KR20220065980A - Sentiment analysis apparatus and controlling method thereof, and newly-coined word and emotion extraction apparatus - Google Patents

Abstract

The present invention relates to an emotion analysis apparatus, which comprises: an analysis unit; and a storage unit connected to the analysis unit. The analysis unit extracts a neologism, an emoticon, and an emotional word collected from each of collected social data, calculates a first emotion value for each of the collected social data by using the extracted emotional word, calculates a first total positive and negative emotion value by adding a first positive and negative emotion value among the first emotion values for each of the social data, calculates a second emotion value for each of the social data by using the extracted neologism and emoticon, calculates a second total positive and negative emotion value by adding a second positive and negative emotion value among the second emotion values for each of the social data, calculates a final total positive and negative emotion value for the entire social data by adding the first total positive emotion value and the second total positive emotion value and adding the first total negative emotion value and the second total negative emotion value, and determines a final emotion for the entire social data by comparing the final total positive emotion value and the final total negative emotion value. According to the present invention, it is possible to improve accuracy of emotion determination.

Description

Sentiment analysis device, its control method, and new word and emoticon extraction device

The present invention relates to an apparatus and control method for emotion analysis, and to a device for extracting new words and emoticons, and to an apparatus and control method for emotion analysis for social data, and an apparatus for extracting new words and emoticons.

Due to the development of wired and wireless networks and the development of portable electronic devices, the use of the Internet is possible regardless of time and place.

For this reason, there is an increasing trend to share opinions and communicate emotions using text messages instead of phone calls using voice.

However, when using a social network service (SNS) or the like through a portable electronic device such as a smart phone, the input unit mainly composed of a touch panel is miniaturized and thus inconvenient to use.

Therefore, in order to reduce the number of characters input, the generation and use of new words such as abbreviations or new words that reduce the number of characters of normal words are increasing, and in addition, emoticons (emoticons) replacing words or sentences expressing emotions or feelings ) is increasing day by day.

Since these new words or emoticons are not normal words like standard words, when these new words or emoticons exist in a sentence, it is difficult to analyze the sentence.

Republic of Korea Patent Registration No. 10-2019756 (Announcement date: September 10, 2019, title of invention: online contextual advertisement intelligent device and method based on language analysis for automatic recognition of new words) Republic of Korea Patent Registration No. 10-1120038 (Announcement Date: March 23, 2012, Title of Invention: Device and Method for Selecting New Words) Republic of Korea Patent Registration No. 10-1740274 (Announcement Date: June 08, 2017, Title of Invention: Emoji Search Method and Terminal)

The problem to be solved by the present invention is to perform sentiment analysis on search results using new words and emoticons.

Another problem to be solved by the present invention is to perform the extraction of accurate neologisms.

One feature of the present invention for solving the above problems includes an analysis unit, and a storage unit connected to the analysis unit, wherein the analysis unit is a new word and An emoticon is extracted and stored in the storage unit, an emotional word is extracted from each collected social data, a first emotional value is calculated for each social data collected by using the extracted emotional word, and in each social data The first positive emotion values are summed among the first emotional values for the first negative emotion values to calculate a first positive emotion sum total and stored in the storage unit, and the first negative emotion values among the first emotion values for each social data are summed for a first negative emotion value. The sum total of emotions is calculated and stored in the storage unit, and a second sentiment value for each social data is calculated using the new words and emoticons extracted from each collected social data, and stored in the storage unit, and stored in each social data. A second sum of positive emotions is calculated by summing the second positive emotion values among the second emotional values for each social data, and a second total negative emotion is calculated by summing the second negative emotion values among the second emotion values for each social data. Stored in the storage, the sum of the first positive sensibility and the sum of the second positive sensibility are added to calculate a final total positive sensibility for all of the at least one social data, and the sum of the first negative sensibility and the sum of the second negative sensibility is added to calculate a final total negative emotion for all of the at least one social data, and compares the final total positive emotion with the final total negative emotion to determine a final emotion for all of the at least one social data.

The analysis unit may delete user information, location information, and hashtags from the collected at least one social data before extracting the neologism, emoticon, and sentimental word.

The analysis unit may perform morpheme analysis on each social data, compare the result of the morpheme analysis of each social data with the sentiment dictionary stored in the sentiment dictionary database, and extract sentiment words for each social data.

The storage unit may store polarity weights for current collection coefficients, and the sentiment analysis unit uses the new word appreciation database and the emoticon emotion database to extract each new word and each emoticon from each social data using the polarity, strength, and current Determine a collection coefficient, determine a polarity weight for the current collection coefficient, calculate the emotion value of each new word and each emoticon using the determined polarity, strength, and polarity weight, each new word for each social data and The total positive emotion value and the total negative emotion value for each social data for each job are calculated using the emotion value of the emoticon, and the total positive emotion value and the total negative emotion value of the social data for each job are compared. 2 Sensitivity values can be calculated.

The sentiment analysis apparatus according to the feature may further include a user input unit connected to the analysis unit and inputting a collection keyword, wherein the analysis unit stores the at least one social data corresponding to the collection keyword to a social network server. can be collected from

The control method of the emotion analysis apparatus according to another feature of the present invention includes the steps of extracting new words and emoticons from each social data collected using a new word emotion database and an emoticon emotion database, extracting a sentiment word from each collected social data, calculating a first emotional value for each social data collected by using the extracted emotional word; calculating a first positive emotional sum by summing the first positive emotional values among the first emotional values for each social data; , calculating a first total negative emotion by summing the first negative emotion values among the first emotion values for each social data; Calculating a second emotional value for each social data using the new words and emoticons extracted from each collected social data, adding up a second positive emotional value among the second emotional values for each social data to make a second positive calculating the sum total of emotions and calculating a second sum total of negative sensibility by summing the second negative sensibility values among the second sensibility values for each social data; calculating a final sum of positive emotions for all one social data; calculating a final total negative emotion for all of the at least one social data by adding the first total negative emotion and the second total negative emotion; Including the step of judging the final sentiment in the whole of the social data.

The control method of the emotion analysis apparatus according to the feature may further include deleting user information, location information, and hashtags from at least one collected social data before extracting the new words, emoticons, and emotional words.

The second emotion value calculation step is a step of determining the polarity, strength, and current collection coefficient of each new word and each emoticon extracted from each social data using the new word appreciation database and the emoticon emotion database, and information stored in the storage unit determining a polarity weight for the current collection coefficient using Calculating the emotional value of each new word and each emoticon by using the determined polarity, intensity, and polarity weight, and using the emotional value of each new word and emoticon for each social data, the total positive emotional value for each job social data and calculating a total negative emotion value, and calculating a second emotion value for each social data by comparing the total positive emotion value and the total negative emotion value of the social data for each task.

An apparatus for extracting new words and emoticons according to another feature of the present invention is connected to a storage unit in which social data is stored and the storage unit, compares the social data with the off-dictionary database, and the words existing in the off-dictionary database and an extraction unit that determines at least one Hangul initial consonant as a neologism.

The social data stored in the storage unit may contain Korean.

The extractor may extract an image tag by extracting an image tag from the social data, extract a character encoded in 4-byte Unicode, and extract it as a 4-byte character emoticon.

According to these characteristics, since the overall sensitivity of social data related to the collected keyword is determined using not only emotional words but also new words and emoticons, the accuracy of emotion judgment is improved.

In addition, since creeds and emoticons are extracted from the social network service and registered in the database, the accuracy of sentiment analysis on social data can be improved.

1 is a schematic block diagram of a sentiment analysis system according to an embodiment of the present invention.
2 is a schematic block diagram of a sentiment analysis apparatus according to an embodiment of the present invention.
3 is a flowchart of an operation of a sentiment analysis apparatus according to an embodiment of the present invention.
4 is a detailed operation flowchart of a sentiment analysis routine for the work social data of FIG. 4 .
5 is a schematic block diagram of an apparatus for extracting new words and emoticons according to an embodiment of the present invention.
6 is an operation flowchart of an apparatus for extracting new words and emoticons according to an embodiment of the present invention.
7 is a diagram illustrating a shape of an image type emoticon and an HTML tag therefor.
8 is an example of an output screen for registering new words and emoticons output to the output unit of the apparatus for extracting new words and emoticons according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In describing the present invention, if it is determined that adding a detailed description of a technique or configuration already known in the field may make the gist of the present invention unclear, some of it will be omitted from the detailed description. In addition, the terms used in this specification are terms used to properly express embodiments of the present invention, which may vary according to a person or custom in the relevant field. Accordingly, definitions of these terms should be made based on the content throughout this specification.

The terminology used herein is for the purpose of referring to specific embodiments only, and is not intended to limit the invention. As used herein, the singular forms also include the plural forms unless the phrases clearly indicate the opposite. As used herein, the meaning of 'comprising' specifies a particular characteristic, region, integer, step, operation, element and/or component, and other specific characteristic, region, integer, step, operation, element, component, and/or group. It does not exclude the existence or addition of

Hereinafter, an apparatus for analyzing emotion, a control method, and an apparatus for extracting new words and emoticons according to an embodiment of the present invention will be described with reference to the accompanying drawings.

First, with reference to FIG. 1 , the sentiment analysis system 100 including the sentiment analysis apparatus 10 according to the present example will be described.

As shown in FIG. 1 , the emotion analysis system 100 of this example includes the emotion analysis device 10 and at least one social network server ( 20) and a database unit 30 may be provided.

The sentiment analysis apparatus 10 of this example may perform sentiment analysis on the searched information by using the sentiment information stored in the database unit 30 .

The emotion analysis device 10 of this example will be described in detail next.

The social network server 20 may be a server built by a company that provides social network services, such as Facebook or Twitter.

Accordingly, the social network server 20 may store social data for users (ie, members and followers) who use a social network service provided by a corresponding company. Here, the social data includes user information (eg, user identification information), user postings, hashtags, types of social data, and location information where social data is stored (eg, uniform resource locator (URL)), etc. can be provided Here, the post may include not only a post written by a member but also a comment of a follower.

The type of social data may indicate the type of social network service to which the corresponding social data belongs, and may be, for example, Twitter or Naver Blog.

The social network server 20 may include a control unit for controlling an operation, a communication unit for communication with other devices, and a storage medium for storing social data. In this case, the storage medium may be a storage in which data is stored, such as a database (DB) or memory.

Since the emotion analysis device 10 of this example is connected to the social network server 20 through a network, it is possible to obtain social data of a desired content from the social network server 20 and analyze the emotion on the obtained social data. can

The database unit 30 may include a new word sentiment database 31 , an emoticon sentiment database 32 , a social data database 33 and a morpheme database 34 , a morpheme dictionary database 35 , and a sentiment dictionary database 36 . can

The neologism sentiment database 31 may include items in which a new word, a polarity and intensity for each new word, and a current collection count (count) are stored, respectively.

In this case, the neologism is a non-standard word, and may be an abbreviation of an existing word or sentence, a newly created word, or a letter consisting of a vowel which is at least one consonant.

Polarity represents the image (ie, emotion) of the neologism, and may include positive, negative and mixed. Here, the mixture may mean having both a positive meaning and a negative meaning.

The intensity (intensity) indicates the degree of polarity numerically, and may have a value of 1 to 5.

The polarity and strength of each neologism may vary depending on the overall meaning of the sentence used.

Since the current collection coefficient may be the current exposure frequency of the corresponding neologism, the value of the collection coefficient may change over time.

For example, if the new word is 'Ipdeok', which means 'I became a fan', the polarity of 'Ipdeok' may be positive and the intensity may be 3. In addition, the current collection coefficient of such an Ibdeok may be 569, that is, 569 times.

In addition, as a neologism composed of at least one consonant, 'ㅎ' or 'ㅎㅋ' means 'deuldeol', which is a mimic of 'cold or fear trembles'. And, depending on the number of 'less', it can be expressed as 'ㅎ' or 'ㅎㅋ'.

The polarity of these 'ㅎ' or 'ㅎㅋ' may be positive, the intensity may be 2, and the current collection coefficient may be 205.

The emoticon emotion database 32 may include items for storing an emoticon, a polarity and intensity for each emoticon, and a current collection count, similar to the neologism emotion database 31 .

), 이미지(예,

) 또는 4바이트 유니코드 문자(예,

) 등으로 이루어질 수 있다.An emoticon is a symbol that expresses emotion, and is an ASCII character (eg,

), images (eg,

) or a 4-byte Unicode character (e.g.,

), and so on.

Polarity, strength, and current collection coefficient for these emoticons may have the same meaning as the polarity, strength, and current collection coefficient of the coined word, respectively.

Accordingly, the polarity of the emoticon includes positive, negative, and mixed, and the strength may have a value of 1 to 5.

A method of extracting new words and emoticons for constructing the new word emotion database 31 and the emoticon emotion database 32 will be described in detail next.

The social data database 33 may be a database in which collected social data corresponding to a desired search word according to the operation of the sentiment analysis apparatus 10 is stored.

At this time, the social data stored in the social data database 33 is social data (eg, social data for work) from which unnecessary information is deleted by preprocessing the social data (eg, raw social data) collected by the social network server 20 . can be

The raw social data may be the same as the form stored in the social network server 20, and may include user information, a user's post, a hashtag, a type of social data, and location information (eg, URL) of the social data. there is.

On the other hand, the social data for work in which unnecessary information is deleted from the raw social data may include only the user's postings and types of social data.

The social data database 33 may be composed of items of an original text number (eg, identification number), a channel, and content.

Here, the item of the original number may be a part in which a key value automatically assigned according to the order is stored, the item of the channel may be a part in which the type of social data is stored, and the item of the content may be a part in which the social data for work is stored. .

In this case, the type of social data is a type of social network service to which the social data belongs, and may include, for example, Twitter and a blog.

Accordingly, a key value sequentially automatically assigned to the corresponding social data may be stored in the column of the original number, and the type of the corresponding social data (eg, Twitter or blog) may be stored in the column of the channel. In addition, the content of the social data for work, that is, a post may be stored in the column of the content.

The morpheme database 34 may store a morpheme analysis result for social data for work.

The morpheme dictionary database 35 is a database in which a morpheme dictionary for morpheme analysis is stored, and the morpheme dictionary may include parts-of-speech and utilization information for each morpheme.

The sentiment dictionary database 36 may store sentiment words in the form of morphemes and their attributes and strengths.

In this example, the sentiment dictionary stored in the sentiment dictionary database 36 may use the Korean sentiment analysis corpus (KOSAC) of Kim Hyung-moon (2013) based on the Korean dictionary.

tag_id morphemes polarity robbery 12142 '/SS;Reality/NNG;A/JKO;Wonyong/NNG;Ha/XSV;B/ETM;Story/NNG;'/SS;Ro/JKB;Man/JX;Boy/VV; denial middle 12143 really/MAG;no/VCN;da/EF; Positive height 12144 truth/NNG;of/JKG;opposite/NNG;at/JKB;be/VV; denial middle 12147 Contemplation/NNG; English/VV; Positive middle 12148 humor/NNG; Positive middle

In [Table 1], 'tag_id' may be a unique number of the sentiment dictionary.

The sentiment analysis device 10 is stored in the social network server 20 and extracts social data having a desired content from among the social data, and then analyzes the emotion on each extracted social data using the sentiment dictionary database 36 . , and again, sentiment analysis can be performed on each extracted social data using the new word sentiment database 31 and the emoticon emotion database 33 .

Then, the final sentiment analysis result for the entire extracted social data can be calculated using the sentiment analysis result using the sentiment dictionary and the sentiment analysis result using the neologism and imidicon.

As shown in FIG. 2 , the sentiment analysis device 10 is a user input unit 11 and an analysis unit 13 connected to the user input unit 11 and including an emotion analysis unit 131 and a morpheme analysis unit 132 . ), a storage unit 15 connected to the analysis unit 13 , an information output unit 17 connected to the analysis unit 13 , and a communication unit 19 connected to the analysis unit 13 .

At this time, the components 11-19 shown in FIG. 2 are not essential, so the sentiment analysis apparatus 10 additionally includes at least one other component other than these components 11-19, or Some of them can be omitted.

The user input unit 11 may generate a signal related to a command or a search word for controlling the operation of the emotion analysis apparatus 10 .

Accordingly, the user may input a collection keyword for searching for a post with a desired content in the social network server 20 using the user input unit 11 .

The user input unit 11 may include a key pad, a dome switch, a touch pad, a jog wheel, a jog switch, or a mouse. there is.

The analysis unit 13 is a control unit that controls the overall operation of the sentiment analysis apparatus 10 , and may be a processor.

The analysis unit 13 may include the sentiment analysis unit 131 and the morpheme analysis unit 132 as described above.

The sentiment analysis unit 131 collects social data corresponding to the collected keyword input using the user input unit 11 from the social network server 20, and performs a pre-processing operation on the collected social data to store the collected social data. By removing unnecessary information contained (eg user information, location information and hashtags), social data for work can be created.

In addition, the emotion analyzer 131 may operate the morpheme analyzer 132 to perform morpheme analysis on the collected social data for work, and may extract new words and emoticons from the social data for work.

Accordingly, the emotion analyzer 131 may calculate the first emotion values p1 and n1 for each of the collected social data for work by performing emotion analysis on the social data for work on which the morpheme analysis has been performed. Here, the first sensibility value may include a first positive sensibility value p1 and a first negative sensibility value n1.

Then, the emotion analysis unit 131 calculates a first positive emotion sum pt1 by summing all the calculated first positive emotion values p1, and sums up all the calculated first negative emotion values n1. The first sum total of negative emotions nt1 may be calculated.

Also, the emotion analysis unit 131 may calculate the second emotion values p2 and n2 for each job social data by using the extracted new words and emoticons. The second sensibility value may also include a second positive sensibility value p2 and a second negative sensibility value n2.

In this case, the emotion analysis unit 131 may calculate the second emotion values p2 and n3 by reflecting the current collection coefficients of the new words and emoticons.

Next, the emotion analysis unit 131 calculates a second positive emotion sum pt2 by summing all the calculated second positive emotion values p2, and sums all the calculated second negative emotion values n2 to make the second 2 The sum total of negative emotions (nt2) may be calculated.

For this reason, the first positive emotion sum (pt1) and the first negative emotional sum total (nt1), which are positive and negative emotion values for all social data for work, are calculated using the sentiment dictionary, and new words and emoticons are used for work A second sum of positive emotions pt2 and a second sum of negative emotions nt2 that are positive and negative sentiment values for the entire social data may be calculated.

Next, the emotion analysis unit 131 adds the first sum total of positive emotions (pt1) and the total number of second positive emotions (pt2) to the final sum total of positive emotions (pf) and the total number of first negative emotions (nt1) for all social data for work. and the second sum total of negative emotions (nt2) may be added to calculate the final sum of negative emotions (nf) for all social data for work.

Then, the emotion analysis unit 131 may determine the final emotion for the entire social data for work by comparing the final positive emotion sum (pt) and the final negative emotion sum (nf).

As already described, the morpheme analyzer 132 may perform morpheme analysis on the social data for work collected using the morpheme dictionary database 35 and store the analysis result in the morpheme database 34 .

In this example, the morpheme analyzer 132 is a part of the analysis unit 13 and constitutes a part of the sentiment analysis device 10 , but unlike this, the morpheme analysis unit 132 may be implemented and operated as a device separate from the sentiment analysis device 10 . there is.

The storage unit 15 is a storage medium for storing data necessary for the operation of the analysis control unit 13 or data generated during the operation, and may be a memory such as a hard disk or a ROM.

In an alternative example, the sentiment analysis apparatus 10 may operate in association with a web storage that performs a storage function of the storage unit 15 on the Internet.

The output unit 17 is for generating an output related to time under the control of the analysis unit 13 , specifically, the emotion analysis unit 131 , and may include a display module.

The display module may display an image corresponding to the image data output from the analysis unit 13 on the screen according to the operation of the analysis unit 13 .

The display module may include at least one of a liquid crystal display, an organic light emitting diode display, a flexible display, and a 3D display.

The communication unit 19 is for communication with external devices such as the social network server 20 and the database unit 30, and may include at least one communication module that enables communication between networks in which the external devices are located. .

For example, the communication unit 19 may include an Internet module.

The Internet module refers to a module for wireless or wired Internet access, and Internet technologies such as WLAN (Wireless LAN) (WiFi) or Wibro (Wireless broadband) may be used.

Next, an operation of the sentiment analysis system 100 having such a structure will be described with reference to FIG. 3 .

For sentiment analysis, a user must first collect social data related to a topic desired by the user, ie, social data having contents desired for sentiment analysis, from the social network server 20 .

Accordingly, when the operation of the emotion analysis device 10 starts for sentiment analysis on social data, the emotion analysis unit 131 of the emotion analysis device 10 searches for input of a collection keyword to the information output unit 17 . A window may be output (S11).

Accordingly, the user may input a keyword (ie, a collection keyword) related to social data of a subject to be collected into the search box output to the information output unit 17 using the user input unit 11 .

When a signal is input from the user input unit 11 by the user's operation, the sentiment analysis unit 131 may read the input signal to determine the word input by the user to determine the input collection keyword ( S12).

In this case, the number of collection keywords input by the user may be one or more.

As an example, the collection keyword input by the user may be 'U20', 'Soccer', 'World Cup', and 'Portugal Korea'.

When the collected keyword input through the user input unit 11 is determined, the sentiment analysis unit 131 accesses the social network server 20 through the communication unit 19 and collects keywords determined as the connected social network server 20 . can be transmitted (S13).

At this time, the sentiment analysis unit 131 transmits not only the collected keyword but also its own identification information, so that the social network server 20 can recognize the device that transmitted the collected keyword.

The social data collection operation by the sentiment analysis unit 131 may use a search API provided by a social network service or a separate crawler.

When the collected keyword is transmitted from the sentiment analysis device 10 by this operation, the social network server 20 accesses the storage medium and social data corresponding to the collected keyword among the social data stored in the storage medium (eg, database). can be searched for.

Then, the social network server 20 may transmit the collected social data (eg, raw social data) to the emotion analysis device 10 through the communication unit 19 .

When raw social data related to a collection keyword is collected from the social network server 20 and transmitted through the operation of the social network server 20 , the emotion analysis unit 131 of the emotion analysis device 10 receives each of the transmitted raw social data By performing a pre-processing operation on , social data for work may be generated and stored in the social data database 33 (S14).

The pre-processing operation for social data may be an operation of deleting information unnecessary for sentiment analysis from among the collected raw social data information.

In one example, the sentiment analysis unit 131 may delete user information (eg, user identification information), location information (eg, URL), and a hashtag from each of the transmitted raw social data.

Due to the pre-processing operation of the emotion analysis unit 131 , the social data for each task may include a post and the type of social network service to which it belongs (eg, Twitter or blog).

In this case, if other information unrelated to a post exists in the transmitted social data other than user information, location information (URL), and hashtag, this information may also be deleted in the pre-processing step.

[Table 2] shows an example of social data for work in which the pre-processing operation has been completed.

raw social data social data for work RT @parfaitfemi: Wow, he's so handsome that he thinks he's watching a movie instead of the Olympics.. #PyeongChangWinterOlympics - 1500m short track men's finalist *** https://t.co/LWPRXGyiBq RT Wow, he's so handsome, he thought he was watching a movie instead of the Olympics.. #PyeongChangWinterOlympic -Short track 1500m men's finalist *** RT @1theleft: #PyeongChangOlympic #PeaceOlympic #DailyMoon https://t.co/X8TPswXKWH Looking at the largest number of athletes in the history of 2,925 athletes from 92 countries supporting TV by 5 billion people around the world RT TV support of 5 billion people worldwide 2,925 athletes from 92 countries. #Pyeongchang Olympics #PeaceOlympic #DailyMoon, the largest number in history

In this way, when social data for work by sentiment analysis is generated, the sentiment analysis unit 131 of the analysis unit 13 operates the morpheme analysis unit 132 to perform morpheme analysis on the social data for each job. .

Accordingly, when a control command is applied by the emotion analyzer 131 , the morpheme analyzer 132 starts an operation, and uses the morpheme dictionary database 35 to store social data for each task stored in the social data database 33 . By performing morpheme analysis on the morpheme, the morpheme analysis result for each job social data may be stored in the morpheme database 34 (S15).

A morpheme analysis operation performed in this example may use a known morpheme analysis method, for example, a komoran morpheme analyzer provided in the form of a Java library may be used.

An example of the morphological analysis result is shown in [Table 3] below.

social data for work morphological analysis result The strongest idol! Black Venus! strongest/NNG;i/NNG;dol/NNG;!/SF;black gold/NNP;sung/XSN;!/SF; I don't like male idols, but I think NCT is good. Man/NNG;Ki/NNG;Dol/NNG;Like/VV;G/EC;B/JX;NOT/VX;Have/EC;Ensi/NNG;T/NNG;A/JX;Mouth/NNG;Duck/ NNG;Till/JX;B/JX;No/VCN;High/EC;,/SP;,/SP;Song/NNG;Ga/JKS;Good/VA;High/EC;Dance/NNG;Lee/JKS; ok/va;ah/EC;

In [Table 3], a tag located after a separated morpheme such as NNG, NNP, XSN, SF, etc. may be a morpheme part-of-speech tag, and in the Cormoran morpheme analyzer, NF (noun estimation category), NV (verb estimation category) and NA (inability to analyze) can be excluded from analysis.

In this way, when the morpheme analysis of the social data for work is completed, the morpheme analysis unit 132 outputs a morpheme analysis completion command to the sentiment analysis unit 131 to notify the emotion analysis unit 131 of the morpheme analysis completion operation. there is.

Accordingly, the sentiment analysis unit 131 may perform an operation of extracting new words and emoticons from the social data for work stored in the social data database 33 (S16)

Therefore, the sentiment analysis unit 131 compares the emoticons stored in the neologism and emoticon sentiment database 33 stored in the neologism sentiment database 31 with the social data for work stored in the social data database 33, After extracting new words and emoticons included in the social data for work, it may be stored in the storage unit 15 .

For example, if the social data for work is 'I don't like male idols, but NCT is good, I don't like it, I like singing and dancing', there are no new words or emoticons in this social data for work.

On the other hand, the social data for work said, ‘Oh, that’s right, the first stage was an idol wearing a uniform, but I just painted it all on. A Korean comeback hahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahaha In the case of 'I'm really pissed off ㅠㅠ', the new words and emoticons included in the social data for this work are 'Bala', 'ㅜㅜ', 'Magic' , 'ㅠ' , 'I'm crazy', 'Army', 'ㅋ' , 'ㅠㅜ' or 'TTㅠ'.

In this way, when the morphological analysis of the social data for work and the extraction of new words and emoticons are completed, the emotion analysis unit 131 determines whether the emotion of the social data (ie, social data for work) related to the collected keyword is 'positive', ' A sentiment analysis operation to determine whether it is 'negative' or 'neutral' may be performed (S17).

In this example, the sentiment analysis operation may be performed through two analysis operations.

For example, sentiment analysis using a sentiment dictionary and sentiment analysis using extracted new words and emoticons are performed, and the final sentiment of all social data related to the collected keywords can be analyzed by synthesizing these two sentiment analysis results. there is.

The sentiment analysis step S17 of the sentiment analysis unit 131 may be subdivided as shown in FIG. 4 .

Accordingly, as shown in FIG. 4 , the sentiment analysis unit 131 compares the sentiment words stored in the sentiment dictionary database 36 and the morphemes of the social data for each job stored in the morpheme database 35 and uses them for each job. It is possible to extract sentiment words belonging to the social data (S171).

Next, the sentiment analysis unit 131 may calculate the first sentiment values p1 and n1 for each social data for each job by using the polarity and intensity corresponding to the extracted sentiment word of the social data for each job (S172). .

To this end, the sentiment analysis unit 131 may calculate a sentiment value for each sentiment word extracted from the social data for each job ( S172 ).

As already described, the polarity for each sentiment word may be one of positive, negative, and neutral, and the intensity may be one of high, medium, and low, as already described. In addition, a value according to each intensity is already stored in the storage unit 15 , for example, a high value is '3', a medium value is '2', and a low value is '1'. can be

Therefore, as an example, when the emotional word is 'fantasy', the polarity of this illusion is positive and the intensity is high. If the emotional word is stored in the dictionary, 3, the value corresponding to high, is the emotional value of 'phantom', and Kind is positive. Therefore, 'fantasy' has a positive emotional value of 3.

As another example, when the emotional word is 'embarrassed', the polarity of this embarrassment is negative and the intensity is medium. it is negative Accordingly, 'embarrassed' has a negative emotional value of 2.

In this way, when an emotional value such as a positive emotional value or a negative emotional value for each emotional word separated from each work social data is calculated (S172), the emotion analysis unit 131 uses the calculated emotional value of each emotional word. Thus, it is possible to calculate a total positive emotional value and a total negative emotional value for the social data for the corresponding task (S173).

Then, the emotion analysis unit 131 may calculate the first emotional values p1 and n1 for the social data for the work by comparing the calculated total positive emotion values and the total negative emotion values of the social data for the corresponding work. (S174). Here, the first emotion values p1 and n1 may have a first positive emotion value p1, a first negative emotion value n1, and a neutral emotion value having a value of '0' according to the type of emotion.

That is, when the total positive emotional value of the social data for the corresponding task is greater than the total negative emotional value, the emotion analysis unit 131 determines that the type of the first emotional value for the social data for the task is positive, and on the contrary, the total negative emotional value is the total negative emotional value. When it is greater than the positive emotional value, the type of the first emotional value for the corresponding work social data may be negative. In addition, if the total positive emotion value and the total negative emotion value of the social data for the corresponding job are the same, the type of the emotion value for the social data for the corresponding job may be neutral. In addition, the determined first emotional value for the sensibility of the corresponding work social data may be an absolute value of the difference between the total positive sensibility value and the total negative sensibility value.

For example, if the total positive emotion value for one job social data is '5' and the total negative emotion value is '1', the emotion type for this job social data is positive, and the emotion value is 4 (=5 -1) can be Accordingly, the first emotional value per unit of the social data for the corresponding task may have a first positive emotional value n1 of 4.

In this way, the sentiment analysis unit 131 may calculate the first sentiment values p1 and n1 for each of the social data for all tasks by using the sentiment dictionary and store the first sentiment values p1 and n1 in the storage unit 15 . As already described, each social data for work may have a first positive emotion value p1 having a corresponding value, a first negative emotion value n1, or a neutral emotion value having a value of 0.

Then, the emotion analysis unit 131 sums up all the calculated first positive emotion values p1 to calculate a first positive emotion sum total (pt1) for all work social data, and the calculated first negative emotion values ( n1) is summed to calculate the first sum total of negative emotions (nt1) for all social data for work, and the first sum total of positive emotions (pt1) and the total number of first negative emotions (nt1) can be stored in the storage unit 15 There is (S175).

Next, the sentiment analysis unit 131 may calculate the second sentiment values p2 and n2 for each job social data by using the new word and imidicon extracted from the social data for each job, and store it in the storage unit 15 . The second sensibility value may also include a second positive sensibility value p2 , a second negative sensibility value n2 , and a neutral sensibility value.

To this end, the emotion analysis unit 131 uses the new word emotion database 31 and the emoticon emotion database 32 to extract the social data for each job and store the polarity for the new word and the emoticon stored in the storage unit 15, It is possible to determine the corresponding intensities (1 to 5) and the current collection coefficient (S176).

Next, the sentiment analysis unit 131 may determine the polarity weight by using the current collection coefficient for the new word or emoticon. In this case, the polarity weight may already be determined in the storage unit 15 according to the value of the current collection coefficient, and as the value of the current collection coefficient increases, the value of the polarity weight may also increase. For example, the minimum value of the polarity mask may be '1', and the maximum value may be '10'.

In this example, the emotional value of each new word or emoticon may be a value obtained by multiplying intensity by a polarity weight.

For this reason, even if it is a new word or emoticon having the same polarity and the same strength, the sensitivity value of the new word or emoticon having a large current collection coefficient is large.

For example, the polarity of the coined word 'Ibdeok' is positive, the intensity is '3', and the current collection coefficient is '569'. ' Assume

In this case, when no polarity weight is applied, the second sensibility values of Ibdeok and the youngest Ontop both have a second positive sensibility value of '3' equal to the intensity.

However, as in this example, it is assumed that the polarity weight is applied, the current collection coefficient is between 1 and 100, the polarity weight is '1', the current collection coefficient is between 400 and 600, and the polarity weight is '2'. In this case, Ibdeok has a second positive emotional value of '3[=3 (strength) × 2 (polar weight)], while the youngest Ontop has a second positive emotional value of '3 (= 3 × 1)'. do.

In this way, the emotion analysis unit 131 uses the polarity weights according to the polarity, intensity, and current collection coefficient of the new word and emoticon of the social data for each task, and the emotion value of the new word and the emoticon of the social data for the corresponding task (that is, positive emotion). value or negative emotional value) can be calculated (S177).

Next, the emotion analysis unit 131 may calculate a total positive emotion value and a total negative emotion value for the social data for the corresponding work by using the calculated emotion values of each new word and emoticon (S178).

Then, the emotion analysis unit 131 may calculate the second emotion values p2 and n2 for the social data for the corresponding job by comparing the calculated total positive emotion value and the total negative emotion value of the social data for the corresponding job. (S179). Here, the second emotion values p2 and n2 may also have a second positive emotion value p2, a second negative emotion value n2, and a neutral emotion value having a value of '0' according to the type of emotion.

The method of calculating the second emotion values p2 and n2 for the social data for the corresponding work is the same as the method of calculating the first emotion values p1 and n1, and the emotion analysis unit 131 calculates the total positive emotion of the social data for the work. By comparing the value and the total negative emotion value, the type of emotion and the size of the emotion value may be calculated.

Then, the emotion analysis unit 131 sums up all the calculated second positive emotion values p2 to calculate a second positive emotion sum total pt2 for all social data for work, and the calculated second negative emotion values ( After calculating the second total negative sentiment nt2 for all social data for work by summing all n2), the calculated total sum pt2, nt2 may be stored in the storage unit 15 (S1710).

Next, the emotion analysis unit 131 adds the first sum total of positive emotions (pt1) and the total number of second positive emotions (pt2) to the final sum total of positive emotions (pf) and the total number of first negative emotions (nt1) for all social data for work. and the second sum total of negative emotions (nt2) are added to calculate the final sum total of negative emotions (nf) for all social data for work, and then the calculated final sum total of negative emotions (pf, nf) can be stored in the storage unit 15 ( S1711).

Then, the sentiment analysis unit 131 may compare the final sum total of positive emotions (pt) and the final sum of negative emotions (nf) to determine the final sentiment for the entire social data for work ( S1712 ).

For example, if the final sum of positive emotions (pt) is greater than the final sum of negative emotions (nf), the overall sentiment for the social data for the task is positive, and conversely, the final sum of negative emotions (nt) is higher than the final sum of positive emotions (pf). If it is large, the overall sentiment toward the social data for the corresponding task may be negative. In addition, the sentiment value for the entire social data for the corresponding work may be an absolute value of the difference between the final sum total of positive emotions (pt) and the final sum of negative emotions (nf), similar to the previously described ones.

In addition, if the final sum total of positive emotions (pt) and the final sum of negative emotions (nf) are the same, the overall emotion for the social data for the corresponding work is neutral, and the emotion value may be '0'.

Then, as shown in FIG. 3 , the emotion analysis unit 131 may output the determined final emotion to the output unit 17 .

Accordingly, the user can accurately and easily determine the public's sensibility for products, such as products, articles, entertainers, etc. related to the corresponding collection keyword, using the result output from the output unit 17 .

In addition, since the overall sentiment of posts related to the collected keyword is determined using not only the sentiment dictionary but also new words and emoticons, the accuracy of sentiment determination is improved. can be even higher.

Next, with reference to FIGS. 5 and 6 , a method of constructing a new word emotion database 31 and an emoticon emotion database 32 by extracting new words and emoticons on a social network service will be described.

As already described, the new words stored in the new word sentiment database 31 and the emoticons stored in the emoji ken emotion database 32 are collected on a social network service (SNS).

An example of a device for extracting new words and emoticons 50 that extracts these new words and emoticons and updates new words and emoticons in the corresponding databases 31 and 32 is shown in FIG. 5 .

As shown in FIG. 5 , the device for extracting new words and emoticons 50 includes a user input unit 51 , a management unit 53 connected to the user input unit 51 , and a storage unit 55 connected to the management unit 53 . ), an output unit 57 and a communication unit 59 may be provided.

The user input unit 51 , the storage unit 55 , the output unit 57 and the communication unit 59 are the user input unit 11 , the storage unit 15 , the output unit 17 and the communication unit ( ) described with reference to FIG. 2 , respectively. 19), and for this reason, a detailed description thereof may be omitted. However, the storage unit 55 of this example may be a large-capacity storage medium such as a database as well as a memory or web storage.

The management unit 53 may be a control unit that controls the overall operation of the extraction device 50 , and may include a collection unit 531 and an extraction unit 532 .

The collection unit 531 may collect and manage collection data for collection of new words and emoticons, ie, social data, on a social network service.

The extraction unit 552 may extract new words and emoticons from the data collected by the collection unit 531 and control the storage operation in the neologism database 31 and the emoticon database 33 .

To understand this, the extraction device 50 of this example is not shown in FIG. 5 , but may be connected to some databases (eg, 31 , 32 , 35 ) of the database unit 30 .

As an example, the management device of this example may collect data for extracting new words and emoticons from social data stored in Twitter and Naver blogs.

Accordingly, new words, text emoticons, and 4-byte text emoticons can be extracted from Twitter, and image-type emoticons can be extracted from Naver Blog.

In the case of Twitter, since a search API is provided, in the case of Twitter, desired social data among social data of Twitter can be searched using the search API provided by Twitter. In this case, as a search condition, tweets in which the writing language of the tweets is Korean may be collected.

However, in the case of a Naver blog, desired social data can be searched from among the social data of the Naver blog by using a crawler that has already been published. In this case, the search category of the Naver blog may be limited to 'entertainment/art', and an operation of collecting social data of the corresponding category may be performed at regular intervals.

Accordingly, the operator can input conditions necessary for the search, such as a desired search condition and a search period, by using the user input unit 51 , and when a corresponding command is input from the user input unit 51 by the operator's operation, the collection The operation of the unit 531 may be started.

When the operation starts, the collection unit 531 accesses the corresponding server of the corresponding social network service connected through the communication unit 59 according to a command input from the user input unit 51 to collect social data that meets the search conditions. can be (S51).

As already described, the social data collected from the servers of corresponding social network services such as Twitter and Naver blog contain information irrelevant to the extraction operation of new words and emoticons.

Accordingly, the collection unit 531 may perform a pre-processing operation of deleting unnecessary information from the collected social data, and then store it in the storage unit 55 ( S52 and S53 ).

By the pre-processing operation of the collection unit 531, it is possible to improve the accuracy of extraction of new words and emoticons and shorten the processing time.

The information deleted by the collection unit 531 may be location information (eg, URL) of social data, user information, and a hashtag. In the case of Naver Blog, image-type emoticons can be composed of HTML tags, so no content is required when extracting image-type emoticons. Accordingly, the collection unit 531 may perform a pre-processing operation of removing all content and tags except for the <IMG> tag from the social data collected from the Naver Blog.

As such, when the collection of social data on the social network service is completed by the extraction unit 531 , the extraction unit 531 may operate to extract new words and emoticons from the pre-processed social data.

In order to extract a neologism from the preprocessed social data, the extraction unit 532 may use the open dictionary database 61 constructed using at least one open dictionary such as Naver Open Dictionary (eg, Naver Open Dictionary).

This open dictionary database 61 stores an off-dictionary including words that receive evaluations of more than a set number (eg, 10) of 'Korean' and 'like' among words registered in open dictionaries such as Naver Open Dictionary. there may be

Accordingly, the extraction unit 532 compares the preprocessed social data of the storage unit 55 with the words stored in the open dictionary database 61, and among the words of the social data, the same words as the words stored in the open dictionary database. can be determined as a neologism and stored in the storage unit 55 (S54).

In addition, the extraction unit 532 may perform an operation of extracting emoticons from the preprocessed social data (S55-S57), and the extracted new words and emoticons may be stored in the storage unit 55 (S58).

Emoticons are emoticons in the form of characters (eg, character emoticons) such as '^^' or 'ㅠ', emoticons in the form of images (eg, image emoticons), and emoticons in the form of 4-byte Unicode characters (eg, 4 bytes character emoticons).

Therefore, in order to extract a text emoticon from the preprocessed social data (S55), the extraction unit 532 separates the preprocessed social data into word units and then tokenizes the word words, that is, word tokenization. can do. In this case, the boundary between the word word and word word may be determined based on spacing.

The extraction unit 53 may remove all alphabets, complete Hangul, numbers, and 4-byte characters from the tokenized word word.

In addition, since it is rare that only one initial consonant is used in the text emoticon, and it is highly likely that a single initial consonant is a typo, the extraction unit 532 may remove the Hangul consonant character from the tokenized word word.

For example, the extraction unit 532 may extract only "^^" as a neologism in the case of "steering ^^", and extract only "ㅜㅜ" as a neologism in the case of "chiin gundi ㅜㅜ".

At this time, since the character emoticon often consists of special characters or Korean consonants, the character emoticon is divided into new words and can be extracted through the operation of extracting new words using the open dictionary database as previously described (S54).

In order to extract an image emoticon from the pre-processed social data ( S56 ), the extraction unit 532 may extract an image tag (eg, <IMG>) tag from the pre-processed social data. In this case, since there is a unique pattern for the image type emoticon for each social network service company, the extractor 532 may extract that the corresponding pattern exists in HTML as an image type emoticon.

For example, in FIG. 7 , when an alt attribute has a value of 'sticker image' in an emoticon tag in the form of an image, it may be extracted as an <IMG> tag.

In order to extract the 4-byte character emoticon (S57), the extraction unit 532 may extract 4-byte Unicode-encoded characters from the preprocessed social data to determine the 4-byte character emoticon.

New words and emoticons extracted through the operation of the extraction unit 532 may be registered in the neologism emotion database 31 and the emoticon emotion database 32 through the operator's operation using the user input unit 51, respectively (S59) .

At this time, the operator uses an input tool operated under the control of the extraction unit 532 to register the new word in the database 31 and 32 as well as the related polarity (at least one of positive, negative and neutral) and weight (1) A value of ~5) and the current collection coefficient may be input together and stored in the corresponding databases 31 and 32 .

In this case, the input tool may include an additional keyword input function, a negative keyword input function, and a dictionary front and back space input function.

The additional keyword input function is a function that can additionally register keywords in the dictionary. The basic word of a neologism, that is, a proposition or a noun is added to the basic neologism, so that the expanded neologism extended from the basic neologism can be registered.

For example, if the basic neologism is 'Hannam', if 'deul', 'chung' and 'nom' are input as additional keywords through the user input unit using the additional keyword function, the neologism dictionary database 31 is As for related neologisms, not only 'Hannam', which is a basic neologism, but also 'Hannams', 'Hannamchung' and 'Hannamnom' can be registered as neologisms.

The negative keyword input function is the opposite of the additional keyword function, and it is a function to register a part of a sentence including a word marked the same as a new word so that the word included in the part of the registered sentence is not searched as a new word.

Therefore, it is to prevent a word that is written the same as a neologism but has a completely different meaning from being registered as a neologism.

For example, 'what do' is an abbreviation for 'what do', and is a neologism. However, in the sentence of 'it can't be helped', 'it' is the same as the new word 'it', but it has a completely different meaning. Therefore, if 'it can't be done' is registered as a negative keyword, 'dot' of 'can't be helped' is not searched as a new word.

The dictionary space input function adds a space to at least one of the front part and the rear part of a dictionary (ie, a neologism or an emoticon) so that the dictionary can be searched like a space.

For example, the name of BTS' fan club is 'Army'. However, if a new word is registered as 'army', 'army' will be searched for even in words or sentences that have no meaning at all with BTS, such as 'I'm sorry'. Therefore, if a space is added to at least one of the front and rear parts of ARMY, only 'Army' related to BTS is searched for 'Ama', 'Army', or 'Army', so that the accuracy of the search can be improved.

An example of an output screen of the output unit 57 for registering such new words and emoticons may be as shown in FIG. 8 .

As shown in Fig. 8, on the registration screen, a new word to be registered and a description thereof are displayed (B11), a type of polarity selection part (B12), a weight (ie, strength) selection part (B13), an additional keyword input It may include a part B141, a negative keyword part B15, and a blank part B16 before and after the dictionary.

Accordingly, the operator can use the user input unit 51 such as a keyboard to perform the type of emotion (that is, the polarity), the value of the weight, the additional keyword input function, the negative keyword input function, and the dictionary front and back space input function.

The following [Table 4] shows an example of storing the new word sentiment database 31 and the emoticon emotion database 32.

No word polarity robbery Add keywords Exclude keywords leading blank back blank One cutie Positive 5 0 0 2 TT.TT mix 2 0 0 3 bala Positive 3 ballad, ballad 0 0 4 □ Positive 5 0 0 5 https://storep-phinf.pstatic.net/ogq_56cdf4f9d81cd/original_11.png?type=p50_50 Positive 3 0 0

The technical features disclosed in each embodiment of the present invention are not limited only to the corresponding embodiment, and unless they are mutually incompatible, the technical features disclosed in each embodiment may be combined and applied to different embodiments.

In the above, embodiments of the present invention have been described. The present invention is not limited to the above-described embodiments and the accompanying drawings, and various modifications and variations will be possible from the point of view of those of ordinary skill in the art to which the present invention pertains. Accordingly, the scope of the present invention should be defined not only by the claims of the present specification, but also by those claims and their equivalents.

10: sentiment analysis device 20: social network server
30: database unit 31: new word sentiment database
32: Emoji Sentiment Database
33: social data database
34: morpheme database
35: Morphological Dictionary Database
36: Sentiment Dictionary Database
11, 51: user input unit 13: analysis unit
131: sentiment analysis unit 132: morpheme analysis unit
15, 55: storage unit 17, 57: output unit
19, 59: Ministry of Communications

Claims (11)

analysis unit; and
a storage unit connected to the analysis unit
including,
The analysis unit is
New words and emoticons are extracted from each social data collected using a new word emotion database and an emoticon emotion database and stored in the storage unit,
extracting sentiment words from each collected social data, and calculating a first sentiment value for each collected social data using the extracted sentiment words;
A first positive emotion sum is calculated by summing the first positive emotion values among the first emotion values for each social data, and stored in the storage unit, and the first negative emotion value among the first emotion values for each social data is summed to calculate the first total negative emotion and store it in the storage unit,
Using the new words and emoticons extracted from each collected social data, a second emotion value for each social data is calculated and stored in the storage unit,
A second sum total of positive emotions is calculated by summing the second positive emotion values among the second emotion values for each social data, and the second negative emotion values are summed among the second emotion values for each social data to sum up the second negative emotion. Calculated and stored in the storage unit,
calculating a final sum of positive emotions for all of the at least one social data by adding the sum of the first positive emotions and the sum of the second positive emotions;
calculating a final sum of negative emotions for all of the at least one social data by adding the total sum of the first negative emotions and the total number of the second negative emotions;
Comparing the total sum of the final positive emotions and the total sum of the final negative emotions to determine the final emotion in the entire at least one social data
Sentiment Analysis Device. According to claim 1,
The analysis unit deletes user information, location information, and hashtags from the collected at least one social data before extracting the neologism, emoticon, and sentimental word. According to claim 1,
The analysis unit performs morpheme analysis on each social data, compares the morpheme analysis result of each social data with the sentiment dictionary stored in the sentiment dictionary database, and extracts sentiment words for each social data. . According to claim 1,
The storage unit stores the polarity weight for the current collection coefficient,
The sentiment analysis unit,
Determining the polarity, strength, and current collection coefficient of each new word and each emoticon extracted from each social data using the neologism appreciation database and the emoticon sentiment database, and determining a polarity weight for the current collection coefficient,
Using the determined polarity, strength, and polarity weight, calculate the emotional value of each new word and each emoticon,
Calculate the total positive emotional value and the total negative emotional value for each social data for each task by using the emotional value of each new word and emoticon for each social data,
Comparing the total positive emotion value and the total negative emotion value of the social data for each task to calculate a second emotion value for each social data
Sentiment Analysis Device. According to claim 1,
It is connected to the analysis unit, further comprising a user input unit for inputting a collection keyword,
The analysis unit collects the at least one social data corresponding to the collection keyword from a social network server. extracting new words and emoticons from each social data collected using a new word emotion database and an emoticon emotion database;
extracting sentiment words from each collected social data, and calculating a first sentiment value for each collected social data using the extracted sentiment words;
A first total positive emotion is calculated by summing the first positive emotion values among the first emotion values for each social data, and the first negative emotion values are summed among the first emotion values for each social data to sum up the first negative emotion calculating ;
calculating a second emotion value for each social data using the new words and emoticons extracted from each collected social data;
A second sum total of positive emotions is calculated by summing the second positive emotion values among the second emotion values for each social data, and the second negative emotion values are summed among the second emotion values for each social data to sum up the second negative emotion. calculating ;
calculating a final sum total of positive emotions for at least one entire social data by adding the sum of the first positive emotions and the sum of the second positive emotions;
calculating a final total negative emotion for all of the at least one social data by adding the first total negative emotion and the second total negative emotion; and
Comparing the total sum of the final positive emotions and the total sum of the final negative emotions to determine the final emotion in the entire at least one social data
A control method of a sentiment analysis device comprising a. 7. The method of claim 6,
Before extracting the new words, emoticons, and emotional words, the method further comprising the step of deleting user information, location information, and hashtags from the collected at least one social data. 7. The method of claim 6,
The second emotional value calculation step is,
determining the polarity, strength, and current collection coefficient of each new word and each emoticon extracted from each social data using the new word appreciation database and the emoticon emotion database;
determining a polarity weight for the current collection coefficient using information stored in a storage unit;
calculating emotional values of each new word and each emoticon by using the determined polarity, intensity, and polarity weight;
calculating a total positive emotion value and a total negative emotion value for each work social data by using the emotion values of each new word and emoticon for each social data; and
Comparing the total positive emotion value and the total negative emotion value of the social data for each task, calculating a second emotion value for each social data
A control method of a sentiment analysis device comprising a. a storage unit in which social data is stored; and
An extracting unit connected to the storage unit, comparing the social data with an off-dictionary database, and determining a word existing in the off-dictionary database and at least one Hangul consonant as a neologism
A device for extracting new words and emoticons that includes. 10. The method of claim 9,
The social data stored in the storage unit is a coined word and emoticon extraction device containing Korean. 10. The method of claim 9,
The extracting unit extracts image tags from the social data to extract image-type emoticons, and extracts 4-byte Unicode-encoded characters and extracts them as 4-byte character-type emoticons.

Images