KR20190136337A - Social Media Contents Based Emotion Analysis Method, System and Computer-readable Medium - Google Patents

Abstract

The present invention relates to a method, a system and a computer-readable medium for social media content based emotion analysis and, more specifically, to a method, a system and a computer-readable medium for social media content based emotion analysis, which can detect an event from social media content, analyze the event according to regions, and analyze and display emotion of the social media content on the event. According to the present invention, the emotion on the event can be analyzed by extracting the event from the social media content generated in real time.

Description

Social Media Contents Based Emotion Analysis Method, System and Computer-readable Medium}

The present invention relates to a social media content based emotion analysis method, system and computer-readable medium, and more particularly, to detect an event from the content of the social media, to analyze the event according to the region and to the social media for the event Social media content based emotion analysis method, system and computer-readable medium for analyzing and displaying the emotion of the content.

As the use of social media sites increases, sites such as Twitter and Instagram are playing various roles beyond simply serving as a space for uploading contents. Social media sites with various characteristics allow users to freely generate large and small contents such as social issues and natural disasters from trivial daily lives without time and space constraints as long as they are connected to the Internet. In recent years, the influence of social media sites is getting stronger, it is possible to obtain information about small events or disasters through social media sites.

Due to the influence of social media sites and ripple effects, various researches are being conducted to analyze the contents generated online. Social media content analysis is performed by various methods with various purposes. Among many content, especially user-written text contains the most direct and important information. Since the content is created by the intention of the user, the creation time of the content is also an important factor in the content analysis.

Among social media sites, Twitter has a very high level of communication about information on social issues, disasters, accidents, etc., because information is delivered in short sentences that do not exceed 140 characters, and anyone can easily obtain and spread the information. Analyzing Twitter with a focus on events such as social issues and disasters enables us to identify events that occurred and to analyze trajectories such as propagation paths of identified events. Moreover, when people's feelings in various regions of the related event are analyzed, a lot of information about the specific event can be obtained. By analyzing the emotional path, it is possible to analyze the emotions according to the place and time, and also to predict the emotions of the users.

The present invention provides a social media content based emotion analysis method, system and computer-readable medium for detecting an event from contents of social media, analyzing the event according to a region, and analyzing and displaying the emotion of social media content for the event. To provide that purpose.

In order to solve the above problems, the present invention, social media content-based emotion analysis method, the event determination step of extracting the event from the content of the social media in real time to determine the keyword for the event; A regional keyword analysis step of performing regional and hourly analysis of contents including the keyword from the contents of the social media; A trajectory analysis step of analyzing the trajectory of the event from regional and hourly analysis results of the content including the keyword; And an emotion analysis step of analyzing the emotion of the content from the content including the keyword. It provides a social media content-based emotion analysis method comprising a.

In the present invention, the emotion analysis step, the emotion can be analyzed by the learned artificial neural network.

In the present invention, the artificial neural network is a convolutional neural network, the content including the keyword may be vectorized and input to the convolutional neural network.

In the present invention, the event determination step, the seedword expansion step of maintaining and managing a seedword set including one or more seedwords based on detecting the event; And an event detecting step of collecting content of a social media and extracting an event based on the seed word to determine a keyword for the event. It may include.

In the present invention, the seed expansion step, the content collection step of collecting information on the Internet; A word count step of deriving the number of words included in the collected information; A topic selection step of selecting a topic based on the number of words derived in the word count step; And a seedword adding step of extracting a seedword associated with the topic and adding the seedword to a seedword set. It may include.

In the present invention, the event detection step, the social media content collection step of collecting the content of the social media; An event classification step of extracting and classifying an event from contents of social media collected based on the seed word set; An event grouping step of analyzing and grouping classified events based on a creation time and a location of the contents of the social media; And a keyword determining step of determining a keyword for the event. It may include.

In the present invention, the region-specific keyword analysis step, it is possible to derive a time zone having the maximum amount of content including the keyword for each time zone according to a predetermined region classification.

In the present invention, the trajectory of the event can be derived based on the sequence of time zones in which the amount of the content derived in the regional keyword analysis step has a maximum value.

According to the present invention, the emotion analysis method may include: a first display step of displaying a display element representing an emotion of the content and a trajectory of the event through a display device; It may further include.

In the present invention, the first display step, it is possible to show the emotion of the content through one or more means of the position, size and color of the display element.

In the present invention, the emotion analysis method, the emotion prediction step of predicting the emotion in the specific region based on the emotion data including the results of the time-specific emotion analysis in a specific region; It may further include.

In the present invention, the emotion prediction step, the region selection step of selecting the specific region to predict the emotion; An emotion data loading step of importing emotion data including a result of time-specific emotion analysis in the specific region; And predicting emotion degree of predicting emotion in the specific region based on the emotion data. It may include.

In the present invention, the predictive emotion extraction step, it is possible to predict the emotion by the learned artificial neural network.

In the present invention, the artificial neural network may be a cyclic neural network based on Long and Short Term Memory (LSTM).

According to the present invention, the emotion analysis method may include: a second display step of showing, through a display apparatus, a predicted emotion in the specific region derived in the emotion prediction step; It may further include.

In order to solve the above problems, the present invention is a social media content-based emotion analysis system, the event determination unit for extracting the event from the content of the social media in real time to determine the keyword for the event; A regional keyword analysis unit for performing regional and hourly analysis of contents including the keyword from the contents of the social media; A trajectory analyzer configured to analyze the trajectory of the event from regional and hourly analysis results of the content including the keyword; Emotion analysis unit for analyzing the emotion of the content from the content containing the keyword; And a first display unit for displaying a display element representing an emotion of the content and a trajectory of the event through a display device. It provides a, social media content-based emotion analysis system.

In order to solve the above problems, the present invention provides a computer-readable medium for social media content-based emotional analysis, the computer-readable medium stores instructions for causing a computing device to perform the following steps: The steps may include: an event determining step of extracting an event from contents of social media in real time to determine a keyword for the event; A regional keyword analysis step of performing regional and hourly analysis of contents including the keyword from the contents of the social media; A trajectory analysis step of analyzing the trajectory of the event from regional and hourly analysis results of the content including the keyword; And an emotion analysis step of analyzing the emotion of the content from the content including the keyword. It provides a computer-readable medium comprising a.

According to an embodiment of the present invention, it is possible to extract an event from social media content generated in real time and to analyze an emotion for the event.

According to an exemplary embodiment of the present invention, an emotion trajectory may be analyzed by analyzing a place and a time of social media content generated in real time.

According to an embodiment of the present invention, it is possible to exert an effect of analyzing the emotion of social media content through an artificial neural network.

According to an embodiment of the present invention, the user can grasp the information at a glance by analyzing the trajectory and the emotion of the event together and showing the analysis result through the display device.

According to an embodiment of the present invention, the artificial neural network may exert an effect of predicting the emotion of the selected region.

1 is a view schematically showing the steps of a social media content-based emotion analysis method according to an embodiment of the present invention.
2 is a view schematically showing the detailed steps of the event determination step according to an embodiment of the present invention.
3 is a view schematically showing the detailed steps of the seedword expansion step according to an embodiment of the present invention.
4 is a view schematically showing the detailed steps of the event detection step according to an embodiment of the present invention.
5 is a diagram schematically illustrating an operation of a region-specific keyword analysis step according to an embodiment of the present invention.
6 is a view schematically showing the operation of the trajectory analysis step according to an embodiment of the present invention.
FIG. 7 is a diagram schematically illustrating vectorization of content in an emotion analysis step according to an embodiment of the present invention.
8 is a diagram schematically illustrating the steps of a social media content based emotion analysis method according to an embodiment of the present invention.
9 is a view schematically showing the detailed steps of the emotion prediction step according to an embodiment of the present invention.
10 is a view schematically showing the operation of the emotion prediction step according to an embodiment of the present invention.
11 is a diagram schematically illustrating an internal structure of a social media content based emotion analysis system according to an embodiment of the present invention.
12 is a diagram schematically illustrating an internal structure of an event determining unit according to an embodiment of the present invention.
13 is a view schematically showing the internal structure of the emotion prediction unit according to an embodiment of the present invention.
14 is a diagram illustrating a display screen in the event determination step according to an embodiment of the present invention.
15 is a diagram illustrating a display screen in the first display step according to an embodiment of the present invention.
16 is a diagram illustrating a display screen in a first display step according to an embodiment of the present invention.
17 is a diagram illustrating a display screen in a second display step according to an embodiment of the present invention.
18 is a diagram illustrating a display screen in the second display step according to an embodiment of the present invention.
19 is a block diagram illustrating an example of an internal configuration of a computing device according to an embodiment of the present invention.

In the following, various embodiments and / or aspects are now disclosed with reference to the drawings. In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of one or more aspects. However, it will also be appreciated by one of ordinary skill in the art that this aspect (s) may be practiced without these specific details. The following description and the annexed drawings set forth in detail certain illustrative aspects of the one or more aspects. However, these aspects are exemplary and some of the various methods in the principles of the various aspects may be used and the descriptions described are intended to include all such aspects and their equivalents.

Moreover, various aspects and features will be presented by a system that may include a number of devices, components, and / or modules, and the like. The various systems may include additional devices, components, and / or modules, etc., and / or may not include all of the devices, components, modules, etc. discussed in connection with the drawings. It must be understood and recognized.

As used herein, “an embodiment”, “an example”, “aspect”, “an example”, etc., may not be construed as any aspect or design described is better than or advantageous to other aspects or designs. . The terms '~ part', 'component', 'module', 'system', 'interface', etc. used generally mean a computer-related entity, for example, hardware, hardware And a combination of software and software.

In addition, the terms "comprises" and / or "comprising" mean that such features and / or components are present, but exclude the presence or addition of one or more other features, components, and / or groups thereof. It should be understood that it does not.

In addition, terms including ordinal numbers such as first and second may be used to describe various components, but the components are not limited by the terms. The terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be referred to as the second component, and similarly, the second component may also be referred to as the first component. The term and / or includes a combination of a plurality of related items or any item of a plurality of related items.

In addition, in the embodiments of the present invention, unless otherwise defined, all terms used herein including technical or scientific terms are generally understood by those skilled in the art to which the present invention belongs. Has the same meaning as Terms such as those defined in the commonly used dictionaries should be interpreted as having meanings consistent with the meanings in the context of the related art, and ideally or excessively formal meanings, unless explicitly defined in the embodiments of the present invention. Not interpreted as

1 is a view schematically showing the steps of a social media content-based emotion analysis method according to an embodiment of the present invention.

Social media content has a feature that spreads very quickly because users can freely create and share it. Users create a lot of opinions or new information about the obtained information and share it with other users. Through this process, social media contents are sensitive to ever-changing social issues and natural disasters, and contents containing various information and emotions are rapidly spreading. The present invention obtains information and emotions from such social media content, detects events such as social issues and natural disasters, and analyzes and shows the trajectories and emotions of social media content for the event to allow the user to see the event at a glance. Make sure you have information about

Referring to Figure 1, the social media content-based emotion analysis method according to an embodiment of the present invention event determination step (S100); Regional keyword analysis step (S200); A trajectory analysis step (S300); Emotion analysis step (S400); And first display step (S500); It includes.

Social media content-based emotion analysis method according to an embodiment of the present invention first performs an event determination step (S100) for extracting the event from the content of the social media in real time to determine the keyword for the event. Since the social media content is generated in real time, the event determination step (S100) determines the event to be analyzed by detecting events such as social issues or natural disasters appearing on the social media by acquiring and analyzing the content of the social media. The keyword for the event may be determined to analyze social media content including the keyword.

Thereafter, a regional keyword analysis step (S200) of performing regional and hourly analysis of contents including the keyword from the contents of the social media is performed. The regional keyword analysis step (S200) analyzes the social media content including the keyword derived in the event determination step (S100) by region and time, thereby generating and propagating the content of the event in the social media. Allow for analysis.

In an embodiment of the present invention, the region-specific keyword analysis step (S200) derives a time zone in which the amount of content including the keyword for each time zone has a maximum value according to a predetermined region classification. In this way, the content including the keyword is classified according to the region classification, the amount of the content is classified according to the time zone, and a time zone having the maximum value of the content is derived, and the content including the keyword becomes an issue. Identify and analyze local areas and time zones. The time zone is a time zone divided according to a preset time unit. For example, the predetermined time unit is 1 minute, and the content is classified in units of 1 minute according to a generation time, and the amount of content is measured for each classified time zone. At this time, a time zone in which the measured amount of content has a maximum value, that is, a specific time and minute may be derived.

Thereafter, a trajectory analysis step (S300) of analyzing the trajectory of the event from the regional and hourly analysis results of the content including the keyword is performed. In an embodiment of the present invention, the movement trajectory of the interest for the event may be derived and analyzed based on the regional distribution and the hourly distribution of the content including the keyword for the event derived through the regional and hourly analysis results of the content. Can be.

In one embodiment of the present invention, the trajectory analysis step S300 derives the trajectory of the event based on the order of time zones in which the amount of the content derived in the regional keyword analysis step S200 has a maximum value. In the region-specific keyword analysis step (S200), the content including the keyword for each region is classified according to time zones, whereby the region and time zone in which the content including the keyword is an issue can be derived, and the trajectory analysis step (S300). ), The trajectory of the event of the keyword can be derived and analyzed by connecting the region where the content including the keyword is an issue for each time zone.

Thereafter, an emotion analysis step S400 of analyzing an emotion of the content from the content including the keyword is performed. By analyzing the content, the emotion of the user who created the content may be derived, and the emotion of the social media users may be analyzed by analyzing the derived emotion.

At this time, the emotion analysis step (S400) in one embodiment of the present invention may analyze the emotion by the learned artificial neural network. By analyzing the emotions through the learned artificial neural network, the emotions of the content creator can be derived from social media content consisting of text or emoticons.

Thereafter, a first display step S500 of displaying a display element indicating an emotion of the content and a trajectory of the event through a display device is performed. In the first display step (S500), the trajectory of the event analyzed in the trajectory analysis step (S300) and the emotions of the users analyzed in the emotion analysis step (S400) are shown on the display at a glance. Make sure that you have information about The emotion analyzed at this time may be shown as a display element on the display device. At this time, the first display step, it is possible to show the emotion of the content through one or more means of the position, size and color of the display element. For example, in showing the emotion of the content in a specific region, the region of generation of the content is indicated by the position of the display element, the amount of the content is indicated by the size of the display element, and the color is expressed by the color of the display element. The emotional state of the content may be displayed.

2 is a view schematically showing the detailed steps of the event determination step according to an embodiment of the present invention.

Referring to FIG. 2, the event determination step S100 according to an embodiment of the present invention includes a seedword expansion step S110 and an event detection step S120.

In social media sites, users share information in their daily lives, so various events are spread in real time. Content added to the timeline in real time has a characteristic that occurs particularly intensively at a time and place where a specific event occurs, and can detect an event using such a characteristic.

In the event determination step (S100), first, a seedword expansion step (S110) of maintaining and managing a seedword set including one or more seedwords as a basis for detecting the event is performed. In the present invention, the seedword is used to detect an event from a social media site. The seed word may be one or more, and the event generated in real time may be detected by analyzing the content of the social media based on the seed word. The seedword expansion step S110 may detect an event by managing a seedword set including one or more seedwords.

This seedword set must be constantly updated to detect events occurring in real time. To this end, in the seedword expansion step S110, an event may be accurately detected from the seedword set by adding a new issue or event classification category to the seedword set as a seedword, or deleting an old seedword.

Thereafter, content of social media is collected, and an event detection step (S120) of extracting an event based on the seed word to determine a keyword for the event is performed. The event detection step (S120) detects and extracts an event from a social media content generated in real time using a seed word included in the seed set, determines a keyword for the event, and associates the event with a social media site. Only content can be extracted.

In the event detecting step (S120) according to an embodiment of the present invention, social media content including the seed word may be classified, and the event may be detected and extracted by classifying the content according to a generation time and a place.

3 is a view schematically showing the detailed steps of the seedword expansion step according to an embodiment of the present invention.

Referring to Figure 3, the seed word expansion step (S110) according to an embodiment of the present invention, the content collection step (S111); A word count step (S112); Topic selection step (S113); And seed addition step 114; It includes.

In the seed word expansion step S110, first, a content collection step S111 of collecting information on the Internet is performed. In the content collection step (S111), by collecting information on the Internet as well as the content of the social media, it is possible to derive information on recent social issues or natural disasters to manage seedwords. Preferably, the content collection step S111 may collect news from one or more news information sites. The news information site may be a site of a press company that provides news directly, or may be a news category of a portal site that aggregates and provides news on the Internet, or a news providing database. As described above, in order to detect an event in social media, it is possible to more accurately detect an event from the social media by identifying a recent issue from information such as news and reflecting it in the seedword set.

Thereafter, a word count step S112 of deriving the number of words included in the collected information is performed. In the word count step S112, the number of words included in the information collected in the content collection step S111 may be derived to extract an issue mainly dealt with in the collected information. In an embodiment of the present invention, in the content collection step (S111), the news is collected from a news information site, and in the word count step (S112), the frequency of use of the words included in the collected news is derived from the latest news. The issue can be derived.

Thereafter, a topic selection step S113 of selecting a topic based on the number of words derived in the word count step S112 is performed. In one embodiment of the present invention, in the topic selection step (S113), the word of interest is extracted from the collected news information, and a recent issue of interest is selected to be selected as a topic. At this time, according to an embodiment of the present invention, in the topic selection step S113, a predetermined word may be excluded from the topic. When a topic is selected from words such as news, which are frequently used, words that cannot be derived from recent issues of interest, such as sentences repeatedly used in standardized texts such as words and articles that cannot convey meaning, such as research, etc. Topics can be selected, resulting in lower accuracy. Therefore, such words should be excluded from the topic to select recent issues of interest as topics.

Thereafter, a seedword adding step of extracting a seedword associated with the topic and adding to the seedword set (S114); It includes. In the seedword adding step S114, a seedword associated with the topic selected in the topic selection step S113 is extracted and added to a seedword set for detecting an event from social media content to increase the detection accuracy of the event. have. The seedword associated with the topic may be the name of the topic or a classification name that classifies the topic. The extracted seedwords are added to the seedword set. At this time, if the seedword is continuously added to the seedset, the size of the seedword set may be too large to be detrimental to management, and as the seedword is added, an existing seedword may be deleted from the seedword set. Can be. At this time, the seedword added to the seedword set first may be deleted, or the score is evaluated by an evaluation score of the seedword (score evaluated according to the degree of detecting an event in the event detection step S120). The seedword with the lowest evaluation score may be deleted.

4 is a view schematically showing the detailed steps of the event detection step according to an embodiment of the present invention.

Referring to Figure 4, the event detection step (S120), the social media content collection step (S121); Event classification step (S122); Event grouping step (S123); And keyword determination step (S124); It includes.

In the event detecting step (S120), first, the social media content collecting step (S121) of collecting the content of the social media is performed. Since the content is generated and registered in real time on the social media, the social media content collection step (S121) also collects the content registered in real time to detect an event from the content.

Thereafter, an event classification step S122 is performed to extract and classify an event from the contents of the collected social media based on the seedword set. The seedword set includes a seedword for extracting an event from the content of the social media. In the event classification step (S122), the event is extracted from the contents of the social media based on the seed word. The seedword set includes seedwords related to various events and the like, and in the event classification step (S122), the seed including the seedword is extracted from the contents of the social media, and the seed is analyzed by analyzing the extracted content. It determines whether an event related to a word is currently occurring and extracts and classifies an event that is currently occurring.

Thereafter, an event grouping step (S123) is performed to analyze and group the classified events based on the creation time and the location of the contents of the social media. Although the content generated in the social media is generated at various times and various places, the content related to a specific event is generated more at the time and place associated with the time and place where the event is occurring, so that the time for creating the social media content and By classifying and grouping according to the place of creation, the time and place of the event can be derived to identify the event.

Thereafter, a keyword determination step S124 of determining a keyword for the event is performed. The keyword determination step (S124) derives a keyword that can represent the specified event through the event grouping step (S123). The keyword may be one word or a combination of two or more words. As such, by determining a keyword for an event that currently occurs, it is possible to extract and analyze the content of the event from social media content through the keyword.

5 is a diagram schematically illustrating an operation of a region-specific keyword analysis step according to an embodiment of the present invention.

In an embodiment of the present invention, the region-specific keyword analysis step (S200) derives a time zone in which the amount of content including the keyword for each time zone has a maximum value according to a predetermined region classification. FIG. 5A illustrates the locations of three regions A1, A2, and A3 in which content including the keyword is generated. In FIG. 5B, the regions in the regions A1, A2, and A3 are shown. A graph showing the change over time of the amount of content including a keyword is shown.

The keyword is a keyword that can indicate a specific event, and the amount of content including the keyword in social media changes according to the occurrence and progress of the event.

Referring to FIG. 5B, the region A1 has a maximum value of the amount of content including the keyword for the time at time t ₁ . Further, in the region A2, the local maximum value of the content including the keyword for the time at time t _{2, and} in the region A3, the maximum of the quantity of the content including the keyword for the time at the time t ₃ is obtained. Have

In the regional keyword analysis step (S200), as shown in FIG. 5, the content including the keyword is classified according to a predetermined region classification, and the amount of content including the keyword for each time zone according to the region classification is a maximum value. To derive the time zone. In the embodiment of FIG. 5, the region-specific keyword analysis step (S200) is a time t ₁ for the A1 region, a time t ₂ for the A2 region, and a time t ₃ for the A3 region to derive a trajectory analysis step (S300). Allows you to analyze the trajectory in.

6 is a view schematically showing the operation of the trajectory analysis step according to an embodiment of the present invention.

In the exemplary embodiment of the present invention, the trajectory analysis step S300 derives a trajectory of the event based on the order of time zones in which the amount of the content derived in the regional keyword analysis step has a maximum value. FIG. 6A illustrates the locations of three regions A1, A2, and A3 in which content including the keyword is generated, as shown in FIG. 5A, and FIG. Times t ₁ , t ₂ , and t ₃ for the A1, A2, and A3 derived through the keyword analysis step S200 are shown in chronological order.

Referring to FIG. 6B, the times t ₁ , t ₂ , and t ₃ derived in the region-specific keyword analysis step S200 have a time sequence of t ₃ , t ₁ , and t ₂ . Therefore, in the trajectory analysis step (S300), a trajectory of an area in which a lot of content including the keyword is generated may be generated by connecting the region A3, the region A1, and the region A2 corresponding to the time in the order of time. . Since the content including the keyword tends to be generated in the place where the event for the keyword occurs or the event is in progress, the trajectory of the event is analyzed by analyzing the trajectory of the region where the content including the keyword is generated. It can exert an effect that can be done.

FIG. 7 is a diagram schematically illustrating vectorization of content in an emotion analysis step according to an embodiment of the present invention.

In an embodiment of the present invention, the emotion analysis step S400 may analyze the emotion by the learned artificial neural network. In more detail, the artificial neural network is a convolutional neural network (CNN), and the content including the keyword may be vectorized and input into the convolutional neural network.

Convolutional neural networks are artificial neural networks that perform very well in image classification problems. In order to analyze the emotions from the contents of the social media, the text constituting the contents of the social media may be converted into a two-dimensional vector and input to the convolutional neural network as shown in FIG. . In one embodiment of the present invention, the content may be vectorized using the word2vec algorithm. The two-dimensionalized vectorized content may be input to the convolutional neural network to learn to classify the content.

The emotion analysis step S400 derives emotions from the input content and classifies them into preset categories. In one embodiment of the present invention, the category may include two categories, a positive category and a negative category, to simply classify the content into positive and negative emotions, or the category may further include three or more categories. You can also analyze your emotions.

8 is a diagram schematically illustrating the steps of a social media content based emotion analysis method according to an embodiment of the present invention.

8, the social media content-based emotion analysis method according to an embodiment of the present invention comprises the event determination step (S100); Regional keyword analysis step (S200); A trajectory analysis step (S300); Emotion analysis step (S400); Emotion prediction step (S600); And second display step (S700); It includes.

The event determination step (S100), the region-specific keyword analysis step (S200), the trajectory analysis step (S300) and the emotion analysis step (S400) are the same as described in Figures 1 to 7, so detailed description thereof will be omitted. do.

Social media content-based emotion analysis method according to an embodiment of the present invention first performs an event determination step (S100) to determine the keyword for the event by extracting the event in real time from the content of the social media, and then, Perform a regional keyword analysis step (S200) for performing regional and hourly analysis of contents including the keyword from contents of social media, and thereafter, track the event from regional and hourly analysis results of the contents including the keyword. After performing the trajectory analysis step (S300) for analyzing, and then, the emotion analysis step (S400) for analyzing the emotion of the content from the content containing the keyword.

Thereafter, an emotion prediction step S600 is performed to predict the emotion in the specific region based on the emotion data including the result of the emotional analysis of the time at the specific region. In the emotion prediction step (S600), subsequent emotions are predicted based on the results of past emotional analysis in a specific region. To this end, the emotion prediction step S600 may extract emotion data including a result of time-specific emotion analysis in a specific region derived in the emotion analysis step S400, and derive a predicted emotion from the emotion data. At this time, the emotion prediction step (S600) in one embodiment of the present invention can predict the emotion by the learned artificial neural network.

Thereafter, a second display step S700 of displaying the predicted emotion in the specific region derived in the emotion prediction step S600 through the display apparatus is performed.

In the second display step S700, the predicted emotion in the specific region derived is illustrated through at least one element of a position, a number, a graph, and a color. For example, in the second display step S700, the derived emotion predicted in the specific region is displayed in a color matched with the emotion predicted on the position of the specific region, and the emotion predicted by the user's selection is displayed. You can also display a numerical value.

9 is a view schematically showing the detailed steps of the emotion prediction step according to an embodiment of the present invention.

Referring to FIG. 9, in one embodiment of the present invention, the emotion prediction step S600 may include a region selection step S610; Emotion data loading step (S620); And predictive emotion extraction step (S630); It may include.

In the emotion prediction step (S600), first, a region selection step (S610) of selecting the specific region to predict the emotion is performed. The region to predict emotion may be determined by the user's selection, or the region may be selected by predicting the region where the trajectory will follow from the previous trajectory.

Thereafter, an emotion data loading step (S620) of importing emotion data including an analysis result of an hourly emotion in the specific region is performed. In order to predict emotions in the specific area, information on past emotions is required. To this end, it is possible to prepare to predict the emotion by loading the emotion data classified by the emotion of the specific region analyzed by the emotion analysis step (S400) by time. In this case, the emotion data may include a result of emotion analysis from a time of loading the emotion data up to a predetermined time interval. For example, when the predetermined time interval is one week, in order to predict the later emotions, data for analyzing emotion analysis data from one week ago to the present so far are selected and arranged, and such Based on the data, it is possible to predict future emotions.

Thereafter, a prediction emotion extraction step S630 of predicting an emotion in the specific region is performed based on the emotion data.

In an embodiment of the present invention, the predicted emotion extraction step (S630) may predict emotion using a moving average. A method of predicting emotion using a moving average may be expressed as in the following equation.

In the above formula, N is the size of the emotion data (time interval), sentiment is an index indicating emotion, and cont_weight is a weight. The weight may have a value between 0 and 1 in consideration of continuity of emotion.

Alternatively, in another embodiment of the present invention, the predicted emotion extraction step (S630) may predict the emotion by the learned artificial neural network. In this case, the artificial neural network may be a recurrent neural network (RNN) based on long and short term memory (LSTM). The LSTM-based cyclic neural network has ordered information as input data. Accordingly, the LSTM-based cyclic neural network used in the predictive emotion extraction step (S630) receives emotion data including the analyzed emotion information at the previous time to predict and derive the subsequent emotions. In addition, the LSTM-based circulatory neural network has no long-term dependency problem as compared to the general circulatory neural network, which is advantageous for predicting emotion through long time emotional data.

Such emotional prediction can be used in various fields such as disaster prediction systems.

10 is a view schematically showing the operation of the emotion prediction step according to an embodiment of the present invention.

10 illustrates changes over time of emotion data in a specific region. In the graph of FIG. 10, the horizontal axis represents time. In the embodiment of Figure 10, the emotion analysis step (S400) analyzes the emotion of the social media content in two categories, positive emotion and negative emotion. In the graph of FIG. 10, the area shaded in gray is a time zone in which negative emotions prevail as a result of emotion analysis, and the area shaded with hatched areas is time zone in which relatively positive emotion is predominant as a result of emotion analysis.

FIG. 10A illustrates a process of performing the emotion prediction step at time t ₁ . As shown in (a) of FIG. 10, the result of analyzing the emotion in the specific region is changed to a positive emotion after continuing a negative emotion, and then, in order to predict the emotion at time t ₁ after the change to a negative emotion, the time t The emotion analysis result from ₁ to a predetermined time interval is set as the emotion data to predict the emotion based on the emotion data. Figure 10 (a) to the result of the analysis of the emotion after t ₁ to the emotion is changed in a positive feeling to the feeling data is inputted to the emotion prediction unit that performs emotion prediction step (S600). The emotion prediction unit predicts an emotion to follow after t ₁ based on the received emotion data. In (a) of FIG. 10, the emotion prediction unit predicted that negative emotions will follow.

10B illustrates a process of performing the emotion prediction step at time t ₂ after time t ₁ . In FIG. 10 (b), after the time t ₁ , the negative emotion continues and then turns into a positive emotion, and then, in order to predict the emotion at the time t ₂ after the change to the negative emotion, the time interval before the predetermined time interval from the time t ₂ is changed. By setting the result of emotion analysis up to emotion data to predict the emotion based on the emotion data. In (b) of FIG. 10, the emotion analysis result of the emotion change from negative to positive and then negatively changed to t ₂ is input to the emotion prediction unit that performs the emotion prediction step (S600). The emotion prediction unit predicts an emotion to follow after t ₂ based on the received emotion data. In FIG. 10B, the emotion prediction unit predicted to be changed to a positive emotion.

FIG. 10C illustrates a process of performing the emotion prediction step at time t ₃ after time t ₁ and time t ₂ . (C) in FIG. 10, the emotional analysis of the time followed by t ₂ positive emotions in order to predict the short duration doedaga feeling at the time t ₃ after a sustained change in a negative emotion to prior to the time period from t ₃ a predetermined distance The result is set as the emotion data to predict the emotion based on the emotion data. In (c) of FIG. 10, a positive emotion is negatively changed, and the result of emotion analysis up to t ₃ is input to the emotion prediction unit that performs the emotion prediction step (S600) as the emotion data. The emotion prediction unit predicts an emotion to follow after t ₃ based on the received emotion data. In FIG. 10C, the emotion prediction part predicted that negative emotions will continue.

As shown in FIG. 10, in the emotion prediction step S600, emotion data classified by the emotions of the specific region analyzed in the emotion analysis step S400 by time is retrieved, and subsequently, based on the emotion data. Predict emotions.

11 is a diagram schematically illustrating an internal structure of a social media content based emotion analysis system according to an embodiment of the present invention.

Referring to FIG. 11, the social media content-based emotional analysis system according to an embodiment of the present invention includes an event determination unit 100 which extracts an event from a content of social media in real time and determines a keyword for the event; A regional keyword analysis unit 200 for performing regional and hourly analysis of contents including the keyword from the contents of the social media; A trajectory analysis unit 300 which analyzes a trajectory of the event from regional and hourly analysis results of the content including the keyword; Emotion analysis unit 400 for analyzing the emotion of the content from the content containing the keyword; A first display unit 500 showing a display element representing an emotion of the content and a trajectory of the event through a display device; An emotion prediction unit 600 predicting an emotion in the specific region based on the emotion data including an analysis result of time-specific emotion in a specific region; And a second display unit 700 showing a predicted emotion in the specific region derived from the emotion prediction unit 600 through a display device. It may include.

The event determination unit 100, the regional keyword analysis unit 200, the trajectory analysis unit 300, the emotion analysis unit 400, the first display unit 500, the emotion prediction unit 600 and Each of the second display units 700 includes an event determination step (S100), a keyword analysis step (S200), a trajectory analysis step (S300), an emotion analysis step (S400), and a first display step described in FIGS. In operation S500, the emotion prediction step S600 and the second display step S700 are performed.

The event determiner 100 includes: a seedword extension unit 110 for maintaining and managing a seedword set including one or more seedwords based on detecting the event; And an event detecting unit 120 collecting content of the social media and extracting an event based on the seed word to determine keywords for the event. It may include.

The seedword extension unit 110 and the event detection unit 120 respectively perform the seedword extension step S110 and the event detection step S120 described with reference to FIG. 2.

12 is a diagram schematically illustrating an internal structure of an event determining unit according to an embodiment of the present invention.

Referring to Figure 12, the seed expansion unit 110 according to an embodiment of the present invention includes a content collection module 111 for collecting information on the Internet; A word count module 112 for deriving the number of words included in the collected information; A topic selection module 113 for selecting a topic based on the number of words derived from the word count module 112; And a seedword adding module 114 which extracts a seedword associated with the topic and adds the seedword to the seedword set. It may include.

The content collection module 111, the word count module 112, the topic selection module 113, and the seedword adding module 114 are the content collection step S111 and the word count step S112 described with reference to FIG. 3, respectively. ), The topic selection step S113 and the seedword addition step S114 are performed.

The seed word derived by the operation of the seed word expansion unit 110 is added to and stored in the seed word set stored in the seed word expansion unit 110 or in a separate storage device.

The event detection unit 120 includes a social media content collection module 121 for collecting the content of the social media; An event classification module 122 for extracting and classifying an event from contents of social media collected based on the seed word set; An event grouping module 123 for analyzing and grouping classified events based on a creation time and a location of the contents of the social media; And a keyword determination module 124 for determining a keyword for the event. It may include.

The social media content collection module 121, the event classification module 122, the event grouping module 123 and the keyword determination module 124 are the social media content collection step (S121), the event described in FIG. A classification step S122, the event grouping step S123, and the keyword determination step S124 are performed.

The event classification module 122 extracts an event from the contents of the social media collected based on the seedword set by loading a seedword set stored in the seedword extension unit 110 or in a separate storage device. can do.

13 is a view schematically showing the internal structure of the emotion prediction unit according to an embodiment of the present invention.

Referring to FIG. 13, the emotion prediction unit 600 according to an embodiment of the present invention includes a region selection module 610 for selecting the specific region for predicting emotion; An emotion data load module 620 for importing emotion data including a result of analyzing emotions by time in the specific region; And a prediction emotion extraction module 630 for predicting the emotion in the specific region based on the emotion data. It may include.

The region selection module 610, the emotion data loading module 620, and the prediction emotion extraction module 630 are the region selection step S610, the emotion data loading step S620, and the prediction described with reference to FIGS. 9 and 10, respectively. The emotion derivation step S630 is performed.

14 is a diagram illustrating a display screen in the event determination step according to an embodiment of the present invention.

FIG. 14 illustrates a screen on which a result of deriving an event in an event determination step S100 according to an embodiment of the present invention is shown through a display apparatus. In the embodiment of FIG. 14, the tweeter including the seedword set is analyzed based on the tweeter data collected in real time. The social media content collecting step (S121) collects the data of the Twitter, and in the event classification step (S122) extracts the event based on the seedword added in the seedword expansion step (S110). In FIG. 14, an event extracted from a seedword “accident” is illustrated.

The extracted events may be grouped through an event grouping step S123, and keywords for the grouped events may be extracted through the keyword determining step S124.

15 and 16 are views illustrating a display screen in the first display step according to an embodiment of the present invention.

15 and 16 show the analysis results of trajectories and emotions for eyes and tornadoes, respectively. 15 and 16 are results of analyzing emotions according to an embodiment of the present invention using Twitter data from April 1 to 30, 2016. In the embodiment of Figs. 15 and 16, the locus is shown using a flag indicating the order of the locus. In addition, the amount of content including keywords in each region is analyzed and displayed in the size of the display element. In one embodiment of the present invention, the display element may be animated on a chronological order and shown on a map. As can be seen in Figures 15 and 16, the results of the trajectory analyzed according to an embodiment of the present invention reflects the movement path of the actual eye and tornado.

In addition, the user's emotional analysis results are expressed in blue and red using a heat map. In the embodiments of FIGS. 15 and 16, the emotion analysis results are classified into two categories, positive and negative. In the illustrated screen, red represents negative emotion and blue represents positive emotion. In addition, the darker the color, the greater the proportion of emotion corresponding to the color.

In the case of the eye of FIG. 15, the negative emotions are relatively high in the severe eye region, but the positive emotion is high in 50.3% and the negative emotion rate is 49.7% in other regions. On the other hand, in the case of tornado, the negative emotion was 81.8%, which showed a strong negative emotion.

According to an exemplary embodiment of the present invention, since the emotion analysis result and the trajectory analysis result for the event generated as shown in FIGS. 15 and 16 are displayed at the same time, the analysis result can be visually identified much more easily.

17 and 18 are views illustrating a display screen in a second display step according to an embodiment of the present invention.

17 and 18 show the results of the prediction of the emotions based on the analysis results of the emotions on the eyes and tornadoes as shown in FIGS. 15 and 15, respectively. FIG. 17 predicts an emotion of a selected region by predicting a region to which the trajectory will follow through analysis of the trajectory, and shows the predicted emotion of the region.

In the embodiment of Fig. 17, the locus predicts the locus following the locus in the order of 1 to 7, followed by Los Angeles in the west, and shows the predicted emotion of Los Angeles in color and number. Predicted results show that positive emotions are high, with 35% positive and 10% negative. Accordingly, in the second display step, the predicted emotion is illustrated through blue and specific numbers representing positive emotions.

In the embodiment of FIG. 18, the predicted emotion of the area selected by the user's input is shown by color and number. Predicted results show that 5% of positive emotions and 65% of negative emotions are very high. Accordingly, in the second display step, the predicted emotion is illustrated through red and specific numbers representing negative emotion. At this time, the color is displayed very strongly, so that the result predicted that the negative emotion is very strong can be visually recognized more clearly.

19 is a block diagram illustrating an example of an internal configuration of a computing device according to an embodiment of the present invention.

As shown in FIG. 19, the computing device 11000 may include at least one processor 11100, a memory 11200, a peripheral interface 11300, and an input / output subsystem ( At least one I / O subsystem 11400, a power circuit 11500, and a communication circuit 11600 may be included. In this case, the computing device 11000 may correspond to an emotion analysis system.

The memory 11200 may include, for example, high-speed random access memory, magnetic disk, SRAM, DRAM, ROM, flash memory, or nonvolatile memory. have. The memory 11200 may include a software module, an instruction set, or other various data necessary for the operation of the computing device 11000.

In this case, accessing the memory 11200 from another component such as the processor 11100 or the peripheral device interface 11300 may be controlled by the processor 11100.

The peripheral interface 11300 may couple the input and / or output peripherals of the computing device 11000 to the processor 11100 and the memory 11200. The processor 11100 may execute a software module or an instruction set stored in the memory 11200 to perform various functions for the computing device 11000 and process data.

Input / output subsystem 11400 may couple various input / output peripherals to peripheral interface 11300. For example, the input / output subsystem 11400 may include a controller for coupling a peripheral device such as a monitor or keyboard, a mouse, a printer, or a touch screen or a sensor, as necessary, to the peripheral interface 11300. According to another aspect, the input / output peripherals may be coupled to the peripheral interface 11300 without passing through the input / output subsystem 11400.

The power circuit 11500 may supply power to all or part of the components of the terminal. For example, power circuit 11500 may include a power management system, one or more power sources such as batteries or alternating current (AC), charging systems, power failure detection circuits, power converters or inverters, power status indicators or power sources. It can include any other components for creation, management, distribution.

The communication circuit 11600 may enable communication with another computing device using at least one external port.

Alternatively, as described above, the communication circuit 11600 may include an RF circuit to transmit and receive an RF signal, also known as an electromagnetic signal, to enable communication with other computing devices.

This embodiment of FIG. 19 is merely an example of the computing device 11000, and the computing device 11000 may include some components shown in FIG. 19, or may further include additional components not shown in FIG. 19, or 2. It may have a configuration or arrangement that combines two or more components. For example, the computing device for a communication terminal in a mobile environment may further include a touch screen or a sensor, in addition to the components illustrated in FIG. 19, and various communication schemes (WiFi, 3G, LTE) in the communication circuit 1160. , Bluetooth, NFC, Zigbee, etc.) may include a circuit for RF communication. Components that may be included in the computing device 11000 may be implemented in hardware, software, or a combination of both hardware and software, including integrated circuits specialized for one or more signal processing or applications.

Methods according to an embodiment of the present invention may be implemented in the form of program instructions that may be executed by various computing devices, and may be recorded on a computer readable medium. In particular, the program according to the present embodiment may be configured as a PC-based program or an application dedicated to a mobile terminal. The application to which the present invention is applied may be installed in the user terminal through a file provided by the file distribution system. For example, the file distribution system may include a file transmitter (not shown) for transmitting the file at the request of the user terminal.

According to an embodiment of the present invention, it is possible to extract an event from social media content generated in real time and to analyze an emotion for the event.

According to an exemplary embodiment of the present invention, an emotion trajectory may be analyzed by analyzing a place and a time of social media content generated in real time.

According to an embodiment of the present invention, it is possible to exert an effect of analyzing the emotion of social media content through an artificial neural network.

According to an embodiment of the present invention, the user can grasp the information at a glance by analyzing the trajectory and the emotion of the event together and showing the analysis result through the display device.

According to an embodiment of the present invention, the artificial neural network may exert an effect of predicting the emotion of the selected region.

Although the embodiments have been described by the limited embodiments and the drawings as described above, various modifications and variations are possible to those skilled in the art from the above description. For example, the described techniques may be performed in a different order than the described method, and / or components of the described systems, structures, devices, circuits, etc. may be combined or combined in a different manner than the described method, or other components. Or even if replaced or replaced by equivalents, an appropriate result can be achieved. Therefore, other implementations, other embodiments, and equivalents to the claims are within the scope of the claims that follow.

Claims (12)

As a social media content based emotion analysis method,
An event determination step of extracting an event from a content of social media in real time and determining a keyword for the event;
A regional keyword analysis step of performing regional and hourly analysis of contents including the keyword from the contents of the social media;
A trajectory analysis step of analyzing the trajectory of the event from regional and hourly analysis results of the content including the keyword; And
An emotion analysis step of analyzing an emotion of the content from the content including the keyword; Social media content-based emotion analysis method comprising a.
The method according to claim 1,
The emotion analysis step,
Emotion analysis method based on social media content, which analyzes emotions by a learned artificial neural network.
The method according to claim 2,
The artificial neural network is a convolutional neural network,
The content including the keyword is vectorized and input to the convolutional neural network, social media content-based emotion analysis method.
The method according to claim 2,
The event determination step,
A seedword expansion step of maintaining and managing a seedword set including one or more seedwords underlying the event detection; And
An event sensing step of collecting contents of social media and extracting an event based on the seed word to determine a keyword for the event; Social media content-based emotion analysis method comprising a.
The method according to claim 4,
The seed word expansion step,
A content collection step of collecting information on the Internet;
A word count step of deriving the number of words included in the collected information;
A topic selection step of selecting a topic based on the number of words derived in the word count step; And
A seedword adding step of extracting a seedword associated with the topic and adding the seedword to a seedword set; Including,
The event detection step,
Collecting social media contents to collect contents of social media;
An event classification step of extracting and classifying an event from contents of social media collected based on the seed word set;
An event grouping step of analyzing and grouping classified events based on a creation time and a location of the contents of the social media; And
A keyword determining step of determining a keyword for the event; Social media content-based emotion analysis method comprising a.
The method according to claim 2,
The regional keyword analysis step,
Deriving a time zone having a maximum amount of content including the keyword for each time zone according to a preset region classification,
The trajectory analysis step,
Social media content-based emotion analysis method of deriving the trajectory of the event based on the sequence of time zones in which the amount of content derived in the regional keyword analysis step has a maximum value.
The method according to claim 2,
The emotion analysis method,
A first display step of displaying a display element representing an emotion of the content and a trajectory of the event; More,
The first display step,
Social media content-based emotion analysis method showing the emotion of the content through one or more means of the position, size and color of the display element.
The method according to claim 2,
The emotion analysis method,
An emotion prediction step of predicting an emotion in the specific region based on the emotion data including an analysis result of time-specific emotion in a specific region; Further comprising, social media content-based emotion analysis method.
The method according to claim 8,
The emotion prediction step,
A region selection step of selecting the specific region for predicting emotion;
An emotion data loading step of importing emotion data including a result of time-specific emotion analysis in the specific region; And
Predicting emotion degree predicting emotion in the specific region based on the emotion data; Including,
The predictive sensitivity extraction step,
Emotion analysis method based on social media content for predicting emotion by the learned artificial neural network.
The method according to claim 9,
The artificial neural network is a cyclic neural network based on long and short term memory (LSTM), social media content-based emotion analysis method.
As a social media content based emotion analysis system,
An event determination unit extracting an event from a content of social media in real time and determining a keyword for the event;
A regional keyword analysis unit for performing regional and hourly analysis of contents including the keyword from the contents of the social media;
A trajectory analyzer configured to analyze the trajectory of the event from regional and hourly analysis results of the content including the keyword;
Emotion analysis unit for analyzing the emotion of the content from the content containing the keyword; And
A first display unit for displaying a display element representing an emotion of the content and a trajectory of the event through a display device; Social media content-based emotional analysis system comprising a.
A computer-readable medium for social media content based emotion analysis,
The computer-readable medium stores instructions for causing a computing device to perform the following steps, the steps:
An event determination step of extracting an event from a content of social media in real time and determining a keyword for the event;
A regional keyword analysis step of performing regional and hourly analysis of contents including the keyword from the contents of the social media;
A trajectory analysis step of analyzing the trajectory of the event from regional and hourly analysis results of the content including the keyword; And
An emotion analysis step of analyzing an emotion of the content from the content including the keyword; A computer-readable medium comprising a.

Images