WO2015005662A1

WO2015005662A1 - Method for analyzing emotional index of text and computer-readable recording medium

Info

Publication number: WO2015005662A1
Application number: PCT/KR2014/006131
Authority: WO
Inventors: Young Hwan Woo
Original assignee: Mezzomedia Co., Ltd.
Priority date: 2013-07-12
Filing date: 2014-07-09
Publication date: 2015-01-15
Also published as: KR101330158B1

Abstract

The present invention relates to a method for analyzing an emotional index of a text and, more particularly, to a method for analyzing an emotional index of a text, the method including the steps of analyzing morphemes of the text, matching the analyzed morphemes with morpheme information stored in a database, and analyzing the emotional index of the text using the results matched with the morpheme information, in which the morpheme information is information representing the degree of association between the morpheme and each item with respect to the properties of the morpheme divided into a plurality of items, and the step of analyzing the emotional index analyzes the degree of association between the plurality of items and the text depending on the morpheme information.

Description

METHOD FOR ANALYZING EMOTIONAL INDEX OF TEXT AND COMPUTER-READABLE RECORDING MEDIUM

The present invention relates to a method for analyzing an emotional index of a text and a computer-readable recording medium storing a program for executing the method and, more particularly, to a method for analyzing a writer’s subjective emotion reflected in a text by analyzing the text.

In general, a piece of writing is done when words are put together to form a sentence and sentences are put together, and the word that is a basic unit of the sentence has a problem with the relationship between one word and another and a problem with the selection of words in the content for writing. The process of solving these two problems can be generally seen as a phenomenon that occurs in the process of writing, and when the second problem, that is, the problem with the selection of words is intensively analyzed, the writer’s emotion can be derived from the analysis.

Each word is transformed in a sentence and implies another meaning in certain situations, but it can be seen that there is an inherent tendency of the word itself. Therefore, the writer may frequently use a word with a certain inherent tendency, and based on literal, cognitive linguistic, and psychoanalytic approaches, it is possible to establish a standard by which the words can be classified through the analysis of the words in accordance with the universal human tendency, thus determining the emotion of the writer based on the classification of the words based on the standard.

Recently, with the rapid increase in smartphone users, social network services such as Twitter and Facebook are activated, and thus there are increasing cases where users express their emotions in simple sentences online or evaluate various products in short sentences online, such as product reviews, movie reviews, restaurant reviews, etc. These sentences have significant effects on other people’s emotions or purchasing tendencies, and thus the analysis and classification of emotions inherent in online sentences become more important than ever in terms of opinion mining or marketing.

Conventionally, a method of extracting positive, negative, and neutral emotions from the sentence has been used. However, the emotions tend to be expressed in various forms and are not simply expressed dichotomously, and thus the conventional method that distinguishes only positive and negative emotions cannot accurately analyze the emotions inherent in the text, which is problematic.

The present invention has been made in an effort to solve the above-described problems associated with prior art, and an object of the present invention is to accurate analyze writers’ emotions inherent in texts written by various entities.

Particularly, another object of the present invention is to collect opinions and comments from various entities by analyzing the emotions of online texts or texts published on SNSs.

To achieve the above-described objects, the present invention provides a method for analyzing an emotional index of a text, the method comprising the steps of: analyzing morphemes of the text; matching the analyzed morphemes with morpheme information stored in a database; and analyzing the emotional index of the text using the results matched with the morpheme information, in which the morpheme information may be information representing the degree of association between the morpheme and each item with respect to the properties of the morpheme divided into a plurality of items, and the step of analyzing the emotional index may analyze the degree of association between the plurality of items and the text depending on the morpheme information. Here, the step of analyzing the morphemes may analyze the text in units of sentences.

Meanwhile, in an embodiment of the present invention, the step of analyzing the emotional index may analyze the emotional index based on at least one of the part of speech of each morpheme, the position of the morpheme in the sentence, the order of arrangement of the morphemes, and the distance between the morphemes and may give a weight value depending on the positions of a negative polarity item and a negative among the morphemes included in a sentence.

In an embodiment of the present invention, the step of analyzing the emotional index may change the plurality of items to items set by a user and analyze the degree of association between the changed items and the text.

Meanwhile, according to the present invention, the morpheme information may include category information, and the step of analyzing the emotional index may analyze the emotional index of the text based on the category.

In another embodiment of the present invention, the method may further comprise, after the step of matching the morpheme information, the step of tracking the emotional index for analyzing the trend of change in the morpheme information over time.

In still another embodiment of the present invention, the morpheme information may be given a weight value depending on the degree of association between the morpheme and a specific property of the morpheme, and the step of analyzing the emotional index may analyze the emotional index when more than a predetermined number of morphemes having a specific weight value are included in the text. Moreover, the part of speech of the morpheme or the position of the morpheme may be considered in giving the weight value.

The present invention may be implemented as a computer-readable recording medium storing a program for executing the above-described method for analyzing the emotional index of the text.

According to the present invention, it is possible to analyze a writer’s emotion included in a text based on various elements.

Particularly, it is possible to collect opinions and comments from various entities by analyzing the emotions of online texts or texts published on SNSs.

Moreover, it is possible to utilize the results of the emotion analysis in the marketing, etc. and accurately analyze pending social issues, etc.

The above and other features and advantages of the present invention will become more apparent by describing in detail exemplary embodiments thereof with reference to the attached drawings in which:

FIG. 1 is a flowchart showing a method for analyzing an emotional index in accordance with an embodiment of the present invention;

FIG. 2 is a block diagram showing an apparatus for implementing a method for analyzing an emotional index in accordance with an embodiment of the present invention;

FIG. 3 is a diagram showing an example of a morpheme matched with morpheme information of the present invention;

FIG. 4 is a diagram showing the analysis results of the emotional index in accordance with an embodiment of the present invention;

FIG. 5 is a flowchart showing a method for analyzing an emotional index in accordance with another embodiment of the present invention;

FIG. 6 is a diagram showing an example in which items of morpheme information are changed in accordance with another embodiment of the present invention; and

FIG. 7 is a flowchart showing a method for analyzing an emotional index in accordance with still another embodiment of the present invention.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a flowchart showing a method for analyzing an emotional index in accordance with an embodiment of the present invention.

The method for analyzing the emotional index in accordance with an embodiment of the present invention comprises the steps of analyzing morphemes of a text, matching the analyzed morphemes with morpheme information stored in a database, and analyzing the emotional index of the text using the results matched with the morpheme information, in which the morpheme information is information representing the degree of association between the morpheme and each item with respect to the properties of the morpheme divided into a plurality of items, and the step of analyzing the emotional index analyzes the degree of association between the plurality of items and the text depending on the morpheme information.

The step of analyzing the morphemes analyzes the morphemes of a text posted or stored online, input by a user, or stored in a terminal (e.g., computer, smartphone, etc.). The analysis of morphemes divides the text into syntactic words (in the case of a single sentence text, the sentence is divided into syntactic words) and classifies the syntactic words based on parts of speech, thus deriving underlying forms of the morphemes. The underlying forms of the morphemes are selected based on the Guidelines for Morphological Analysis Corpus Construction of Modern Korean Language published by the National Institute of the Korean Language, which may be stored in a separate storage means of an apparatus in which the present invention is implemented or accessible to a separate external server or storage means for use as a standard for morphological analysis. Moreover, the morphemes may be extracted from emoticons, expletives, slang, neologisms, etc. which are widely used in daily life.

When the present invention is used to analyze in real time the texts containing opinions, emotions, etc. of various entities about online issues by crawling text information stored in servers in which a variety of digital information is uploaded and stored, especially, in SNS servers, its effectiveness is high.

The morphemes analyzed in the step of analyzing the morphemes are matched with the morpheme information stored in the database.

The morpheme information is information representing the degree of association between the morpheme and each item with respect to the properties of the morpheme divided into a plurality of items. The properties of the morpheme represent the emotions expressed by a certain morpheme, and examples thereof are shown in FIG. 3. In the examples shown in FIG. 3, ten kinds of emotions such as “satisfaction, relief, pleasure, interest, pride, dissatisfaction, fear, sadness, disgust, and anger” are set to a plurality of items to represent the properties of the morpheme. In the present invention, the properties of the morpheme are divided into the plurality of items, and thus various emotions can be analyzed more objectively and comprehensively than conventional methods that divide the emotions into positive, negative, and neutral.

The morpheme information of each morpheme is stored in the database. The database may be constructed by systematically organizing the degree of association between each morpheme and a certain item among the plurality of items, and the morpheme information stored in the constructed database is matched with the morpheme and used to analyze the emotional index.

After the analyzed morphemes are matched with the morpheme information, the emotional index of the text is analyzed using the result.

The step of analyzing the emotional index analyzes the degree of association between the plurality of items and the text depending on the morpheme information. The morpheme information of the full sentences or text is derived based on the morpheme information of each morpheme, and then the degree of association between the text and each item in the morpheme information is analyzed, thus analyzing the emotional index. In particular, the emotional index is analyzed based on the properties of the morpheme derived from the morpheme that expresses the nature, emotion, etc. such as a verb or adjective among the analyzed morphemes, and thus the intention of the text writer can be accurately identified.

In the following, the method for analyzing the emotional index according to the present invention will be described as an example of another actual text. For example, when analyzing the emotional index of a text such as “Christina is known for favoring provocative looks, but her music has always satisfied people's ears.”, the morphemes of the text are analyzed such that ‘favoring’ is ‘favor+ing, ‘satisfied’ is ‘satisfy+ed’, and ‘people's is ‘people+'s’. Other syntactic words in the text are analyzed in the above manner.

The analyzed morphemes are matched with the morpheme information stored in the database to derive the properties of each morpheme as shown in FIG. 3. Particularly, in the above example, the morphemes such as “provocative” and “satisfied” are important for identifying the subjective emotion of the text writer, and thus based on the properties of the morphemes, the emotional index of the text is analyzed by representing the emotion that the writer wants to express through the text as the degree of association between the morpheme and the plurality of items. In the following, the method for analyzing the emotional index in accordance with other embodiments of the present invention will be described in more detail.

In an embodiment of the present invention, when analyzing the morphemes of the text, the text is analyzed in units of sentences. Conventionally, the positive/negative emotions are determined in units of texts by identifying only the number of morphemes with positive meanings and the number of morphemes with negative meanings in the full text. However, the basic unit of meaning in the text is the sentence, and thus in the present invention, the morphemes are analyzed in units of sentences in the full text, and the emotional index is analyzed in units of sentences. Therefore, the intention of the text writer can be accurately determined.

Meanwhile, in an embodiment of the present invention, the step of analyzing the emotional index analyzes the emotional index based on at least one of the part of speech of each morpheme, the position of the morpheme in the sentence, the order of arrangement of the morphemes, and the distance between the morphemes. The present embodiment will be described as an example of a text such as “A few day ago, I felt the iced coffee was cool, but the steaming hot coffee is good for me yesterday and today”. In this text, the morphemes critical for analyzing the emotion of the writer are “iced coffee”, “cool”, hot”, and “good”. The morphemes such as “iced coffee” and “cool” are similar compared to other words with different morpheme information, and when both morphemes are considered at the same time, they supplement the meanings of each other, thus making it possible to accurately identify the emotion of the writer. If the morphemes such as “iced coffee” and “hot” are considered at the same time, the intention of the text writer is wrongly reflected, and thus the accuracy of the emotion analysis is lowered.

Even in such a sentence in which various morphemes are combined, the intention of the writer may be interpreted differently depending on the part of speech of each morpheme, the position of the morpheme in the sentence, the order of arrangement of the morphemes, and the distance between the morphemes, and accordingly the emotional index may be analyzed differently. In order to accurately analyze the emotional index, it is necessary to clearly and objectively identify the intention of the writer, and it is possible to increase the reliability of the analysis of the emotional index by analyzing the emotional index based on the above-described elements.

Meanwhile, in another embodiment of the present invention, it is possible to give a weight value depending on the positions of a negative polarity item and a negative among the morphemes included in a sentence. The reason for this is to analyze the emotional index by accurately reflecting the text writer’s intention to represent a strong positive due to a double negative.

Examples of the negative polarity items may include “not”, “never”, “not at all”, “nothing”, “nobody”, “far from”, “let alone”, “hardly”, “rarely”, etc. Each of these negative polarity items is used itself as the negative meaning, but when it is combined with a negative predicate, it creates an expression that emphasizes positive emotion.

When the emotional index is analyzed without considering the meaning of the double negative, the meaning of the text may be interpreted reversely. For example, when a sentence such as “Anybody does not deny the fact” is interpreted literally, the morphemes “anybody” and “does not” all contain negative meanings, and thus it is considered that a negative emotion is reflected in this sentence.

However, as in the present embodiment, when the positions of the negative polarity item and the negative are considered and a weight value is given to a strong positive intention due to a double negative, the above sentence is interpreted as having a positive meaning, like “Everybody recognizes the fact”, and the emotional index is analyzed as positive.

Examples of the results of analyzing the emotional index of the texts in accordance with the embodiments of the present invention are shown in FIG. 4. Referring to FIG. 4, the emotional index of the text is analyzed and schematized depending on the properties of the morpheme divided into a total of 10 items. It can be determined how much the text coincides with each item and how the emotion is generally distributed. The emotional index may be expressed by the graph shown in FIG. 4 or expressed numerically. The expression method is not limited. Meanwhile, as shown in FIG. 4, the plurality of items may be divided into positive items and negative items to analyze whether the text contains a positive emotion or a negative emotion.

FIG. 5 is a flowchart showing a method for analyzing an emotional index in accordance with another embodiment of the present invention.

In the present embodiment, the plurality of items that represent the properties of the morpheme may be changed by a user’s setting, and when changed, the emotional index of the text may be analyzed by analyzing the degree of association between the items set by the user and the text. The present embodiment will be described in more detail with respect to FIGS. 4 and 6.

In the above-described FIG. 4, the plurality of items are set depending on ten kinds of emotions (satisfaction, relief, pleasure, interest, pride, dissatisfaction, fear, sadness, disgust, and anger). These are the emotions set as the default in the above embodiment. However, in the analysis of the emotional index, it is necessary to analyze the emotional index of the text based on the emotion that the user wants to identify, if necessary, other than the preset emotions.

The present embodiment is to achieve this purpose and enables the user to analyze the degree of emotion that the text can express with respect to the items set by the user (e.g., beauty, sophistication, elegance, neatness, and charm in FIG. 6). That is, the items of the emotional index may be set depending on the user’s needs, and the text may be analyzed depending on the set items.

As an example for implementing this embodiment, the emotional index of the full text is analyzed depending on the items set as the default (10 items in FIG. 4), and the rate of concordance between the analyzed emotional index and the items set by the user is analyzed. The items set by the user correspond to the morphemes, and thus the items set by the user can be expressed by the items set as the default. That is, it is possible to derive the degree of association between the analysis results of the emotional index of the text, the items set by the user, and the items set as the default, and thus it is possible to derive the correlation and the rate of concordance between the emotional index of the text and the items set by the user. The results are shown in FIG. 6.

In the above example, the comparison between the degree of association with “beauty” according to the items set as the default and the degree of association with each item according to the analysis results of the text shows a rate of concordance of about 70%, wherein the “charm” shows a rate of concordance of 15%, the “neatness” shows a rate of concordance of 20%, the “elegance” shows a rate of concordance of 25%, and the “sophistication” shows a rate of concordance of 50%. Accordingly, the analyzed text can be represented as the rate of concordance with the items set by the user, and the degree of a certain emotion that the user reflects in the text can be determined.

As another example for implementing this embodiment, it is possible to analyze to what extent the text contains the emotion of the item set by the user by determining the rate of concordance between the morpheme information of each morpheme analyzed from the text and the morpheme information of each item set by the user, analyzing the text based on the items set by the user, instead of the items set as the default, and determining the degree of association. Moreover, in another embodiment of the present invention, the text may be analyzed in accordance with the purpose of the emotional index analysis by reversely tracking the morpheme that affects each emotion when the results of the entire emotional index analyzed from the text are calculated.

Meanwhile, in another embodiment of the present invention, category information may be further included in the morpheme information. In this case, the step of analyzing the emotional index may analyze the emotional index of the text based on the category. That is, the emotional index of a text for a specific category is analyzed by matching the morpheme included in the text with a corresponding category among various categories such as politics, people, economy, responses, places, brands, hobbies, etc. (while the noun in the morphemes is a main factor that determines the category, the category may be determined by other parts of speech).

Therefore, it is possible to analyze which emotion a specific text has in the political stance and which emotion the text contains for a specific person.

The present embodiment further comprises, after the step of matching the morpheme information in the previous embodiment, the step of tracking the emotional index for analyzing the trend of change in the morpheme information over time. The step of tracking the emotional index is performed independently from the step of analyzing the emotional index. In the case of a common noun among the morpheme information, the emotion contained in the common noun varies due to various reasons. Therefore, it is necessary to periodically update the morpheme information stored in the database, and in the present embodiment, the trend of change in the morpheme information due to the update of the morpheme information is analyzed and provided to the user. According to the present embodiment, it is possible to track people's' perception about a certain person, company, product, brand, etc. in an SNS, and thus the results of the emotion analysis can be effectively utilized in the marketing, etc.

In another embodiment of the present invention, the morpheme information is given a weight value depending on the degree of association between the morpheme and a specific property of the morpheme, and the step of analyzing the emotional index analyzes the emotional index when more than a predetermined number of morphemes having a specific weight value are included in the text. The present embodiment is to increase the reliability of the emotional index analysis and will be described with reference to the following table.

Word	Satis-faction	Relief	Pleasure	Interest	Pride	Happy/ Positive	Dissatis-faction	Fear	Sadness	Disgust	Anger	Unhappy/Negative	Weight value
Happy	393	36	500	0	36	965	0	0	0	0	36	36	5
Sad	26	0	26	0	0	53	26	53	763	53	53	947	5
Glad	257	29	629	57	0	971	0	0	0	0	29	29	5
Lovable	0	83	750	0	0	833	0	0	167	0	0	167	4
Full	694	102	82	0	0	878	20	61	0	0	41	122	4

Referring to the above table, the degree of association between each morpheme and a plurality of items sets as the default is described by numerical values. The plurality of items are classified into specific properties (happy/positive and unhappy/negative), and the weight value is determined depending on the numerical value. In the above table, the weight value is determined as 5 for more than 900 points out of 1000 points and as 4 for 800 to 900 points. That is, the weight value is an index indicating a specific emotion (happy or unhappy) expressed by the plurality of items, and when the corresponding weight value is higher, it represents a strong emotion.

In the present embodiment, the emotional index is analyzed only when more than a predetermined number (e.g., 2) of morphemes (e.g., happy or glad) having a specific weight value (e.g., 5) are included in the text. In the absence of the morpheme that represents a strong emotion, it is highly likely that the significance of the text as the object of the emotional index analysis is low and the text describes a fact only, and thus the corresponding text is filtered out in the analysis of the emotional index according to the present invention.

In addition to this, the present invention may give weight values differently in terms of the part of speech of the morpheme or the position of the morpheme. As an example, the words such as very, extremely, too, most, etc. increase the emotion of the following morpheme, and when these words are used, for example, in the case of “very happy” or “very sad”, the degree of association may be set to be higher, and the weight value may also be calculated to be higher (e.g., twice) than the weight value of the morpheme.

The method for analyzing the emotional text according to various embodiments of the present invention may be implemented in the form of a program for executing the method, and the program may be implemented in the form of a computer-readable recording medium. The computer-readable recording medium does not simply mean a storage device such as CD, HDD, etc., but includes an apparatus, server, or system accessible through a computer or terminal to utilize data.

FIG. 2 is a block diagram showing an apparatus for implementing a method for analyzing an emotional index in accordance with an embodiment of the present invention.

An emotional index analyzer of the present invention may comprise a morpheme analyzing module for analyzing morphemes of a text, a morpheme information matching module for matching the analyzed morphemes with morpheme information stored in a database, and an emotional index analyzing module for analyzing the emotional index of the text using the result matched with the morpheme information. The emotional index analyzer may be driven in conjunction with the database storing the morphemes matched with the morpheme information. That is, the method for analyzing the emotional index according to the present invention may be executed by a computer-readable recoding medium and may also be executed by an independent apparatus such as the emotional index analyzer.

The embodiments of the present invention have been described for illustrative purposes, and those skilled in the art will appreciate that various changes, modifications, and additions are possible within the technical scope of the invention and are within the scope of claims of the present invention.

Claims

A method for analyzing an emotional index of a text, the method comprising the steps of:
analyzing morphemes of the text;
matching the analyzed morphemes with morpheme information stored in a database; and
analyzing the emotional index of the text using the results matched with the morpheme information,
wherein the morpheme information is information representing the degree of association between the morpheme and each item with respect to the properties of the morpheme divided into a plurality of items, and
wherein the step of analyzing the emotional index analyzes to what extent each item of the properties of the morpheme, which form the morpheme information, is included in the text.
The method of claim 1, wherein the step of analyzing the morphemes analyzes the text in units of sentences.
The method of claim 2, wherein the step of analyzing the emotional index analyzes the emotional index based on at least one of the part of speech of each morpheme, the position of the morpheme in the sentence, the order of arrangement of the morphemes, and the distance between the morphemes.
The method of claim 2, wherein the step of analyzing the emotional index gives a weight value depending on the positions of a negative polarity item and a negative among the morphemes included in a sentence.
The method of claim 1, wherein the step of analyzing the emotional index changes the plurality of items to items set by a user and analyzes the degree of association between the changed items and the text.
The method of claim 1, wherein the morpheme information further includes category information, and the step of analyzing the emotional index analyzes the emotional index of the text based on the category.
The method of claim 1, further comprising, after the step of matching the morpheme information, the step of tracking the emotional index for analyzing the trend of change in the morpheme information over time.
The method of claim 1, wherein the morpheme information is given a weight value depending on the degree of association between the morpheme and a specific property of the morpheme, and
wherein the step of analyzing the emotional index analyzes the emotional index when more than a predetermined number of morphemes having a specific weight value are included in the text.
The method of claim 8, wherein the method gives weight values in terms of the part of speech of the morpheme or the position of the morpheme.
A computer-readable recording medium storing a program for executing the method for analyzing the emotional index of the text described in any one of claims 1 to 9.