KR20140101906A - Apparatus and Method for social data analysis - Google Patents

Abstract

In order to achieve the above purpose, an apparatus for analyzing social data according to the present invention includes: a data collecting unit which collects data on content present in a server group through a communication network, and stores the collected data; a message analysis unit which analyzes the data in morphological units matched to a morpheme DB and an emotional vocabulary DB, tags the data with parts of speech in the morphological units, and searches out and stores emotional vocabularies; an emotion evaluation unit which receives, from the message analysis unit, unregistered vocabularies which are not tagged with parts of speed, performs emotion evaluation, and stores, in an emotion DB, standard language evaluation vocabularies corresponding to the unregistered vocabularies; an emotion analysis unit which performs an emotional engineering analysis of an emotion measurement target, based on a plurality of emotional vocabularies provided by the message analysis unit and the emotion evaluation unit, and extracts a main emotion image from the emotion measurement target; and an analysis result display unit which displays the emotion measurement target and the main emotion component on an image space in real time. The present invention can naturally identify mixed emotions based on ordinary language, slang, jargon, abbreviation, emoticon, etc. and can measure the hidden emotions of consumers by utilizing documents, internet comments, SNS message text data spontaneously made by the consumers.

Description

[0001] Apparatus and Method for Social Data Analysis [

The present invention relates to an apparatus and a method for analyzing social data received via a communication network. More specifically, the present invention relates to a new concept, a brand concept, a naming, a design, a person or an object such as an entertainer or a politician, etc. from a document, an internet comment, and a social network service The present invention relates to an apparatus and method for emotional measurement and analysis capable of automatically quantifying, evaluating, and analyzing detailed images and detailed emotions.

With the expansion of smartphone and the Internet nowadays, digital data is explosively increasing to be difficult to handle in the conventional way, and who extracts the value from big data first determines the success of the enterprise and the nation. . However, the mechanical classification method for big data including current social media is mass producing results, and emotional and emotional analysis should be done for big data to help. The emotional measurement and analysis techniques related to this are the 'Sentiment Analysis' technology which classifies the emotions into two or three big categories such as positive / negative and the 'emotional' Engineering technology '.

Emotional analysis is also called 'Opinion Mining' or 'reputation analysis'. And whether the content contained in the text such as a message is subjective or objective and if the subject is subjective, it is analyzed to determine whether the content is positive or negative or whether the content is neutral. In other words, the article written by the person who wrote the article is analyzed to determine whether it is positively or negatively in the context, and is used to grasp the user's reaction or opinion on a specific product or service. This emotion analysis is known to be able to grasp online opinion like social media relatively quickly, and it can save time and money compared to existing open-line survey, and can easily grasp and predict people's opinion.

However, the prior arts such as Korean Patent Publication No. 2012-0108095 have the following problems. First, dividing into three categories of positive, negative, or positive, negative, and neutral is a big constraint on the application. In other words, emotions such as 'retro', 'new generation', 'feminine' and 'masculine' are important emotions in image appreciation for specific products or characters, but they are simply classified as 'neutral' And feelings can not be grasped. Rather, polarity judgments undermine the accuracy of emotional analysis by forcing ambiguous emotions into two (positive / negative) or three (positive / negative / neutral) categories. Second, existing emotion analysis techniques exclude profanity, abbreviation, abbreviation, and emoticon filtering and analyze only standard vocabulary. However, many articles (texts) such as SNS, homepage comments, and blogs contain spoken words, idiomatic expressions, and emoticons. By filtering these texts, data including various emotions are excluded from the emotional analysis process. Therefore, detailed images and detailed emotions of people and objects, including data that are out of the standard vocabulary categories such as profanity, fluency, abbreviation, etc., are collected automatically from collected documents, Internet comments, SNS (Social Network Service) There is a need for social data analysis devices and methods that can quantitatively measure, evaluate and analyze.

Disclosure of Invention Technical Problem [8] Accordingly, the present invention has been made to solve the above problems of the related art by providing a social data analysis apparatus and method capable of naturally grasping complex emotions based on everyday language, profanity, abbreviation, abbreviation, emoticon, The purpose.

It is another object of the present invention to provide a social data analysis apparatus and method for measuring and evaluating the hidden emotions of consumers from documents prepared by a plurality of consumers, Internet comments, and SNS message text data.

It is another object of the present invention to provide an apparatus and method for analyzing social data that displays real-time and continuous analysis results by locating several main emotional image factors on an image space for a specific time point or a specific time period.

According to an aspect of the present invention, there is provided a social data analysis apparatus comprising: a data collection unit for collecting and storing data of contents existing in a server group through a communication network; Analyzing the data in a morpheme unit matched with the morpheme DB and the emotional vocabulary DB, tagging parts of speech by the morpheme unit, extracting emotional vocabulary and storing the sensed vocabulary; An emotion evaluation unit for performing emotional evaluation on an unregistered vocabulary that is not partly tagged in the message analyzing unit and storing a standard word emotional vocabulary corresponding to the unregistered vocabulary obtained as a result of the emotional evaluation in an emotional DB; An emotional analysis unit for extracting a main emotional component, which is a representative emotional component for the emotional measurement object, after the emotional engineering analysis of the emotional measurement object based on the emotional vocabulary provided by the message analysis unit and the emotional evaluation unit; And an analysis result display unit for displaying the sensibility measurement object and the main sensibility component on an image space in real time.

In this case, the data collecting unit may receive a keyword of data or a period of collecting data, collect and store data corresponding to the input, and store the classified data according to the source of the data or the uploader of the data.

At this time, the data collecting unit supports a parallel processing technique for collecting and storing data by utilizing a plurality of computers, a multiprocessor or a multi-thread simultaneously, deletes redundant data among data collected through the parallel processing technique, If the parallel processing technique is interrupted during the process, the data can be automatically connected to continuously collect and store the data after the interrupted data.

In this case, the message analyzing unit may include an input filter module for filtering or eliminating or repeating short patterns without repetition or repetitive use of meaningless special symbols. And an unregistered word processing module for delivering an unregistered vocabulary corresponding to at least one of profanity, hangeul, abbreviation, and emoticon that has not been tagged with part-of-speech to the sensitivity evaluation unit.

At this time, the message analyzing unit may support a parallel processing technique that utilizes a plurality of computers, a multiprocessor, or a multi-thread simultaneously.

At this time, the message analyzing unit can distinguish the vocabulary related to the emotion and emotion based on the morpheme DB and the emotional vocabulary DB, and can separate and store the subject and object of the vocabulary related to the emotion and emotion.

At this time, the emotion evaluation unit performs emotional evaluation using a pair of standard word emotional vocabularies, and an evaluator can evaluate the standard word emotional vocabulary pairs through an online connection.

At this time, the emotion evaluation unit may select a standard word that can be substituted for an unregistered vocabulary that is not part-tagged by the evaluator through online connection, as a weighted sum of two or more standard words.

At this time, the emotion evaluation unit may directly receive the intensity of the emotion and emotion for the unregistered vocabulary that has not been tagged with the part-of-speech, or may be evaluated through an online questionnaire evaluation.

At this time, the emotional analysis unit performs sparse principal component analysis (SPCA) based on a combination of the emotional vocabulary of the emotional measurement target and the frequency and the emotional intensity of the vocabulary, The components can be extracted.

At this time, the extraction of the main emotional component may be performed by setting a specific point in time or a specific period.

At this time, the analysis result display unit may be configured such that a plurality of principal emotion components extracted from the emotion analyzing unit are configured as two-dimensional or three-dimensional axes, and the position on the image space with respect to a specific time point of the emotion measurement object is displayed in real time , And the analysis result according to the change of time during a specific period can be displayed in an animation format.

At this time, the analysis result display unit can simultaneously display several sensibility measurement objects on an image space composed of common emotional and emotional image factors.

According to another aspect of the present invention, there is provided a method of analyzing social data, comprising: collecting and storing data of contents existing in a server group through a communication network; Analyzing the data in a morpheme unit matched with the morpheme DB and the emotional vocabulary DB, tagging parts of speech by the morpheme unit, searching and storing emotional vocabulary; Performing emotional evaluation on an unregistered vocabulary that is not part-tagged in the detecting and storing step and storing a standard word emotional vocabulary corresponding to the unregistered vocabulary obtained as a result of the emotional evaluation in an emotional DB; Extracting and storing a main emotion component as a representative emotion component for the emotion measurement object after emotion engineering analysis of the emotion measurement object based on the emotion word provided in the step of extracting and storing the emotion word and storing the emotion word in the emotion DB; And displaying the sensibility measurement object and the main sensibility component on the image space in real time.

In this case, the collecting and storing step may include receiving a keyword of data or a period of collecting data, collecting and storing data corresponding to the input, and classifying the data by the source of the data or the uploader of the data A parallel processing technique for collecting and storing data by using a plurality of computers, a multiprocessor or a multi-thread simultaneously is supported, and redundant data among data collected through the parallel processing technique is deleted, And if the parallel processing technique is interrupted during the process, it is automatically connected to continue collecting and storing the data after the interrupted data.

At this time, the extracting and storing step distinguishes vocabulary related to emotion and emotion based on the morpheme DB and the emotional vocabulary DB, separates and stores the subject and object of the vocabulary related to the emotion and emotion, , Parallel processing techniques that utilize multiprocessor or multithread simultaneously can be supported.

At this time, the extracting and storing step may be performed by eliminating or shortening the repetition of a short pattern without spacing or repetitive use of a meaningless special symbol, and analyzing it by using an abbreviated word, abbreviation, abbreviation, or emoticon Only when there is at least one unregistered vocabulary corresponding to at least one of the unregistered vocabularies is stored in the emotional DB.

At this time, in the step of storing in the emotional DB, emotional evaluation is performed using a pair of standard word emotional vocabulary, and the evaluator is allowed to evaluate the standard word emotional vocabulary pair through online connection, The standard word that can replace the unregistered vocabulary which is not tagged with the part-of-speech is selected as a weighted sum of two or more standard words, and the intensity of emotion and emotion for the unregistered vocabulary that is not tagged with the part- .

At this time, the extracting step may be performed by performing Sparse Principal Component Analysis (SPCA) based on a combination of emotional vocabulary for an emotional measurement target and frequency and emotional intensity of the vocabulary, Extracting the sensible component, and extracting the main sensible component may be performed by setting a specific point in time or a specific period.

At this time, in the displaying step, a plurality of principal emotion components extracted in the extracting step are constituted by two-dimensional or three-dimensional axes, and the position on the image space with respect to a specific time point of the sensitivity measurement object is displayed in real time , An analysis result according to the change of time during a specific period is displayed in an animation format, and a plurality of objects for sensitivity measurement can be simultaneously displayed on an image space composed of common emotional and emotional image factors.

The present invention has an effect of providing a social data analysis apparatus and method capable of naturally grasping complex emotions based on everyday language, profanity, abbreviation, abbreviation, emoticon, and the like.

Further, the present invention provides an apparatus and method for analyzing social data that measures and evaluates the hidden emotions of the consumers, utilizing the documents, Internet comments, and SNS message text data that are created by a plurality of consumers.

In addition, the present invention has the effect of displaying real-time and continuous analysis results by placing several main emotional image factors on the image space for a specific time point or a specific time period.

1 is a view for explaining a configuration of a social data analysis apparatus according to the present invention.
FIG. 2 is a view for explaining an example of collecting contents data from a data collecting unit server group of a social data analyzing apparatus according to the present invention.
FIG. 3 is a diagram for explaining the configuration and operation of a message analyzing unit of the social data analyzing apparatus according to the present invention.
4 is a flowchart illustrating an operation of the method for analyzing social data according to the present invention.
FIG. 5 is a diagram illustrating an example of a standard word emotional lexical pair used in the social data analyzing apparatus according to the present invention.
6 is a diagram showing an example of principal component analysis used in the sensitivity evaluation unit of the social data analysis apparatus according to the present invention.
7 is a diagram showing an example of the output of the analysis result display unit.
8 is a diagram showing another example of the output of the analysis result display unit.
9 is a diagram showing another example of the output of the analysis result display unit.

The present invention will now be described in detail with reference to the accompanying drawings. Hereinafter, a repeated description, a known function that may obscure the gist of the present invention, and a detailed description of the configuration will be omitted. Embodiments of the present invention are provided to more fully describe the present invention to those skilled in the art. Accordingly, the shapes and sizes of the elements in the drawings and the like can be exaggerated for clarity.

Using emotional analysis, it is known that online opinion such as social media can be understood relatively quickly, and it is known that it can save time and money compared to existing offline opinion poll, and can easily grasp and predict opinions of people. The main techniques for analyzing positive / negative are the following techniques, and some of these techniques are analyzed together. The Support Vector Machines (SVMs) method is a method of classifying affirmative / negative opinions of text with training sets previously classified as affirmative / negative, and N-grams or Part Of Speech is classified into N-grams word structure There is a sentence called "I do not like to drink tea" in the way of finding affirmation / negation and separating this sentence into Bi-gram Decomposition, and I-do, do-not and not-like . At this time, N is mainly used up to 1, 2, and 3. Lexicon-based Approach is a technique to identify affirmation and negation with the pre-defined positive / negative Bag of Words (1-grams or Uni-grams) Linguistic Inquiry and Word Count) or POMS (Profile of Mood States). In addition, the Linguistic Approach is a technique for discriminating the polarity by grasping the grammatical structure of text, and is mainly used in conjunction with the Lexicon-based Approach method. It is characterized by distinguishing the polarity by grasping the context and the like.

In addition to the emotional analysis of extracting emotions and emotions from the text data, it is also known as human sensibility ergonomics (sensibility ergonomics / image technology), which aims to quantitatively measure emotions directly to various products and designs Is currently being activated. Sensibility engineering is a technique that scientifically measures and analyzes a complex sensibility such as comfort, comfort, discomfort, and inconvenience of an external physical stimulus obtained through an individual experience, so that the product or environment is convenient and comfortable It is a field to develop comfortably. It is largely divided into biometrics, human sense of the five senses and emotional processing, emotional design, micro processing, and usability evaluation technology. Emotion engineering techniques are generally used to quantitatively measure emotions using the Semantic Differential (SD) method, which presents a large number of pairs of emotional adjectives to the evaluator, (Multi Dimensional Scaling), and the like, the results of the emotion map type are displayed. It has been widely used in automotive instrument panel, interior, housing, cosmetic container, electric car exterior, yacht exterior design, etc. It can also be used for diagnostic evaluation of brand, concept, design and naming, image diagnosis evaluation of objects and people .

Hereinafter, the structure and functions of the social data analysis apparatus according to the present invention will be described.

1 is a view for explaining a configuration of a social data analysis apparatus according to the present invention.

Referring to FIG. 1, the social data analysis apparatus 100 can know that it collects content data from the server group 10, and includes a data collecting unit 110, a message analyzing unit 120, a morpheme DB 121, A sensibility database 130, a sensibility DB 131, an emotion analyzing unit 140, and an analysis result display unit 150. The emotion analyzing unit 140,

FIG. 2 is a view for explaining an example of collecting contents data from a data collecting unit server group of a social data analyzing apparatus according to the present invention.

1 and 2, the data collecting unit 110 collects and stores data of contents existing in the server group 10 through a communication network 20 such as the Internet network. In this case, the term 'social data' refers to data uploaded by users who use various online services such as a blog, a cage, a portal, a show mall, and a Twitter . The data collecting unit 110 may collect social data using various methods. For example, the social data can be collected using a URL request method, a method using a search engine, a method of directly executing a browser, and the like. In addition, the data collection unit 110 receives a data keyword or a period of time for collecting data, and inputs a social media message including a keyword, a text in a homepage, a text in a homepage, And the like can be automatically collected and stored in the DB of the data collecting unit. In addition, a message, URL, information (gender, region) of the uploaded person can be typed and extracted from each collected social media message, text comment, opinion, and media data bundle and stored in the DB. The data collecting unit 110 collectively collects a large amount of data within a unit time by using a parallel processing technique that simultaneously utilizes a plurality of computers, a multiprocessor, and a multithread, and the data collected by each computer, It is possible to delete the redundant data among the data and integrate the data by time or type. At this time, the data collecting unit 110 automatically reconnects when the communication connection of the Internet or the like is disconnected while the parallel processing technology is in progress, and has a function of continuously collecting data after the interrupted data.

FIG. 3 is a diagram for explaining the configuration and operation of a message analyzing unit of the social data analyzing apparatus according to the present invention.

Referring to FIGS. 1 and 3, the message analyzer 120 analyzes the texts and messages collected and stored in the data collecting unit 110 by morpheme units to distinguish parts of speech, tagging. The part - of - speech (word of speech) divides the word according to its grammatical function, form, and meaning. This is because of the fact that there are two kinds of names: Noun (Name), Daemyung (Daemma), Sasa (Sam), Survey (Toshi), Verb (Mang), Adjective Divided. The message analyzing unit 120 may include an input filter module 123 for filtering or eliminating short repeated patterns or repetitive use of meaningless special symbols, and an input filter module 123 for analyzing shortened patterns, And an unregistered word processing module 125 for transmitting an unregistered vocabulary corresponding to at least one of the emoticons to the emotion evaluation unit. A sentence separating module for separating other sentences, a morpheme analysis module for analyzing the morpheme, and a speech tagger module 124 for tagging the part of speech. Also, as in the data collecting unit 110, a parallel processing technique that simultaneously utilizes a plurality of computers, a multiprocessor, or a multi-thread is supported to increase the efficiency of analysis. 3, the message analyzing unit 120 distinguishes vocabulary related to emotion and emotion by using the morpheme DB 121 and the emotional vocabulary DB 122, It has a function to separate and store subjects and objects. At this time, the unregistered vocabularies such as profane words, abbreviations, abbreviations, and emoticons that are not stored in the morpheme DB 121 and the emotional vocabulary DB 122, To the emotion evaluation unit 130. The morpheme DB and emotional vocabulary DB can be updated and stored and deleted later. Therefore, as described above, the message analyzing unit 120 searches the emotional vocabulary for all the data collected and stored in the data collecting unit 110 except for the unregistered vocabulary.

FIG. 5 is a diagram illustrating an example of a standard word emotional lexical pair used in the social data analyzing apparatus according to the present invention.

The emotion evaluation unit 130 selects a standard word that can be substituted for an unregistered vocabulary such as profanity, hermit, abbreviation, or emoticon that has not been tagged with a part of speech by the unregistered word processing module 125 of the message analysis unit 120 Save emotional vocabulary. When the emotional DB 131 is stored in the emotional DB 131 of the emotional evaluation unit 130, the emotional DB 131 can be updated and stored and deleted later. The emotion evaluation unit 130 may classify and store the unregistered vocabulary according to the type such as silver, profanity, abbreviation, and emoticon, and may store the unregistered vocabulary so that the evaluator can perform the sensitivity evaluation using the standard word emotional vocabulary pair. do. At this time, each swear word, profanity, abbreviation, and emoticon is displayed on the screen of an output device such as a monitor through online connection, and the evaluator evaluates the standard word sentence lexical pair or the combination of the standard word by input device such as a mouse or a keyboard The evaluation can be carried out online. Referring to FIG. 5, in order to map (or translate) an adjective word of 'standardized word' to a standard word, for example, the standard word emotional word pair is presented as pairs (having opposite meanings) (Usually 20 to 40 pairs) to the subjects. This is a method used in the SD method (Semantic Differential method). The SD method is one of the method of irradiating and is used for measuring various images and preferences, and is abbreviated as the SD method. Use the adjective-opposite words such as big-small, good-bad, etc. to investigate to what degree a certain brand or product gives a close feeling to either side. A relatively stable result can be obtained with an average of about 30 persons. In addition, if the sensibility evaluation unit 130 can not select one substitute standard word that can replace each word or abbreviation, or if the meaning is unclear, The strength of emotion and emotion is evaluated through online questionnaire evaluation and the result is stored in the emotional DB. The emotional DB may contain one or a combination of standard words and weights that can replace the meaning of each of the words, abbreviation, abbreviation, and emoticon, and each vocabulary or emoticon, and vocabulary related to emotion and emotion, Emotional vocabulary combinations and their weights are recorded, and profanity can include emotional and emotional intensity.

6 is a diagram showing an example of principal component analysis used in the sensitivity evaluation unit of the social data analysis apparatus according to the present invention.

Principal Component Analysis is one of the methods of multivariate statistical analysis. When the variables related to each other are observed, a small number of new variables (principal components) can be explained to maximize the information they have. It is a way to find out. In FIG. 5, the word 'poojae' has the meaning or emotion through the evaluation result of the emotional lexical pair (adjective pair) below it, and the number of the emotional lexical pair (adjective pair) becomes too much. Therefore, referring to FIG. 6, it is possible to represent slang-based profanity or the like in abbreviated form by two or three pairs through principal component analysis. (It is not simply reduced to two or three of the emotional lexical pairs shown above, but it is abbreviated as a composite adjective pair.

In the case of the word "having a puffy" in FIG. 5,

The first main ingredient: (lagged) - (soft and unformed)

Second main ingredient: (masculine) - (feminine)

In other words, the adjective pair that can explain the most part of the profanity of "poojae" is the first main ingredient, and the first main ingredient, the second main ingredient, is used until the fourth main ingredient if the explanation is insufficient. Through this process, the sensibility of 'poojae' can be expressed as a standard sentiment combination such as 0.8 * (each lucky form) + 0.2 * (masculine), and such data can be stored in the emotional DB 131 do.

The emotional analysis unit 140 extracts a main emotional image after emotional engineering analysis on the emotional measurement object based on the plurality of emotional vocabulary provided by the message analysis unit 120 and the emotional evaluation unit 130.

Emotional engineering analysis refers to analytical methods used in product or brand evaluation in the area of emotional engineering. The evaluation and analysis of the profligate word "poverty" is based on emotional evaluation using emotional engineering techniques. Emotion engineering techniques are generally used to quantitatively measure emotions using the Semantic Differential (SD) method, which presents a large number of pairs of emotional adjectives to the evaluator, (Multi Dimensional Scaling), and the like, the results of the emotion map type are displayed. It has been widely used in automotive instrument panel, interior, housing, cosmetic container, electric car exterior, yacht exterior design, etc. It can also be used for diagnostic evaluation of brand, concept, design and naming, image diagnosis evaluation of objects and people .

The emotional analysis unit 140 combines the emotional vocabulary of the emotional measurement target with the frequency and emotional intensity of the emotional vocabulary and performs sparse principal component analysis (SPCA) The sensitive component is extracted. As shown in FIG. 5, the emotion evaluating unit 130 uses a principal component analysis (PCA) because it performs evaluation without missing portions of various emotional lexical pairs. However, the data collected from the Internet or the SNS is shown in FIG. 5 The emotion analyzing unit 140 extracts the main emotional component using the sparse principal component analysis because evaluation data for a sufficient emotional lexical pair can not be obtained.

The analysis result display unit 150 displays the main emotion component extracted by the emotion analyzing unit 140 and the emotion measurement object on the image space in real time. At this time, the analysis result display unit 150 constructs factor axes of two or three dimensions from the main several emotion components extracted from the emotion analysis unit 140, Or display the analysis result according to the transition of the time during the period in an animation format. Further, it is possible to simultaneously display several sensibility measurement objects on an image space composed of common emotion and emotion components.

7 is a diagram showing an example of the output of the analysis result display unit. 8 is a diagram showing another example of the output of the analysis result display unit. 9 is a diagram showing another example of the output of the analysis result display unit.

Referring to FIGS. 7 and 8, for example, when emotion is specified and analyzed with respect to three models (A, B, and C) of automobiles and two main emotion components are extracted as follows

1) (traditional-modern)

2) (looking uneasy and uncomfortable - looking stable and comfortable)

The three models with the two main emotional components as the X and Y axes are as follows. It can be shown that the emotion changes according to the time. (For example, the emotions of concept cars before the launch of the models, the emotions after seeing and watching the actual models are changed after the launch, etc.) In addition, the change in emotion over time can also be expressed by animation. Figures 7 and 8 are output in 2D form, with ellipses representing the 95% significance level range of the model. Fig. 8 is a screen in which Fig. 7 is output in a simplified information display mode. In addition, referring to FIG. 9, a case in which the main emotional component is extracted as three types (traditional-modern), (unstable and uncomfortable-stable and comfortable-looking), and (masculine-feminine) . Referring to FIGS. 7, 8, and 9, since the evaluation is performed at a specific time point, a change can be confirmed according to a change in time, and a change according to the change in time can be output in an animation form.

Hereinafter, the operation of the method for analyzing social data according to the present invention will be described.

4 is a flowchart illustrating an operation of the method for analyzing social data according to the present invention.

Referring to FIG. 4, the social data analysis method collects and stores data of contents existing in a server group through a communication network (S10), and stores the collected and stored data in the form of a morpheme DB Tagging, and emotional vocabulary is searched (S20). The part-of-speech tagging can not be performed on data that is not stored in the morpheme DB and emotional vocabulary DB. At this time, since the unregistered vocabulary such as profanity, fluency, abbreviation, and emoticon can not be tagged with part-of-speech, it is determined whether there is an unregistered vocabulary that is not part-tagged (S30). Thereafter, if an unregistered vocabulary is found, the unregistered vocabulary is subjected to a sensitivity evaluation, and the standard word sensible vocabulary capable of replacing the unregistered vocabulary is stored in the sensibility DB (S40). If the unregistered vocabulary is not found or the unregistered vocabulary is found, the main sensibility component is extracted based on the sensibility vocabulary provided in steps S20 and S40 (S50). Thereafter, the main extracted emotional components and the sensibility measurement object are displayed on the two- or three-dimensional image space (S60), and the social data analysis method ends.

As described above, the social data apparatus and method according to the present invention are not limited to the configuration and method of the embodiments described above, but the embodiments can be applied to all of the embodiments Or some of them may be selectively combined.

10; Server group
20; communications network
110; The data collecting unit
120; Message analysis section
121; Morpheme DB
122; Emotional Vocabulary DB
123; Input filter module
124; Partial tagging module
125; Unregistered word processing module
130; The emotion evaluation unit
131; Emotional DB
140; Emotional analysis department
150; Analysis result display

Claims (20)

A data collecting unit for collecting and storing data of contents existing in a server group through a communication network;
Analyzing the data in a morpheme unit matched with the morpheme DB and the emotional vocabulary DB, tagging parts of speech by the morpheme unit, extracting emotional vocabulary and storing the sensed vocabulary;
An emotion evaluation unit for performing emotional evaluation on an unregistered vocabulary that is not partly tagged in the message analyzing unit and storing a standard word emotional vocabulary corresponding to the unregistered vocabulary obtained as a result of the emotional evaluation in an emotional DB;
The message analyzing unit and the emotional vocabulary and standard
An emotional analysis unit for extracting a main emotional component, which is a representative emotional component for the emotional measurement object, after the emotional engineering analysis of the sensation target based on the emotional vocabulary; And
And an analysis result display unit for displaying the sensibility measurement object and the main sensibility component on an image space in real time.
The method according to claim 1,
Wherein the data-
Wherein the input unit receives input of a keyword of data or a period of time for collecting data, collects and stores data corresponding to the input, and stores the classified data by the source of the data or the uploader of the data.
The method according to claim 1,
Wherein the data-
A parallel processing technique for collecting and storing data by utilizing a plurality of computers, a multiprocessor or a multi-thread simultaneously is supported, and redundant data among data collected through the parallel processing technique is deleted, and the data is integrated And automatically collecting and storing data after the suspended data when the parallel processing technique is stopped during the process.
The method according to claim 1,
The message analyzing unit,
An input filter module for filtering or repeating short patterns without repetition or repeated use of meaningless special symbols; And
And an unregistered word processing module for transmitting an unregistered vocabulary corresponding to at least one of a profanity word, a fluent word, an abbreviation, and an emoticon that has not been tagged with part-of-speech to the emotion evaluation unit.
The method according to claim 1,
The message analyzing unit,
Characterized in that a parallel processing technique which utilizes a plurality of computers, a multiprocessor or a multi-thread simultaneously is supported.
The method according to claim 1,
The message analyzing unit,
Wherein the system is configured to distinguish emotional and emotional vocabularies based on a morphological database and an emotional vocabulary DB, and to separate and store subjects and objects of the vocabulary related to the emotional and emotional states.
The method according to claim 1,
Wherein the emotion evaluation unit
And the evaluator performs an emotional evaluation using the standard word emotional vocabulary pair, and the evaluator evaluates the standard word emotional vocabulary pair through an online connection.
The method according to claim 1,
Wherein the emotion evaluation unit
Wherein the evaluator selects the standard word that can replace the unregistered vocabulary that has not been tagged with the part-of-speech by a weighted sum of two or more standard words through an online connection.
The method according to claim 1,
Wherein the emotion evaluation unit
Wherein the strength of emotion and emotion for the unregistered vocabulary that has not been tagged with the part-of-speech is directly input or is evaluated through an online questionnaire evaluation.
The method according to claim 1,
Wherein the emotional analysis unit comprises:
Wherein the main sensory component of the sensibility measurement object is extracted by sparse principal component analysis (SPCA) based on a combination of emotional vocabulary of the sensibility measurement object and frequency and emotional intensity of the vocabulary Social data analysis device.
The method of claim 10,
The extraction of the main emotional component,
And a specific time or a specific period is set.
The method according to claim 1,
As a result of the analysis,
The main sensory components extracted from the emotional analysis unit may be configured as two-dimensional or three-dimensional axes, and the position on the image space with respect to a specific time point of the sensibility measurement object may be displayed in real time, Wherein the analysis result is displayed in an animation format.
The method according to claim 1,
As a result of the analysis,
Wherein the plurality of sensibility measurement objects are simultaneously displayed on an image space composed of common emotion and emotion components.
Collecting and storing data of contents existing in a server group through a communication network;
Analyzing the data in a morpheme unit matched with the morpheme DB and the emotional vocabulary DB, tagging parts of speech by the morpheme unit, searching and storing emotional vocabulary;
Performing emotional evaluation on an unregistered vocabulary that is not part-tagged in the detecting and storing step and storing a standard word emotional vocabulary corresponding to the unregistered vocabulary obtained as a result of the emotional evaluation in an emotional DB;
Extracting and storing the main emotional component as a representative emotional component for the emotional measurement object after the emotional engineering analysis on the emotional measurement object based on the emotional vocabulary and the standard word emotional vocabulary provided in the step of storing in the emotional DB step; And
And displaying the sensibility measurement object and the main sensibility component on an image space in real time.
15. The method of claim 14,
Collecting and storing the data comprises:
A plurality of computers, a multi-processor or a multi-thread, and a plurality of computers, a plurality of computers, a multi-processor or a multithreader, for collecting and storing data corresponding to the input, A parallel processing technique for collecting and storing data by utilizing the parallel processing technology is supported, and redundant data among data collected through the parallel processing technology is deleted, the data is integrated by time and type, And automatically collecting and storing the data after the interrupted data when it is interrupted.
15. The method of claim 14,
The step of extracting and storing comprises:
It is possible to distinguish emotional and emotional vocabularies based on morpheme DB and emotional vocabulary DB, to separate subjects and objects of vocabulary related to emotions and emotions, to store them in parallel, to use multiple computers, multi-processors or multi- Processing technology is supported.
15. The method of claim 14,
The step of extracting and storing comprises:
A repetition of a short pattern without spacing or a repetitive use of a meaningless special symbol is excluded or shortened to analyze and an unregistered vocabulary corresponding to at least one of profanity, The method comprising the steps of: storing the sensed data in the emotional DB;
The method according to claim 1,
Wherein the storing in the emotional DB comprises:
And the evaluator is made to evaluate the standard word emotional lexical pair through an online connection, and the evaluator is allowed to perform evaluation of the part marking tagging through the online access, The standard word that can replace the unregistered vocabulary is selected as a weighted sum of two or more standard words and the strength of emotion and emotion for the unregistered vocabulary that is not tagged in the part is directly input or is evaluated through an online questionnaire evaluation A social data analysis method characterized by.
Claim 1:
Wherein the extracting comprises:
(SPCA) based on a combination of the emotional vocabulary of the sensibility measurement object and the frequency and the emotional intensity of the vocabulary, thereby extracting the main sensibility component representative of the sensibility measurement object However,
Wherein the extraction of the main sensible component is performed by setting a specific time point or a specific time period.
The method according to claim 1,
Wherein the displaying comprises:
A plurality of principal emotion components extracted in the extracting step are constituted by two-dimensional or three-dimensional axes, the position on the image space with respect to a specific time point of the sensitivity measurement object is displayed in real time, And displays a plurality of sensibility measurement objects on an image space composed of common emotion and emotion components at the same time.

Images