KR102398683B1 - System and Method for Constructing Emotion Lexicon by Paraphrasing and Recognizing Emotion Frames - Google Patents

Abstract

The present invention provides a method for constructing an emotion dictionary and recognizing an emotion structure performed on a recording medium recorded in a computer-executable language, through paraphrasing, to construct an emotion dictionary including emotion categories in phrase units. Step, and based on the built-up emotion dictionary, the emotional structure including the emotional category for the emotional expression appearing in the received text, the emotional subject corresponding to the emotional expression, and the emotional object corresponding to the emotional expression are recognized including the steps of

Description

Construction of an emotion dictionary using paraphrasing and a system and method for recognizing emotional structures in text using the same {System and Method for Constructing Emotion Lexicon by Paraphrasing and Recognizing Emotion Frames}

The present invention relates to a system and method for constructing an emotion dictionary through paraphrase and recognizing an emotion structure in a text using the same. More specifically, semi-automatically constructs a dictionary that collects emotion vocabulary for each emotion category belonging to Ekman's basic emotion model (joy, sadness, fear, anger, surprise, disgust), and uses the built emotion dictionary and syntax analysis technique Thus, it relates to a system and method for finding the various emotions appearing in the text, the subject who feels each emotion, and the object that induces the emotion.

The technology for recognizing emotions from texts written in natural language is recognized as an important technology for analyzing product reviews, social phenomena, and markets. In order to develop such emotion recognition technology, studies are being conducted to recognize emotions from various text types (dialogs, SNS posts, blogs, etc.). Attempts to analyze emotions from texts have been mainly focused on recognizing the polarity of emotions (positive/negative) of texts using sentiment analysis technology (Registration Patent No. 10-1555-0390000, (2015.09. 16.), Registered Patent No. 10-1346-1150000, (2013.12.23.), Registered Patent No. 10-1540-6830000, (2015.07.24.)). However, information on the sentiment of simple emotions such as positive or negative has its limitations in deep emotion analysis. Therefore, the need for technology that recognizes various deduction categories for more detailed emotion analysis is emerging.

The present invention has been devised to solve the above problems, and a system and method for constructing a phrase-level emotion dictionary composed of emotion vocabulary for various emotion categories and recognizing various emotion structures appearing in text using this Its purpose is to provide

However, the problem to be solved by the present invention is not limited to the above purpose, and may be variously expanded without departing from the spirit and scope of the present invention.

A method for constructing an emotion dictionary and recognizing an emotion structure performed on a recording medium recorded in a language executable on a computer according to an embodiment of the present invention for solving the above problems is, through paraphrasing, a phrase unit Building an emotion dictionary including an emotion category, and based on the built emotion dictionary, an emotion category for an emotional expression appearing in the received text, an emotional subject corresponding to the emotional expression, and the emotional expression corresponding to the emotional expression and recognizing an emotional structure including an emotional object.

In some embodiments of the present invention, the step of constructing the emotion dictionary may include expanding the emotion word in phrase units through paraphrase.

In some embodiments of the present invention, the step of constructing the emotion dictionary may be performed to exclude an expression other than the emotional expression corresponding to the same word is selected.

In some embodiments of the present invention, the step of constructing the emotion dictionary may be performed to exclude a case in which emotion polarities of two expressions of a paraphrase pair are different from each other.

In some embodiments of the present invention, the step of recognizing the emotional structure may be performed to exclude an expression other than an emotional expression corresponding to a synonymous word among expressions in the text from being recognized.

In some embodiments of the present invention, the step of recognizing the emotional structure may be performed to recognize the subject of the emotion and the object of the emotion related to the recognized emotion expression.

Other specific details of the invention are included in the detailed description and drawings.

According to an embodiment of the present invention, various types of emotional vocabulary having the same meaning and emotional category can be extracted by taking into account the meaning of a predetermined categorized emotional vocabulary through paraphrase.

In addition, according to an embodiment of the present invention, it is possible to extract an emotion vocabulary with high accuracy in consideration of synonyms.

In addition, according to an embodiment of the present invention, it is possible to extract an emotional expression composed of a plurality of expressions by constructing a phrase-based emotion dictionary including words.

In addition, according to an embodiment of the present invention, it is possible to filter on the incorrectly extracted emotion vocabulary.

In addition, according to an embodiment of the present invention, it is possible to perform detailed emotion analysis by recognizing various emotional structures appearing in text.

In addition, according to an embodiment of the present invention, emotion recognition can be performed on an emotional expression composed of a plurality of words by using a phrase-based emotion dictionary.

In addition, according to an embodiment of the present invention, it is possible to solve the problem of identical words that may occur in the process of comparing the vocabulary of the emotion dictionary with the vocabulary on the text.

However, the effects of the present invention are not limited to the above effects, and may be variously expanded without departing from the spirit and scope of the present invention.

1 is a flowchart of a method of constructing an emotion dictionary through paraphrase and recognizing an emotion structure of a text according to an embodiment of the present invention.
2 illustrates the structure of a system for constructing an emotion dictionary through paraphrase according to an embodiment of the present invention.
3 is a flowchart of a system for recognizing an emotion structure on a text according to an embodiment of the present invention.

Advantages and features of the present invention and methods of achieving them will become apparent with reference to the embodiments described below in detail in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but will be embodied in various different forms, and only these embodiments allow the disclosure of the present invention to be complete, and common knowledge in the art to which the present invention pertains It is provided to fully inform those who have the scope of the invention, and the present invention is only defined by the scope of the claims.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS [0012] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS [0010] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS [0010] Reference is made to the accompanying drawings, which show by way of illustration specific embodiments in which the present invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the present invention. It should be understood that the various embodiments of the present invention are different but need not be mutually exclusive. For example, certain shapes, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the present invention with respect to one embodiment. In addition, it should be understood that the position or arrangement of individual components in each disclosed embodiment may be changed without departing from the spirit and scope of the present invention. Accordingly, the detailed description set forth below is not intended to be taken in a limiting sense, and the scope of the present invention, if properly described, is limited only by the appended claims, along with all scope equivalents as those claimed.

The terminology used herein is for the purpose of describing the embodiments, and is not intended to limit the present invention. In this specification, the singular also includes the plural unless specifically stated otherwise in the phrase. As used herein, "comprises" and/or "comprising" refers to the presence of one or more other components, steps, operations and/or elements mentioned. or addition is not excluded.

Unless otherwise defined, all terms (including technical and scientific terms) used herein may be used with the meaning commonly understood by those of ordinary skill in the art to which the present invention belongs. In addition, terms defined in a commonly used dictionary are not to be interpreted ideally or excessively unless specifically defined explicitly.

Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the accompanying drawings. The same reference numerals are used for the same components in the drawings, and repeated descriptions of the same components are omitted.

1 is a flowchart illustrating a method for constructing an emotion dictionary through paraphrasing and recognizing an emotion structure of a text according to an embodiment of the present invention.

In the method for constructing an emotion dictionary through paraphrasing and recognizing an emotion structure of text according to an embodiment of the present invention, each step is recorded in a program language that can be executed on a computer, and can be executed on the computer.

Referring to FIG. 1 , the method for constructing an emotion dictionary through paraphrase and recognizing an emotion structure according to an embodiment of the present invention includes a step of constructing an emotion dictionary through paraphrase (S100), a step of receiving text from a user (S200), and recognizing the emotion structure from the input text using the built emotion dictionary (S300).

A recording medium in which the above steps are recorded in a language executable by a computer or the computer program itself may be an embodiment in which the present invention is implemented. First, the step of constructing an emotion dictionary through paraphrase according to an embodiment of the present invention will be described.

Step of constructing an emotion dictionary through paraphrase (S100)

2 is a flowchart for explaining in detail the step of constructing an emotion dictionary through the paraphrase of FIG. 1 according to an embodiment of the present invention.

In the step (S100) of constructing an emotion dictionary through paraphrase according to an embodiment of the present invention, paraphrase is performed based on an emotion vocabulary of a predetermined word level corresponding to an emotion expression for a predetermined emotion category (see Table 1 below). A step of constructing an emotion dictionary by generating various types of emotional vocabulary through It consists of a step (S130) of filtering the emotional vocabulary containing errors among the extracted emotional expressions, and a step (S140) of building the extracted emotional expression in the form of a dictionary.

First, the step of extracting a paraphrase pair from the parallel corpus ( S110 ) is a step of extracting a phrase-unit paraphrase pair from a parallel corpus composed of two languages. A paraphrase pair in the present invention means an expression pair paired with a phrase-unit "original expression" and an expression composed of another word form that preserves its meaning.

For paraphrase pair extraction, in the present invention, a phrase alignment process is performed between an original sentence constructed in parallel on a corpus and a translated sentence for it. Phrase-wise alignment refers to a process of “semantic alignment” between the original and translated phrases of phrases that have the same meaning in parallel sentences.

(Example 1)

For example, a pair of parallel sentences < "I am happy." , "I am happy." > from < "I" , "I" > , < "happy" , "am happy" > , < "." , "." In the form of >, phrase-wise expression sorting can be performed.

As a computer program for sphere unit alignment, for example, GIZA++ may be used. The computer program calculates and aligns the translation probability between the two language expressions by using the frequency of the expressions co-occurring in both languages over several translation sentence pairs. In the present invention, in order to solve the problem that the accuracy of word alignment is damaged because it is regarded as a different word even though it is the same expression due to a change in the morphology of one word and is not reflected in the frequency calculation, only the stem of each word is extracted and word alignment is performed.

(Example 2)

For example, in the case of "happy", only the stem "happy" is extracted so that it is aligned with "am happy".

After alignment, the translation probability between two different language expressions can be derived by the following Equation (1).

[Equation (1)]

In an embodiment of the present invention, k is a Korean expression, e means an English expression, count(e) is the number of occurrences of the corresponding English expression in the corpus, and count(k,e) is a parallel corpus between the Korean expression and the English expression. It refers to the frequency of simultaneous occurrences in a translation pair.

After aligning phrase-by-phrase expressions, only those that are grammatically the same grammatical category (tense, voice, case, etc.) among two expressions sharing the same translation are considered as a paraphrase pair and stored. Identification of grammatical categories can be confirmed by examining the suffixes interlaced in the stem. In a paraphrase pair, the paraphrase relationship probability between the original expression and the paraphrase expression can be derived by the following Equation (2).

[Equation (2)]

Next, the step of extracting the emotional expression from the paraphrase pair ( S120 ) is a process of extracting the emotional expression of a phrase unit from the generated paraphrase pair based on predetermined categorized emotional vocabulary. To this end, first, various words expressing emotions are collected, and a process of determining the emotional category to which they belong from at least three evaluators is performed. The emotion category in the embodiment of the present invention is defined using six emotion categories (joy, sadness, fear, anger, surprise, disgust) of the Ekman model, but these emotion categories are not limited to the example of the Ekman model. Table 1 below summarizes the emotional expressions assigned to each emotion category as an embodiment of the present invention.

emotion category expression of emotion pleasure lovely, fun, delightful sadness grieve, grieve, mourn fear afraid, anxious, anxious anger rage, resent, tremble surprised admiring, bewildered, bewildered aversion contempt, displease, shiver

The emotion category to which each word belongs is determined by the emotion category selected by the largest number of evaluators, and the probability that the corresponding word belongs to the assigned emotion category is defined as the ratio selected as the emotion category from the total number of appraisers.

In the present invention, manual evaluation is used to categorize predetermined words, but in some cases, if a pre-built categorized emotion dictionary exists, it may be used. does not

And, if there is a stem of the emotion word collected in the original expression of the paraphrase pair, all the pairs are extracted and the corresponding emotion category is assigned. This is because expressions including the corresponding emotional word are more likely to have the same emotional category. The reason for extracting based on the stem is to prevent a case in which the original expression of a paraphrase pair cannot be extracted due to a change in the word form due to affixation even though it has the same stem as the collected word.

(Example 3)

Collected emotion word: "Angry" [Emotion Category: Anger]

The stem of the emotion word: "Angry"

Extracted paraphrase pairs: < "He looked angry" , "I was angry" > [Emotional Category: Anger]

In the present invention, the probability that the original expression of the paraphrase pair is an assigned emotion category is the same as the probability of the corresponding emotion word. The probability of the assigned emotion category for the paraphrase expression is defined as the joint probability of the emotion category probability of the original expression and the paraphrase probability of the paraphrase expression. In the present invention, by generating an emotional expression having the same meaning through paraphrase, it is possible to more accurately assign an emotional category.

Next, the step of filtering the emotional vocabulary containing the error ( S130 ) is a process of removing the expression pairs containing the error with respect to the emotional expression candidate paraphrase pair extracted in the above process. The following five types of errors are derived through the error analysis process for the present invention.

First, when searching for the original expression of the paraphrase including the stem of the emotional word in the above-mentioned step, it is an error that occurs when the original expression, not the emotional expression, is found by the same word. As an example, in the case of the word "figure" corresponding to the emotional category "sadness", the original expression of the paraphrase pair is matched with the stem "numerous" of an expression such as "measure the number", which is a meaning related to a number, and is used as an emotional expression. in case of recognizing

In order to solve the above-mentioned problem, in the present invention, the case where the part-of-speech of the stem of the emotional word is different or the case where the translation expression shared by the paraphrase pair is not the expression having the polarity of the emotion is excluded. If the translated expression is not an emotional expression, it is highly likely that the paraphrase pair is also not an emotional expression. When the word ambiguity problem is limited to emotional words, it can be helpful to solve the above-mentioned problem because the problem related to the same word occurs with a very low probability. In the embodiment according to the present invention, the part-of-speech recognition may be performed through a morpheme analyzer, and in the case of an emotional expression, it may be performed using a dictionary constructed in relation to the polarity of a conventional emotion.

Second, it is a case in which meaning is added or lost to the original expression or paraphrased expression, so that even though it contains emotional words, it is not a pure emotional expression. As an example, the original expression “Nobel Peace Prize” containing the word “peace” for “joy” is not an expression of emotion. Also, when an object is added to an emotional expression, it is difficult to see it as a pure emotional expression. As an example, in the case of the word “fun” corresponding to “joy”, it is difficult to see the original expression “my friend is having fun” containing the root of the word as a pure emotional expression. Therefore, in the present invention, a paraphrase pair having an entity name or an expression including a proper noun/pronoun is excluded. The process of recognizing an entity name and recognizing a part-of-speech of a noun may be performed using an entity name recognizer and a morpheme analyzer.

Third, it is a case where the paraphrase expression has a completely different meaning from the original expression due to a paraphrase failure. In order to solve the above-mentioned problem, in the embodiment of the present invention, the case where the paraphrase probability of the paraphrased expression is less than or equal to a specific threshold is excluded.

Fourth, it is a case where the sentiments of the two emotional expressions of the paraphrase pair are different. Since the paraphrase expression has the same meaning as the original expression, it is self-evident that the original expression has the polarity of emotion. Therefore, if the two expressions have different polarities, the corresponding paraphrase pair is excluded. Polarity analysis of expressions may be performed using a dictionary related to various emotional polarities proposed in the prior art.

The process of the above-described error filtering step can solve problems related to synonyms that have not been considered in conventional studies, and can contribute to improving the quality of the emotion dictionary according to error analysis.

Subsequently, the step of constructing the emotion dictionary ( S140 ) is a process of constructing the emotion words and the extracted emotion expressions in the form of an emotion dictionary configured in the following format.

“Emotional category: Emotional expression: Emotional expression annotated with part-of-speech of morpheme: Stem of emotional expression annotated with part-of-speech of morpheme: Emotion category probability”

(Example 4)

go. Joy: Fun: Fun/VA Da/EF: Fun/VA: 1.0

me. Joy: I felt good: I felt good/NNG was/JKS was good/VA was/EP was everything/EF: I was feeling/NNG was good/VA: 1.0

Step of receiving text input from the user (S200)

A text according to an embodiment of the present invention includes a paragraph or a document consisting of a single sentence to a plurality of paragraphs.

Recognizing the emotional structure from the received text (S300)

3 is a flowchart of a system for recognizing an emotion structure on a text according to an embodiment of the present invention. In the step of recognizing the emotional structure of the text (S300), the emotional expression in the text is recognized and the emotional category is assigned using the phrase-based emotion dictionary built in the above-described step, and the subject of the emotion and the object of the emotion related to the emotion are identified. This is the recognition stage. This includes a text syntax analysis step (S310), an emotion category recognition step (S320), and an emotion subject and object recognition step (S330). Through this, it is possible to recognize the detailed emotional structures existing in a part of the text, rather than recognizing the emotion expressed by the sentence or the entire document.

In the text syntax analysis step S310, sentences are separated from the text input by the user, and morpheme analysis, phrase structure grammar analysis, and dependency grammar analysis are performed for each sentence. For this purpose, a stemming analyzer, a phrase structure parser, a dependency parser, or the like may be used.

The emotion category recognition step S320 is a process of finding an emotion expression from each sentence through the emotion dictionary built in the above step and assigning an emotion category. When searching for an emotional expression in a sentence from an emotion dictionary, a problem with the same word may occur as shown in the following example.

(Example 5)

Sentiment Dictionary Vocabulary: Sadness: Shame: Shame/NNG: Shame/NNG: 0.6666

Given sentence: “According to the figures investigated, this is a serious problem.”

In order to solve the problem of cognate words, in the present invention, only a word having a stem corresponding to the fourth field of the emotion dictionary vocabulary and a corresponding part-of-speech is recognized as an emotional expression. Also, in the case of phrases, there are cases where the expression of emotion is made clear by various qualifiers (eg: I felt a sense of shame), but in the case of words, the problem of synonymous words can arise more seriously. Therefore, only in the case of words, in addition to the part-of-speech agreement, if there is an expression with emotional polarity among words that directly or indirectly depend on the word in the sentence, the emotion category recorded in the dictionary is assigned by assuming that the expression is the emotional expression in the sentence. .

The emotional subject and object recognition step S330 is a process of searching for the emotional subject and object related to the emotional expression searched for in the above-described step. In an embodiment of the present invention, a noun phrase including a subject having a direct or indirect dependence relationship with the searched emotional expression is recognized as the subject of emotion, and a noun phrase or entity including an object, and an adverb phrase including an event target can be recognized.

(Example 6)

he/[NP_SBJ;noun-phrase_subject]/[NP_AJT;noun-phrase_verb-modifier] vehemently/AP_AJT[adverb-phrase_verb modifier] angry/[anger]

Dependency 1: he (including subject) -> angry

Dependency 2: at me (including object) -> angry

In the above, the embodiment has been mainly described, but this is only an illustration and does not limit the present invention, and those of ordinary skill in the art to which the present invention pertains are not exemplified above in the range that does not depart from the essential characteristics of the present embodiment. It can be seen that various modifications and applications are possible. For example, each component specifically shown in the embodiment can be implemented by modification. And differences related to such modifications and applications should be construed as being included in the scope of the present invention defined in the appended claims.

S100: Steps of building an emotion dictionary through paraphrase
S200: Step of receiving text input from the user
S300: Recognizing the emotional structure from the received text

Claims (9)

As a method of constructing an emotion dictionary and recognizing an emotion structure performed on a recording medium recorded in a language executable by a computer, the method comprising:
constructing an emotion dictionary including emotion categories in phrase units through paraphrasing; and
recognizing an emotional structure including an emotional category for an emotional expression appearing in the received text, an emotional subject corresponding to the emotional expression, and an emotional object corresponding to the emotional expression, based on the built-up emotion dictionary; including,
The step of building the emotion dictionary is,
extracting a phrase-unit paraphrase pair from a parallel corpus composed of two languages;
extracting an emotional expression from the extracted paraphrase pair; and
Constructing the extracted emotional expression in the form of a dictionary; Containing, a method of constructing an emotion dictionary and recognizing an emotion structure. According to claim 1,
The step of constructing the emotion dictionary is performed including the step of expanding emotion words in phrase units through paraphrase, emotion dictionary building and emotion structure recognition method. 3. The method of claim 2,
The step of constructing the emotion dictionary is performed so as to exclude selection of an expression other than the emotional expression corresponding to the same word, the emotion dictionary building and emotion structure recognition method. 3. The method of claim 2,
The step of constructing the emotion dictionary is performed so as to exclude cases in which the emotion polarities of the two expressions of the paraphrase pair are different from each other. According to claim 1,
The step of recognizing the emotional structure is performed to exclude an expression other than an emotional expression corresponding to the same word among expressions in the text from being recognized. 6. The method of claim 5,
The step of recognizing the emotional structure is performed to recognize the subject of the emotion and the object of the emotion related to the recognized emotion expression, the method of constructing an emotion dictionary and recognizing the emotion structure. According to claim 1,
The step of extracting the paraphrase pair,
A method for constructing an emotion dictionary and recognizing an emotion structure, performing phrase alignment between an original sentence constructed in parallel on the corpus and a translated sentence therefor. 8. The method of claim 7,
The step of extracting the paraphrase pair,
A method of constructing an emotion dictionary and recognizing an emotion structure, calculating a translation probability between the two language expressions by using the frequency of expressions that appear together in the two languages over a plurality of translation sentence pairs, and performing the phrase unit alignment. 9. The method of claim 8,
A method for constructing an emotion dictionary and recognizing an emotion structure, wherein the translation probability between the two language expressions satisfies the following equation.

(k: first language expression, e: second language expression, count(e): number of appearances of second language expression in corpus, count(k,e): first and second language expression translation on parallel corpus frequency of occurrence in pairs)

Images