, Similarly, the second component may also be referred to as the first component.

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In describing the embodiments of the present invention, if it is determined that a description of a related known function or configuration may unnecessarily obscure the subject matter of the present invention, the description is omitted.

1 is a diagram showing the configuration of a content clustering system based on synonyms according to an embodiment of the present invention. As shown in FIG. 1 , the synonym-based content clustering system 10 includes a tag extraction unit 100 , a placement unit 200 , an anchor unit 300 and a classification unit 400 .

The tag extraction unit 100 extracts tags related to the same content from content included in a predetermined population. The tag extraction unit extracts a tag related to the same content from content including predefined tagging information in text form in order to show a search result based on a user's text form query.

The arrangement unit 200 is a component that arranges tags in a hierarchical structure having a semantic hierarchy in the tag extraction process. In the process of tag extraction, the arranging unit arranges concepts in a hierarchical structure having semantic hierarchies, and stores inclusion relationships and synonym relationships in a database.

The anchor unit 300 configures anchors that connect semantically close tags in the process of arranging tags. In the process of arranging tags, the anchor unit configures anchors that horizontally connect semantically adjacent tags to connect adjacent concepts.

The classification unit 400 assigns a weight to the connection adjacency of the tag concept for connection through the anchor unit, classifies the concept according to the connection strength, and clusters the contents. The classification unit enables the classification unit to link and store meta data including tags or descriptions for linguistically describing the contents of the data together with the original data of the contents. Also, in the process of tag extraction, each concept is arranged in a hierarchical structure having a semantic hierarchy, and the inclusion relationship and the synonym relationship are stored in the database.

The content clustering system based on synonyms according to an embodiment of the present invention classifies content elements of narrative stories based on synonyms and converts them into databases. In order to construct an efficient database, the following conditions are required. In constructing a library of audio, video, interactive contents, etc., predefined tagging information in text form is required to show search results based on a user's text form query. For linguistic explanation of such non-linguistic data, meta data such as tags or descriptions for linguistically explaining the contents of the data along with the original data must be linked and stored. The linguistic tag and description information must be able to express the contents of non-linguistic data in as much detail as possible, and for this purpose, classification criteria under appropriate conditions are required.

2 is a diagram illustrating text information tagging in a content clustering system based on synonyms according to an embodiment of the present invention. 2 shows an example of text information tagging for video. In order to build an efficient database, predefined tagging information in the form of text is required, meta data must be interlocked and stored, and classification criteria under appropriate conditions are required. need this

However, even if linguistic metadata as detailed as possible is configured for these various classification conditions, a problem arises in that various types of linguistic metadata can be configured for similar information.

3 is an example of words that can be selectively or simultaneously used in the meta data configuration of image information. As shown in FIG. 3 , an image displayed on a video may be expressed in several vocabularies having different categories. When different tagging is performed depending on the context, it becomes difficult to sufficiently express such hierarchical information with a general SQL query.

4 is a diagram illustrating a situation in which expressions of the same meaning but notations on the database may be different, and FIG. 5 is a diagram showing associations between words. As shown in FIG. 5, there is an example of a situation in which the concept of Eo-hee cannot be expressed in horizontal and vertical categories. Word-to-word associations should be expressed as quantitative values with multiple scales, not simple vertical categories.

As shown in FIGS. 4 and 5, meta data can be created manually or automatically by a machine, but both of these methods are one of various expressions in linguistically describing a specific situation of image information. You are put in a situation where you have to choose one or some of them. This problem occurs due to the following reasons.

When different expressions are semantically similar and can be selectively used, when semantically located higher or lower and refer to the same object, but the range of expressions that can be interpreted can be different, when a plurality of words are the same Although it can refer to an object, the scope of meaning defined by each vocabulary includes cases where it is different from each other.

If such similar data is defined as other linguistic meta data, it is impossible to provide comprehensive search results of all similar databases according to the queryer's intention with only a basic search query of a general database such as SQL. This is a major obstacle in semantically classifying image databases for various purposes and situations, and means that a means to comprehensively recognize query data is required.

Statistical guesses, such as frequency-based text analysis, can theoretically solve this problem very well if there are records of all languages used in human activities, but in reality there are limits to the textual information that can be collected and the The specific language usage habits of subgroups of language use, which can be divided into fields of interest, occupations, etc., have linguistic diversity that can be missed by this general analysis, which is used in statistical guessing methods to predict the similarity of expressions. Databases also mean that you can't use the same one in every situation.

6 is a diagram for explaining a hierarchical arrangement structure and anchor connection of a synonym-based content clustering system according to an embodiment of the present invention. This database may be configured as shown in FIG. 6 . Each concept can be semantically arranged in a hierarchical structure, and through this, basic inclusion and synonym relationships can be stored in the database. In addition, in the case of adjacent concepts that are difficult to define as hierarchical concepts, existing on different hierarchies but semantically close, all possible adjacent concepts can be connected by constructing anchors that are connected horizontally, and weighted for anchor connections. As a result, it is possible to specify the connection strength of tighter connections and vague connections with relatively few adjacencies.

The initial anchor configuration can use a statistical method, but this method, which can provide visualized connectivity that humans can understand, implements an easy supervised learning procedure through human supervised learning, so that the amount of data learning is small or it is not suitable for general use. It is possible to build a fast concept classification machine when there is a need to connect appropriate adjacent concepts in a special situation that is not the case.

As shown in FIG. 6, a tree-structured arrangement of vocabularies or concepts having a hierarchical relationship creates a horizontal anchor for a lower concept that cannot be fully covered by a concept of a higher category, thereby providing an organic connection. . A synonym includes a display providing a visual and intuitive view of the synonym concepts of each concept.

Meanwhile, in the content clustering method based on synonyms according to the present embodiment, as shown in FIG. 7 (701), tags related to the same content are extracted from content included in a predetermined population. Next, in the tag extraction process (703), tags are arranged in a hierarchical structure having a semantic hierarchy. Next, in the process of arranging (705) tags, an anchor connecting semantically close tags is configured. Next, in step 707, the connection adjacency of the tag concept is weighted for the connection through the anchor unit, and the concepts are classified according to the connection strength to cluster the contents.

Expected effects of the synonym-based content clustering system according to an embodiment of the present invention are as follows. A synonym database based on human collection and modification can save the storage and computing resources required for big data and machine learning, and can be flexibly modified according to the type of content collected. It has the advantage that the database administrator can intuitively check the evidence and add improvements. In order for statistical analysis based on big data to work meaningfully, text information of at least tens of gigabytes or more must be kept and continuously updated. , and it is impossible to prove whether or not the collected text data is a suitable sample for analyzing the image database, the extended search query method based on the synonym database can be said to have great utility in relatively small and medium-sized projects. .

A content clustering system based on synonyms according to an embodiment of the present invention creates a database for expressions of equal level with similar meanings, and classifies and creates a database for expressions with different categories of expressions that can be hierarchically arranged. can

The present invention described above is not limited by the above-described embodiments and the accompanying drawings, and various substitutions, modifications, and changes are possible without departing from the technical spirit of the present invention will be apparent to those skilled in the art.

100: tag extraction unit
200: placement unit
300: anchor part
400: classification unit

Claims (6)

a tag extraction unit for extracting a tag related to the same content from among content included in a predetermined population;
an arrangement unit for arranging tags in a semantically hierarchical hierarchical structure in the tag extraction process;
an anchor unit constituting an anchor connecting semantically adjacent tags in the process of arranging the tags;
A classification unit for clustering content by assigning weights to connection adjacencies of tag concepts for connections through the anchor unit and classifying concepts according to connection strength;
The classification unit constructs anchors that are horizontally connected in the case of adjacent concepts that exist on different hierarchies but are semantically close, cluster all possible content of adjacent concepts, and define the connection strength according to the connection adjacency for the connection of the anchors A content clustering system based on synonyms, characterized in that for doing. According to claim 1,
The tag extraction unit extracts a tag related to the same content from among content including predefined tagging information in a text form in order to show search results based on a user's text form query. based content clustering system. According to claim 1,
The synonym-based content clustering system according to claim 1 , wherein the classification unit stores metadata including tags or descriptions for linguistically explaining the content of the data together with the original data of the content. According to claim 1,
The content clustering system based on synonyms, characterized in that the arrangement unit arranges each concept in a hierarchical structure having a semantic hierarchy in a tag extraction process and stores inclusion relationships and synonym relationships in a database. According to claim 1,
The content clustering system based on synonyms, characterized in that the anchor unit configures anchors that horizontally connect semantically adjacent tags in the process of arranging tags to connect adjacent concepts. (a) extracting a tag related to the same content from among content included in a predetermined population;
(b) arranging tags in a hierarchical structure having a semantic hierarchy in the tag extraction process;
(c) constructing anchors connecting semantically close tags in the process of arranging the tags;
(d) a classification step of clustering content by assigning weights to connection adjacencies of tag concepts for connections through the anchors and classifying concepts according to connection strength;
In the step (d), in the case of adjacent concepts that exist on different hierarchies but are semantically close, configure anchors that are horizontally connected, cluster the contents of all possible adjacent concepts, and connect the anchors according to the connection adjacency A content clustering method based on synonyms characterized by defining strength.

Images