Opinion network having content or direction and implementation method thereof

Abstract

The present invention relates to an opinion network, a sentiment network, or an emotion network having content (positive, negative, neutral, etc.) or direction, and an implementation method thereof, and more specifically, to an opinion network having content or direction capable of simultaneously and visually showing positive and negative opinions of relations, which appear in various articles, with people, issues, institutions, and the like, and an implementation method thereof.

Description

Opinion network with content or direction and its implementation method

[0001] The present invention relates to an opinion network having content (positive, negative, neutral, etc.) or direction, or a sentiment network, an emotion network, and a method for implementing the same, and more specifically, to an opinion network having content or direction that can simultaneously and visually show positive and negative opinions about relationships between people, issues, institutions, etc. that appear in various texts, and a method for implementing the same.

[0002] Although text network analysis exists, it has the limitation of showing the co-occurrence of simple words (Doerfel & Barnett, 1999) and cannot see the content, direction, and relationship of positive and negative feelings between objects (people, issues, institutions, etc.).

[0003] For example, a text network according to conventional technology, as shown in Figure 1, lacks arrows. This means that it has limitations in that it is impossible to know the positive and negative relationships between opinions or emotions, as well as their direction and magnitude.

[0004] Although there are many studies on the co-occurrence of words, there is no technology that shows the relationship between words, that is, the content and direction of the positive and negative relationship between objects (people, issues, institutions, etc.).

[0005] Accordingly, as shown in Fig. 2, there is a need for technology development for an opinion network that has content or direction that can simultaneously and visually show positive and negative opinions regarding relationships with people, issues, institutions, etc. that appear in various texts.

[0006] [Prior Art Literature]

[0007] [Non-patent literature]

[0008] (Non-patent literature 1) - Butts, CT (2008) Social network analysis with sna. J Stat Softw 24(6):1-51.

[0009] (Non-patent Document 2) - Doerfel, ML, & Barnett, GA (1999). A semantic network analysis of the International Communication Association. Human communication research, 25(4), 589-603.

[0010] (Non-patent Document 3)- Liu, B. (2011). Tutorial: Sentiment Analysis and Opinion Mining. Retrieved September 6th, 2015, from source Web site: http: / www.cs.uic.edu / ~liub / FBS / Sentiment-Analysis-tutorial-AAAI-2011.pdf

[0011] (Non-patent Document 4)- Liu, B. (2012) Sentiment Analysis and Opinion Mining. Morgan & Claypool Publishers.

[0012] (Non-patent Document 5) - Wasserman S, & Faust K. (1994). Social network analysis: Methods and applications. Cambridge University Press

[0013] (Non-patent literature 6) - Ahn Yeon-seon, & Song Jae-hong. (2020). Knowledge network analysis of [educational psychology research]: Focusing on the structural relationships among research topics, researchers, research subjects, and research methods. Journal of Educational Psychology Research, 34(4), 677-703.

[0014] The purpose of the present invention is to provide a method for implementing an opinion network having content or direction that allows one to easily grasp the relationships between people, issues, institutions, etc. appearing in various documents, writings, and texts at a glance without having to read the full text of the documents, writings, and texts by visually showing both positive and negative opinions.

[0015] In order to achieve the above-described object, a method for implementing an opinion network having content or direction according to the present invention comprises: a data collection step of collecting data containing content regarding a subject of interest; an edge list creation step of coding the collected data according to a predetermined protocol to create an edge list regarding the subject of interest; and a relationship visualization step of visually displaying the relationship between the subject of interest and other objects based on the created edge list; wherein the edge list creation step comprises: an opinion existence determination step of determining whether an opinion regarding the subject of interest exists in the collected data; an opinion content determination step of determining whether the content of the opinion is positive or negative if the opinion exists; and an opinion directionality determination step of extracting a subject expressing the opinion and an object receiving the opinion from a sentence to determine the directionality of the opinion; wherein the relationship visualization step is characterized in that the directionality of the opinion is displayed using the direction of an arrow between the subject and the object, and the content of the opinion is displayed by varying the shape or color of the arrow between the subject and the object or by displaying it with a predetermined character near the arrow.

[0016] Here, the edge list creation step is characterized by creating a list based on sentences in which opinions exist in the collected data, and creating the edge list by entering a set of writing items into each of the created lists.

[0017] Here, the edge list creation step includes the subject, the object, the content of the opinion, and the intensity of the opinion in the creation items of the edge list, and the intensity of the opinion is expressed as a number according to a set standard, and if the content of the opinion is positive, it is expressed as +, and if it is negative, it is expressed as -, and the relationship visualization step is characterized by calculating the total, maximum, minimum, or average value of the intensity of the opinion expressed as a number, and displaying it by changing the shape or color of the arrow between the subject and the object, or displaying it as a set character near the arrow.

[0018] Here, the edge list creation step is performed by a plurality of coders, and the method for implementing an opinion network having the content or direction is characterized in that it further includes an inter-coder reliability measurement step for measuring reliability between coders who performed the edge list creation step.

[0019] Here, the edge list creation step is characterized in that it is performed by an AI coder using artificial intelligence technology.

[0020] In addition, the method for implementing an opinion network having content or direction according to the present invention further includes a preprocessing step of preprocessing the collected data before the edge list creation step; wherein the preprocessing step is characterized by including a relevance criteria extraction step of determining and extracting data recognized as being related to a specific event or specific topic determined by a user among the collected data.

[0021] Here, the relevance criteria extraction step is characterized by arranging the plurality of data whose relevance is recognized in chronological order, and the edge list creation step is characterized by analyzing the plurality of data in chronological order to create the edge list.

[0022] Here, the relevance criterion extraction step is characterized in that, if the relevance is not determined within a sentence, the relevance is determined based on the paragraph containing the sentence.

[0023] Here, the preprocessing step is characterized by further including a period-based extraction step for extracting data corresponding to a period range set by the user among the collected data.

[0024] Here, the edge list creation step is characterized in that a list is created based on a sentence in which an opinion exists, and if the sentence is composed of a complex sentence, a compound sentence, or a multiple sentence, the sentence is divided into multiple short sentences and the list is created by repeating the number of short sentences, if there are multiple opinions in one sentence, the list is created by repeating the number of opinions, and if there are multiple subjects or objects in one sentence, the list is created by repeating the number of products of the subjects and the objects.

[0025] Here, the edge list creation step is characterized in that, if the subject, the object, the content of the opinion, or the intensity of the opinion included in the creation items of the edge list is not determined within a sentence, the subject, the object, the content of the opinion, or the intensity of the opinion is determined based on the paragraph including the sentence.

[0026] Here, the edge list creation step is characterized by assigning a representative name that can represent a name or expression having identity, and creating the edge list based on the assigned representative name.

[0027] Here, the edge list creation step is characterized in that the creation items of the edge list further include at least one of a document number assigned to each of the collected data, a sentence number within the document assigned the document number, the date and time the collected data was published or posted, and whether it is a topic sentence and picture.

[0028] By the above configuration, the opinion network having content or direction according to the present invention and the method for implementing the same can show at a glance the positive and negative content and direction and relationship between objects (people, issues, institutions, etc.) appearing in a text, and as a result, the complex relationship between objects can be quickly grasped at a glance, and even an illiterate person can easily grasp the relationship between objects in a newspaper, magazine, etc. text at a glance without reading the text, and has the advantage of reducing language barriers as an opinion network.

[0029] Furthermore, the opinion network and its implementation method, which have content or directionality according to the present invention, have the advantage of being applicable to newspapers, magazines, social media, marketing, online e-commerce, and consumer opinions on products of retailers.

[0030] In addition, it is believed that applying the opinion network having content or direction according to the present invention and the method for implementing the same to other foreign languages ​​(English, Japanese, Chinese, German, French, Russian, etc.) will greatly contribute to promoting communication between people that is otherwise disconnected.

[0031] In other words, when we read foreign articles or newspapers (English, Chinese, French, German, Russian, etc.), it is difficult to understand the relationships between people because their names are unfamiliar. However, if we can quickly understand the basic content through the Opinion Network, it is expected that acquisition of the text will be faster and global communication will become easier.

[0032] Figure 1 is an example of a text network according to the prior art.

[0033] Figure 2 is an example diagram of implementing (b) of Figure 1 as an opinion network with content or directionality.

[0034] Figure 3 is a flowchart of a method for implementing an opinion network having content or direction according to the first embodiment of the present invention.

[0035] Figure 4 is an example of an edge list creation item according to one embodiment of the present invention.

[0036] Figure 5 is a flowchart of an edge list creation step according to one embodiment of the present invention.

[0037] Figure 6 is an example of edge list input and relationship visualization according to one embodiment of the present invention.

[0038] Figure 7 is an example diagram showing an example of inputting an example from a newspaper into an edge list according to an embodiment of the present invention and visualizing the relationship.

[0039] Figure 8 is an example diagram of an opinion network having content or directionality according to an embodiment of the present invention using Netminer as a means of relationship visualization.

[0040] Figure 9 is an example of relationship visualization according to one embodiment of the present invention that reflects the strength of opinions.

[0041] Figure 10 is a flowchart of a method for implementing an opinion network having content or direction according to a second embodiment of the present invention.

[0042] Figure 11 is a flowchart of a method for implementing an opinion network having content or direction according to a third embodiment of the present invention.

[0043] Figure 12 is an example diagram of an opinion network having content or directionality implemented through a relevance criteria extraction step in a preprocessing step according to the third embodiment of the present invention.

[0044] Figure 13 is an example diagram of an opinion network having content or directionality implemented through a period-based extraction step in a preprocessing step according to the third embodiment of the present invention.

[0045] A method for implementing an opinion network having content or directionality according to the present invention comprises: a data collection step of collecting data containing content regarding a subject of interest; an edge list creation step of coding the collected data according to a predetermined protocol to create an edge list regarding the subject of interest; and a relationship visualization step of visually displaying a relationship between the subject of interest and other objects based on the created edge list; wherein the edge list creation step comprises: an opinion existence determination step of determining whether an opinion regarding the subject of interest exists in the collected data; an opinion content determination step of determining whether the content of the opinion is positive or negative if the opinion exists; and an opinion directionality determination step of extracting a subject expressing the opinion and an object receiving the opinion from a sentence to determine the directionality of the opinion; wherein the relationship visualization step is characterized in that the directionality of the opinion is displayed using the direction of an arrow between the subject and the object, and the content of the opinion is displayed by varying the shape or color of the arrow between the subject and the object or by displaying it with a predetermined character near the arrow.

[0046] The present invention has different features, advantages, and strengths from existing ones in that it can be performed by an AI coder (or an equivalent technology, e.g., ChatGPT) using sentences that are easy for ordinary people to understand and commonly used, rather than existing formulas, etc., compared to other patents or inventions.

[0047] Hereinafter, a method for implementing an opinion network having content or direction according to the present invention will be described in more detail with reference to an embodiment illustrated in the drawings.

[0048] FIG. 3 is a flowchart of a method for implementing an opinion network having content or directionality according to a first embodiment of the present invention, FIG. 4 is an example diagram of an edge list creation item according to an embodiment of the present invention, FIG. 4 is a flowchart of an edge list creation step according to an embodiment of the present invention, FIG. 6 is an example diagram of edge list input and relationship visualization according to an embodiment of the present invention, FIG. 7 is an example diagram of inputting an example from a newspaper into an edge list and visualizing the relationship according to an embodiment of the present invention, FIG. 8 is an example diagram of implementing an opinion network having content or directionality according to an embodiment of the present invention using Netminer as a means of relationship visualization, and FIG. 9 is an example diagram of relationship visualization according to an embodiment of the present invention that reflects the strength of opinion.

[0049] Referring to FIG. 3, a method for implementing an opinion network having content or direction according to the first embodiment of the present invention includes a data collection step (S10), an edge list creation step (S30), and a relationship visualization step (S50).

[0050] The above data collection step (S10) is a step for collecting data containing information on the subject of interest.

[0051] Here, the subject of interest can be a person, issue, or institution, and the form of data collected can include various documents, articles, texts, etc.

[0052]

[0053] The above edge list creation step (S30) is a step of creating an edge list for the subject of interest by coding the data collected in the above data collection step (S10) according to a set protocol.

[0054] That is, the edge list creation step (S30) is a step of creating the edge list by creating a list based on sentences with opinions in the data collected in the data collection step (S10) and entering a set of writing items (cells) into each of the created lists.

[0055] Meanwhile, the edge list creation step (S30) according to one embodiment of the present invention may be configured to determine the subject, object, content of the opinion, or intensity of the opinion included in the edge list creation items, based on the paragraph including the sentence, if the subject, object, content of the opinion, or intensity of the opinion is not determined within the sentence.

[0056] The above relationship visualization step (S50) is a step that visually displays the relationship between the object of interest and other objects based on the edge list created in the above edge list creation step (S30).

[0057] In one embodiment of the present invention, the relationship visualization step (S50) displays the direction of the opinion using the direction of the arrow between the subject and the object, as in the example diagram of FIG. 2, and the content of the opinion can be displayed by changing the shape or color of the arrow between the subject and the object, or by displaying it with a specified character near the arrow, and it is also possible to visualize the subject and the object by inserting them as a picture or photograph.

[0058] In addition, in one embodiment of the present invention, the relationship visualization step (S50) can be realized by uploading the edge list data created in the edge list creation step (S30) to Netminer and then running it.

[0059] The steps for uploading data to Netminer can be broadly divided into the following three steps.

[0060] 1) Input of Main Node Set

[0061] 2) Input of edge list written in 1-mode network

[0062] 3) Data analysis and visualization

[0063] However, the core of the present invention is the creation of an edge list for inputting into the above netminer, and the above netminer itself is a known technology and is not the core of the present invention, so a detailed description thereof will be omitted.

[0064]

[0065] [A brief explanation of the items in the Edge List]

[0066] According to one embodiment of the present invention, the items to be written in the edge list may include, as shown in FIG. 4, 1. document number (Document), 2. sentence number in the document (ID-in Document), 3. subject (Senders), 4. object (Receivers), 5. content of opinion (Content), 6. strength of opinion (Level), 7. time (Time), and 8. whether it is a topic sentence and picture, etc. (Topic). A brief description of each written item is as follows.

[0067] 1. Document: Document Number

[0068] 2. Sentence number in the document (ID-in Document): Opinion sentences in editorials were coded with a number in chronological order.

[0069] 3. Senders: The starting point (person, issue, institution, etc.) expressing their opinion.

[0070] 4. Receivers: The destinations (people, issues, organizations, etc.) that receive opinions (Code nodes representing a sender's expressed opinion (coded with node name following the list of sample nodes).

[0071] 5. Content: Positive or negative opinion.

[0072] 6. Level of opinion: Magnitude of opinion; coded as a number -3, -2, -1, 1, 2, 3, etc.

[0073] 7. Time: Document's publication day (Month, Day, Year, etc.)

[0074] 8. Topic: Whether a sentence was a topic (coded as 1) or not (coded as 0)

[0075]

[0076] The key points in the creation items of the edge list according to one embodiment of the present invention are 3. Subject (Senders), 4. Object (Receivers), 5. Content of the opinion, and 6. Level of the opinion.

[0077] Among the above edge list creation items, 3. Subject (Senders) and 4. Object (Receivers) indicate directionality in the existing network, and 5. Content of opinion and 6. Level of opinion allow us to see the relationship between positive and negative and its level of intensity.

[0078]

[0079] [Precautions before coding]

[0080] The above edge list creation step (S30) may be performed manually by multiple coders, or may be performed by an AI coder using artificial intelligence technology (or an equivalent technology, e.g., ChatGPT, etc.).

[0081] The following are precautions before coding in the above edge list creation step (S30).

[0082] 1) Because the protocol is complex, coders must repeatedly familiarize themselves with the protocol's core content before beginning each coding session to maximize consistency. Meanwhile, if coding is performed by an AI coder, it is advisable to regularly train and learn the protocol.

[0083] 2) Before coding, read it first, and then read it a second time, understanding the overall context of the text before coding, so that you read it more than twice and make a judgment rather than making a judgment in one go.

[0084] 3) If ambiguous sentences are identified during coding, they are marked in a specific color (e.g., red) and discussed later. If both parties' coding is identical, the data can be accepted and used for analysis. However, if the coding differs, the final decision on whether to include or exclude the sentence in the analysis is made through discussion and agreement between the researcher and the coder. In other words, the decision is made through sufficient logical discussion and consensus, regardless of one party's authority or expertise. If agreement is difficult, the opinion of a third-party coder (or researcher, etc.) can be followed.

[0085] 4) You can use the protocol here or create additional protocols based on it by agreeing with the coder and researcher.

[0086]

[0087] [Basic Principles]

[0088] The edge list creation step (S30) according to one embodiment of the present invention includes an opinion existence determination step (S31), an opinion content determination step (S32), and an opinion direction determination step (S33), as illustrated in FIG. 5.

[0089] The above opinion existence judgment step (S31) is the first step, which is a step to determine whether an opinion on the subject of interest exists in the data collected in the data collection step (S10).

[0090] The above opinion content judgment step (S32) is the second step, and if it is determined in the opinion existence judgment step (S31) that an opinion exists, it is the step of judging whether the content of the opinion is positive or negative.

[0091] The above opinion direction judgment step (S33) is a third step in which the subject (Senders) who expresses the opinion and the object (Receivers) who receives the opinion are extracted from the sentence to judge the direction of the opinion.

[0092] Fundamentally, it is not only difficult to determine whether a sentence has an opinion or not, but also difficult to determine who is the subject and who is the object.

[0093] Therefore, in the opinion network according to the present invention, one additional step is added to the existing steps 1 and 2, making a total of three steps.

[0094] The existing steps 1 and 2 are to determine whether the sentence in step 1 has an opinion or not, and step 2 is to determine whether it is positive or negative (Liu 2012).

[0095] However, the present invention goes one step further than the existing sentiment analysis and adds three steps by applying the existing network analysis (Wasserman & Faust, 1994).

[0096] That is, the third step is to extract from complex sentences what are 3. the subject (Senders, the starting point of expressing opinions) and 4. the object (Receivers, the target of receiving opinions) and to show the direction of opinions.

[0097] These steps 1, 2, and 3 become the basic principles of coding the edge list creation step (S30).

[0098] In addition, the following may be included as basic principles of coding in the edge list creation step (S30).

[0099] 1) Basically, we assume and analyze a sentence unit appropriate for determining whether it is positive or negative (Liu 2012). That is, we assume that a single opinion holder expresses one opinion in one sentence and code it as a basic sentence unit (Liu 2012). (If this assumption is not met, we will explain in detail in the sentence situation-specific coding section that follows.)

[0100] 2) Basically, code the subjects (senders, sources, starting point of expressing opinions) and objects (receivers, target of receiving opinions) specified in the text.

[0101] For example, if the 10th sentence of document 5 posted on April 17, 2016 contains the words "I hate it" and "They like it", it is coded in the edge list format as shown in Figure 6 (a).

[0102] And, based on the edge list of Fig. 6 (a), an opinion network having content or directionality through the relationship visualization step (S50) can be implemented as in Fig. 6 (b).

[0103] 3) Network terminology: Nodes are basically subjects (senders, sources) and objects (receivers, targets).

[0104] A node, a term used in networks, is also called an object, a point, an agent, or an actor. It can represent any object, person, thing, or even a specific subject. When there's directionality between nodes, they're divided into two categories: senders (sources) and receivers (targets). The subject is the entity that performs an action or exhibits a state, while the object is the recipient of the subject's action or state.

[0105] 4) The subject of coding is a negative or positive opinion, excluding neutrality. The object expressing the opinion is called the subject (also known as senders, sources, or starting one), and the object receiving the opinion is called the object (also known as receivers, target, or ending one).

[0106] To help with understanding, examples of newspapers related to the Sewol Ferry incident (Hankyoreh, Chosun Ilbo, and other newspapers) will be provided below.

[0107] Fig. 7 (a) is an example diagram of inputting an example from a newspaper into an edge list according to an embodiment of the present invention, and Figs. 7 (b)(c) are example diagrams of implementing an opinion network having content or directionality through a relationship visualization step (S50) based on the edge list of Fig. 7 (a).

[0108] Here, Fig. 7 (b) is an example diagram of an opinion network in which the subject and object are each independent and each has a direction and expresses a negative or positive opinion, and Fig. 7 (c) is an example diagram of an opinion network in which Fig. 7 (b) is displayed as one.

[0109] When the above results are accumulated, it becomes possible to implement an opinion network with content or direction using Netminer as a means of relationship visualization, as shown in Figure 8.

[0110] Each point in Fig. 8 may display the name or photo of the subject or object indicated at the point in Fig. 7 (b) or Fig. 7 (c), and a photo may be used instead of the point.

[0111]

[0112] [Situational Coding of Sentences]

[0113] As mentioned above, the analysis is conducted assuming a sentence unit (Liu 2012) that is basically appropriate for determining whether it is positive or negative. However, since sentences basically include various types of sentences such as single sentences, compound sentences, complex sentences, and compound-complex sentences, the following will examine in detail the coding of sentences according to their situations.

[0114] In one embodiment of the present invention, the edge list creation step (S30) is characterized in that a list is created based on a sentence having an opinion, and if the sentence is composed of a complex sentence, a compound sentence, or a multiple sentence, the sentence is divided into multiple short sentences and the list is created by repeating the process as many times as the number of short sentences, if there are multiple opinions in one sentence, the list is created by repeating the process as many times as the number of opinions, and if there are multiple subjects or objects in one sentence, the list is created by repeating the process as many times as the product of the subjects and the objects.

[0115]

[0116] A. [Between Sentences]: Repeated measurements of complex sentences, compound sentences, and compound-complex sentences.

[0117] 1. Complex sentences, compound sentences, and compound-complex sentences are basically measured repeatedly.

[0118] 1) Basically, if the subject and object appear repeatedly in a sentence connected by coordinating conjunctions (and, but, or, so, because, etc.) and in correlative and subordinating conjunctions, repetition is measured.

[0119] Ex) 2) Hong Moon-jong, Secretary General of the Saenuri Party, criticized the opposition party on the 2nd, saying, "It is truly a cowardly political offensive to fan criticism and anger against the 3)4) President and 5)6) the government ahead of the local elections."

[0120] * Additional explanation: A sentence spoken by a subject other than the newspaper company codes both the object outside and the target object inside.

[0121] 31.02 Saenuri Party - Representative Hong Moon-jong Opposition party negative - 131.03 Opposition party President Park Geun-hye negative - 131.04 Opposition party President Park Geun-hye negative - 131.05 Opposition party government negative - 131.06 Opposition party government negative - 1

[0122] B. [Within a sentence]: Repetitive measure of phrases and clauses within a sentence 1. If two or more opinions or sentiments appear in a sentence, they are coded using repeated measures. That is, in the case of complex or compound sentences, if there are multiple senders and receivers, all of them are coded and entered. This is similar to the separate coding of multiple subjects and objects in a sentence, which will be described later.

[0123] ex) If the subject (or object) is accompanied by another person or subject (or object), the subject is coded separately.

[0124] ex) 7) It is said that on the 16th, the day of the accident, Rep. Lee Yoon-seok of the New Politics Alliance for Democracy, 8) was on a Coast Guard patrol vessel with three aides touring the accident area, and was met with strong protests from the families of the missing.

[0125] 6.07 General Victims' Families: Lee Yoon-seok (negative) - 16.08 General Victims' Families: Lee Yoon-seok (3 aides) (negative) - 1

[0126] * Additional explanation: In this case, if the general victim criticized Rep. Lee Yoon-seok and his aides, they are coded separately. 2. If multiple subjects (or objects) appear in one sentence, each subject (or object) is coded and displayed as a separate sentence. That is, if the object (Receivers) is singular and the subject is plural, repeat measurements are performed as many times as the number of subjects. If the subject (Senders) is singular and the object is plural, repeat measurements are performed as many times as the number of objects. ex) Former Rep. Roh Hoe-chan tweeted on the 17th, "Unless you plan to carry an oxygen tank and engage in rescue activities, 7) politicians and 8) candidates should refrain from visiting the site.

[0127] * Additional explanation: If you repeatedly criticize one target in one sentence, code it as 2, 3, etc.

[0128] 1. Document 2. Sentence number in document (ID-in Document) 3. Subject (Senders) 4. Object (Receivers) 5. Content of opinion (Content) 6. Level of opinion (Level) 7. Time (Time) 8. Topic 117 Justice Party - Former Assemblyman Roh Hoe-chan Politician Infidelity - 12014.04.190 118 Justice Party - Former Assemblyman Roh Hoe-chan Candidate Infidelity - 12014.04.190

[0129] 3. Words that appear to be one word but have two meanings are coded as having separate meanings. Ex) Search and seizure

[0130] * Additional explanation: Since “seizure” and “search” have separate meanings, the subject codes them as two actions to the object.

[0131]

[0132] C. [Within a paragraph] Inference

[0133] 1. If the main sentences in a paragraph express positive or negative opinions for objective and repetitive reasons (e.g., laws, etc.), repetitively code them as positive or negative based on the relevant sentences.

[0134] ex) The special investigation teams of the Incheon District Prosecutors' Office and the Busan District Prosecutors' Office are pursuing structural corruption in the 32)33) Sewol Ferry supervisory agency, 34)35) and the shipping industry. They are reviewing six charges against the Yoo Byung-eon family, the actual owners of Chonghaejin Marine, 36)37)38)39)40)41)41-1)41-2)41-3)41-4)41-5)41-6) including embezzlement, breach of trust, tax evasion, overseas asset flight, coercion, and bribery.

[0135] 2. Even if the subject or object mentioned in the preceding sentence is omitted, if the omitted subject or object can be inferred from the context “within one paragraph,” it can be coded as a positive-negative sentence with a subject or object.

[0136] ex) At 9:45 AM on the 16th, the day of the accident, 15) The Central Disaster and Safety Countermeasures Headquarters, headed by Minister of Public Administration and Security Kang Byung-kyu, was activated, but criticism continued that they were not even able to count the numbers properly. 17) Even after a full day had passed since the accident, the number of passengers and missing people continued to change.

[0137]

[0138] * Additional explanation: It is coded that Minister of Public Administration and Security Kang Byung-kyu and the Central Disaster and Safety Countermeasures Headquarters were repeatedly criticized behind the scenes.

[0139]

[0140] [Detailed description of the items in the Edge List]

[0141] A. Document

[0142] The document is assigned a serial number. The scope and method of setting the document can be freely determined by the coder.

[0143]

[0144] B. Sentence number in document (ID-in Document)

[0145] 1. Enter the order in which sentences containing opinions or sentiments appear in the document. Since this is not a core analysis target for the opinion network, the ordering is not critical. However, if order is important for analysis purposes or for coding convenience, referencing the method below can make coding easier. The specific order is as follows.

[0146] 2. Recommended method for entering and displaying sentence numbers in the Edge List and documents

[0147] Although not essential to coding, it is helpful to number sentences to make them easier to distinguish and look up later.

[0148] 1) Code sentences that are considered opinion-based by assigning numbers to them. While this may lower the level of agreement between coders if strict adherence is maintained, it's not a key factor affecting the quality of the opinion network data, so it's irrelevant. Therefore, when "if an ambiguous sentence is discovered during coding, mark it in red or a preferred color and discuss it later," it's a good idea to include numbers like 1-1 and 1-2 if a sentence is missing from the original.

[0149] 2) A sentence with a period is always considered as one sentence, and the opinion content in the detailed sentences within it (complex sentences, compound sentences, compound-complex sentences, phrases and clauses within the sentence, etc.) can be indicated as 1-1, 1-2... (However, this is difficult because it numbers sentences that do not need to be coded instead).

[0150] 3) Numbers can be placed where there are senders and receivers rather than at the beginning of the sentence, or if there is no subject, numbers can be placed where there are objects.

[0151] (1) If the object (receiver) is singular and the subject is plural, it is better to enter a number in the subject.

[0152] (2) If the subject (Senders) is singular and the object is plural, it is better to put a number in the object.

[0153] ex) Former lawmaker Roh Hoe-chan pointed out on Twitter on the 17th, "Unless they plan to carry out rescue operations while carrying oxygen tanks, 7) politicians and 8) candidates should refrain from visiting the site."

[0154] 1. Document 2. Sentence number in document (ID-in Document) 3. Subject (Senders) 4. Object (Receivers) 5. Content (Content) 6. Level (Level) 7. Time (Time) 8. Topic 117 Justice Party - Former Assemblyman Roh Hoe-chan Politician Infidelity - 12014.04.190 118 Justice Party - Former Assemblyman Roh Hoe-chan Candidate Infidelity - 12014.04.190

[0155] (3) If both the subject and the object are plural, put a number equal to the product of the subject and the object (subject Υ object) at the beginning of the sentence (or as agreed upon by the coder). ex) The National Assembly's investigation into the Sewol Ferry disaster is suffering from difficulties without even getting off to a proper start due to 1)2) differences of opinion between the ruling and opposition parties. Opinions are running parallel over whether to specify 3)4) witnesses and 5)6) reference persons in the investigation plan, or 7)8) to first adopt the plan and decide on the witness issue later (Hankyoreh 73).

[0156] 73.03 Saenuri Party Opposition Negative-173.04 Saenuri Party Opposition Negative-173.05 Saenuri Party Opposition Negative-173.06 Saenuri Party Opposition Negative-173.07 Saenuri Party Opposition Negative-173.08 Saenuri Party Opposition Negative-1

[0157] * Additional explanation: 1)2) shows that the subject and object of the ruling and opposition parties are in conflict over issues 3)4), 5)6), 7)8) (the ruling party was the Saenuri Party at the time). C. Node

[0158] Meaning of Node

[0159] - Nodes are expressed as subjects (Senders, the starting point of expressing opinions), objects (Receivers, the target of opinions) (Wasserman & Faust, 1994), tail / head (Butts, 2008), or sometimes Beginning and End (Liu 2012). In this protocol, the terms subject and object are used for ease of understanding.

[0160]

[0161] Node Basics

[0162] 1. Nodes are essentially used as they appear in the original text. However, node names vary depending on the sentence, context, and time. Therefore, for consistency, it may be helpful for coders to create a thesaurus of common node names appropriate for the document, making future analysis easier. (The following is an example, so it's less mandatory than the "basic principles" and "sentence-specific coding" outlined above, and is for reference only.)

[0163] - A thesaurus like this can be created by inputting a thesaurus into a network program (e.g. NetMiner, UCInet, etc.), or if that is not possible, the coder can input each node according to the thesaurus while looking at the thesaurus during input.

[0164] - That is, the edge list creation step (S30) according to one embodiment of the present invention is characterized by assigning a representative name that can represent a name or expression having identity, and creating the edge list based on the assigned representative name.

[0165] ex) Different expressions such as 'Sewol Ferry Incident' and 'April Ferry Accident' are unified as 'Sewol Ferry Accident'

[0166] 2. The subject and object are basically people or institutions, but if there is no specific subject, the person who wrote the article (newspaper, author of the article (newspaper reporter, author written in the document, etc.)) is considered. If there is no author of the article, the company that published the article (newspaper, etc.) is considered. However, in the case of social media such as Instagram and Facebook, the subject is the person who posted the article, not the social media, because there is no author of the article.

[0167] 3. If it is not in the given category, find the most similar category in the existing category and enter it. However, if it is not in the existing category, a new node is created.

[0168] ex) Navigator, helmsman vs. existing classification - Sewol ferry crew

[0169] In this situation, the final input can be "Sewol Ferry Crew - Navigator, Sewol Ferry Crew - Helmsman."

[0170] 4. Specific terms and technical terms for each coding issue can be agreed upon between the researcher and the coder.

[0171] ex) 1st agreement on the Sewol Ferry Special Act - 8.7

[0172] Second Agreement on the Sewol Ferry Special Act - August 19

[0173] Third Agreement on the Sewol Ferry Special Act - September 30

[0174]

[0175] D. Coding of Subjects (Senders, the starting point for expressing opinions) and Objects (Receivers, the target of opinions)

[0176] 1. When a sub-department is mentioned for each coding issue, the name of the parent department can be listed alongside the coding. This allows for grouping of numerous nodes to some extent.

[0177] ex) Ministry of Public Safety and Security - Central Disaster and Safety Countermeasures Headquarters

[0178] * Additional explanation: Instead, when analyzing later, you may want to divide this into groups and analyze it in the Node List.

[0179] 2. Individuals in different positions within an organization can be entered as different nodes, specifying their position, such as their organization. If the position changes, enter the new position (as described in the document) by coding it.

[0180] ex) Democratic Party - National Assemblyman Hong Gil-dong, Democratic Party - Party Leader Hong Gil-dong

[0181] Saenuri Party - Hwang Woo-yea, Co-Chairperson of the Saenuri Party Election Strategy Committee", "Saenuri Party - Hwang Woo-yea, Saenuri Party Leader"

[0182] * Additional explanation: You can do it differently or, since they are the same person, you can do it as “Saenuri Party - Representative Hwang Woo-yea.”

[0183] 3. If there is any confusion, the abbreviation can be entered as the full name.

[0184] ex) The ruling and opposition parties

[0185] * Additional explanation: “Ruling and opposition parties” are coded separately as “ruling party” and “opposition party.”

[0186] 4. In cases where a subject is not specified but can be seen as a vague person, a general term such as “citizen” can be used as a superordinate concept to cover it.

[0187] ex) Rumors and slander

[0188] * Additional explanation: Rumors and slander are coded as “Citizen - Rumor” and “Citizen - Slander.”

[0189] 5. Involve ordinary people in coding too.

[0190] ex) Citizen - Hong Gil-dong

[0191] 6. When referring to a vague object (person or object), especially an inanimate object or subject, specify it as a node (this can be ignored depending on the purpose of data collection and research).

[0192] ex) It appears that from the time of departure to the accident, 1) negligence, 2) moderation, and 3) evasion of responsibility all overlapped, leading to the disaster.

[0193] The culprits were 1) moderate and 2) irresponsible.

[0194] - However, if the context explicitly states that an inanimate object belongs to a node, the subject to which it belongs is considered a node and inserted. Even if the object being referred to is ambiguous, if the context clearly states it, it is inserted.

[0195] ex) National disaster management system, government command and control function

[0196] * Additional explanation: The country or government becomes the subject or object.

[0197] ex) Wrong practices in our society

[0198] * Additional explanation: Coding as a critique of our society.

[0199] 7. [Guessing when the node is unclear] If the subject is clearly stated and can be inferred from the context (within a paragraph, by looking at the title alone), code it as a clear subject or object (or input positive or negative) (refer to the preprocessing step (S20) described later).

[0200] 8. Social and political issues or their targets are considered as subjects and objects and are coded as subjects and objects.

[0201] ex) diving bell, Sewol Special Act, agreement between the ruling and opposition parties, Sewol, Jindo, Ansan City, Sewol Ferry disaster, Sewol, pro-Park, pro-Roh, local elections, economy, Gyeonggi (meaning economy...), economic industry (accommodation, travel, etc. - refer to Chosun Editorial No. 13, 48, Hankyoreh No. 44), Sewol Ferry National Assembly investigation, Sewol Special Act, protest (candlelight protest...), hearing, salvage of Sewol Ferry, corporate responsibility law, yellow ribbon, political situation, state affairs, ..., politics, tripartite consultative body, Sewol Special Act Enforcement Decree, rally, amendment to Sewol Special Act Enforcement Decree, Sewol Special Act Enforcement Decree, special law, agreement between the ruling and opposition parties, giving the Sewol Ferry Truth Investigation Committee the right to investigate and prosecute

[0202] 9. For other detailed examples of subjects and objects, please refer to [Appendix 2] described below.

[0203]

[0204] E. Content: Negative & Positive Opinion

[0205] The meaning of opinion

[0206] - The meaning of opinion can be seen as opinion in Korean, but it encompasses a broader meaning. Opinion means "a positive or negative sentiment, view, attitude, emotion, or appraisal about an entity or an aspect of the entity from an opinion holder" (Liu, 2011).

[0207]

[0208] Positive and Negative Opinion Basics

[0209] 1. Basically, opinion in a sentence encompasses a positive or negative sentiment, view, attitude, emotion, or appraisal.

[0210] 2. Basically, determine the author's intention at the time, whether positive or negative.

[0211] 3. If the subject or object is clearly stated and the positive or negative opinion can be inferred from the context (within a paragraph), code the subject or object (or positive or negative).

[0212] 4. If the target of criticism or affirmation is not specified but can be guessed and identified within the paragraph, it is coded as subject or object (or affirmation or negation).

[0213] 5. Even if the content is factual, the author's subjective opinion can be judged by the positive or negative tone of the adjectives and adverbs.

[0214] ex) Prime Minister Chung Hong-won made a listless address to the nation, urging only the normalization of the National Assembly without even mentioning the issue of enacting the Sewol Ferry Special Act.

[0215] ex) 12) Saenuri Party's 13) Joo Ho-young, the policy committee chairman, disparaged the Sewol ferry disaster as "basically a traffic accident," but

[0216] * Additional explanation: If the content is objective and factual, but contains the author's subjective opinion and has a clearly negative adjective or adverbial tone, it is coded as negative.

[0217] ex) I can't help but wonder who they are trying to show off by making such a fuss.

[0218] 6. The title or picture is the subject or object (or positive or negative) of the title or picture.

[0219]

[0220] E-1. Negative

[0221] 1. When a subject expresses a negative opinion toward an object, it is coded as negation.

[0222] 2. The general criteria for judging negativity are as follows (Liu, 2012, p. 55-56).

[0223] - less or decreased positive (Liu 2012 p. 55))

[0224] - more or increase negative (Liu 2012 p. 55)

[0225] - no low, less or decreased quantity of positive potential items (PPI) (Liu 2012 p. 55)

[0226] - Large, larger or increased quantity of negative potential items (NPI) (Liu 2012 p. 55)

[0227] - Undesirable fact (Liu 2012 p. 56)

[0228] - deviate from the desired value range (Liu 2012 p. 56)

[0229] - Produce no, little or less resource (Liu 2012 p. 56)

[0230] - Produce some or more waste (Liu 2012 p. 56)

[0231] - Consume a large quantity of more resources (Liu 2012 p. 56)

[0232] - Consume no, little or less waste (Liu 2012 p. 56)

[0233] 3. Claims of duty and responsibility, such as "must", are coded as accusatory, that is, negative. For example, "-must" is coded as negative, as in "-didn't" or "-didn't", or as suggesting responsibility and alternatives to a specific target.

[0234] ex) 5)6) The Ministry of Oceans and Fisheries' ship licensing and management, 7)8) The Coast Guard's ship emergency training inspection and initial disaster response should be investigated.

[0235] * Additional explanation: - It is coded as negative because it must be done.

[0236] ex) What is the government's raison d'être? It goes without saying that it is to protect the "lives and safety of the people."

[0237] * Additional explanation: Code as negative because it claims responsibility.

[0238]

[0239] Other situations

[0240] 1. The subject's use of the object is coded as negative.

[0241] ex) Some leftist forces are currently trying to use the22) sorrow and 22.1) anger of the families of the missing Sewol ferry victims as a fuse for political struggle, just like during the mad cow disease incident in 2008 (Chosun 38).

[0242] ex) However, 16) the opposition party must be aware that if it tries to use the Sewol tragedy for the election, it may face a backlash (Chosun 42).

[0243] 2. Using metaphors to criticize is coded as negative.

[0244] ex) 28) ‘Surely not, disease’ is a chronic disease that brings tragedy to our society.

[0245] * Additional explanation: Coded as criticizing our society.

[0246] 3. Indirect expressions (such as sarcasm) are coded as negative, as they are considered criticism.

[0247] ex) It would be a blessing if even just one of the 46 lifeboats on the Sewol opened.

[0248] * Additional explanation: The expression "I am so grateful" is used as an indirect criticism, as it is a sarcastic expression. However, if "I am so grateful" is not present, it can be considered a general fact.

[0249] ex) Chief of Staff Kim Jang-soo's remarks clearly show the current perception and primary focus of the Blue House. In short, he is saying that he does not want to be hastily involved in this incident and have the issue of responsibility raised. 15.1) Chief of Staff Kim's quick evasion of responsibility is likely not just for his own benefit. It also has the nature of preventing the sparks of responsibility from flying to President Park Geun-hye. His intention is that President Park's only job is to reprimand, nag, and severely punish from her "lofty position," and to avoid getting her toes wet in the mud of responsibility. 21) This demonstrates incredible loyalty and 22) astute political sense.

[0250] * Additional explanation: The last sentence can be inferred as a criticism based on the context within the paragraph.

[0251] 4. Lament is coded as negative in context.

[0252] ex) 24)25) How long must our society repeat these shocking incidents and shocking responses?

[0253] 5. Certain words can be coded as double negation.

[0254] a) Words like "conflict" are double negative

[0255] ex) There was an incident where about 13) civilian divers who volunteered without receiving a single penny withdrew after a conflict with the Coast Guard.

[0256] 14.14 Korea Coast Guard Sewol Rescue - Civilian Diver negative-114.15 Korea Coast Guard Sewol Rescue - Civilian Diver negative-1

[0257] 6. It is also possible for the subject and object to be the same. Ex 1) When the subject brings about an unjust and bad result for the subject himself.

[0258] ex) 31) If the opposition party repeats the behavior that makes the people feel disgusted this time,31.1) it may end up in a situation of self-destruction (Chosun 126).

[0259]

[0260] E-2. Positive

[0261] 1. When a subject expresses a positive opinion to an object, it is coded as positive.

[0262] 2. The general criteria for judging positivity are as follows (Liu, 2012, p. 55).

[0263] - less or decreased negative (Liu 2012 p. 55)

[0264] - more or increased positive (Liu 2012 p. 55)

[0265] - No low, less or decreased quantity of NPI (Liu 2012 p. 55)

[0266] - A large, larger or increased quantity of PPI (Liu 2012 p. 55)

[0267] - Desirable fact (Liu 2012 p. 56)

[0268] - within the desired value range (Liu 2012 p. 56)

[0269] - Produce a large quantity of or more resources (Liu 2012 p. 56)

[0270] - Produce no, little or less waste (Liu 2012 p. 56)

[0271] - Consume no, little or less resources (Liu 2012 p. 56)

[0272] - Consume a large quantity of or more waste (Liu 2012 p. 56)

[0273]

[0274] Other situations

[0275] 1. Mourning, feeling sad for someone, is coded as a positive emotion.

[0276] 2. Code apologizing to someone as positive.

[0277] ex) Public apology

[0278] * Additional explanation: Coded as positive for the public.

[0279] 3. Encouraging someone is coded as positive.

[0280] ex) [Editorial] 1) We need to support the Ansan community so that it can overcome the shock and rise again.

[0281] 4. Empathy is coded as positive.

[0282] ex) When I think that even at this very moment, children might be crying out for help in the cold water,1) my heart breaks.

[0283] 5. The Sewol Ferry victims' memorial service is positive.

[0284] ex) 39) Hundreds of middle and high school students held a candlelight vigil in Seoul's Cheonggye Plaza to commemorate the victims of the Sewol ferry disaster (Hankyoreh 34).

[0285] 6. Corruption is basically seen as having a positive effect on the relevant organization.

[0286] ex) Private organizations have become prey to retired civil servants (Chosun 62).

[0287] 7. Certain words can be coded as bilaterally positive.

[0288] a) Words like "black connection" are coded as bilateral positive.

[0289] ex) Deep-rooted dark connections between the Ministry of Oceans and Fisheries and the shipping industry (33. Munhwa Ilbo)

[0290] 33.07 Ministry of Oceans and Fisheries Shipping Sector - Shipping Industry Positive-133.08 Shipping Sector - Shipping Industry Ministry of Oceans and Fisheries Positive-1

[0291] b) Words related to "collusion" are coded as positive for both parties. For example, it is necessary to investigate why a major shareholder of a company that went bankrupt in 1997 received such special treatment, and whether there was any collusion or corruption with public officials during that time. (Dong-A Ilbo, 29)

[0292] 27.18 Civil servant Cheonghaejin Shipping - Major shareholder positive1 27.19 Cheonghaejin Shipping - Major shareholder civil servant positive1

[0293] E-3. Neutral

[0294] It is not coded because it does not fit the nature of the opinion network, but researchers or coders can code it when necessary for analysis and research. - It is possible to input level 0 at this time, but it is difficult to input the direction because it is simply a simultaneous occurrence.

[0295] 1. Objective sentences do not express any emotions or opinions, so they are considered neutral sentences and are not coded.

[0296] ex) According to the transcript, the first communication was made at 9:07 AM. This was 12 minutes after 8:55 AM, when the Sewol notified the Jeju Control Center under the Ministry of Oceans and Fisheries that "the ship is capsizing." Who neglected those 12 minutes? Jindo VTS's first communication was an urgent question: "Are you sinking?" The Sewol immediately replied, "Sinking."

[0297] 2. Simple facts are considered neutral sentences and are not coded.

[0298] ex) At the beginning of the accident, the divers searched in the strong current and then returned to work without being able to rest properly on the rocking ship, (Hankyoreh 17)

[0299] 3. Do not code them as simple general principles, propositions, or other neutral sentences.

[0300] ex) "The policy decisions and performance of duties by the government or state agencies must always be subject to the public's scrutiny and criticism, and this is possible only when the freedom of the press, whose main task is scrutiny and criticism, is sufficiently guaranteed." This is a principle shared not only by our Supreme Court but also by modern democratic nations.

[0301]

[0302] Other situations

[0303] 1. Opinions disguised as simple propositions are inserted by understanding the context and determining whether they are positive or negative.

[0304] ex) The tripartite consultative body is a negotiating body, not a decision-making body. (Hankyoreh 171)

[0305] 2. If there is no obligation to do something and it is not done at one's discretion, it is considered a statement of fact and is considered a neutral sentence and is not coded.

[0306] ex) The court did not find Captain Lee guilty of murder, saying there was insufficient evidence to show that he escaped alone because he thought it was okay for the passengers to die.

[0307] [Note] Something that should have been done but was not done (there is a duty to do something, i.e. something that should have been done but was not done) is considered a criticism and is coded as negative. For more information, refer to "3. Asserting duty and responsibility such as "must"" in A. Negative.

[0308] 3. In context (within a paragraph), vague criticism without a target is not indicated.

[0309] ex) We must thoroughly investigate whether the crew took appropriate evacuation measures, including a thorough post-accident response, and hold them strictly accountable. The situation was as shabby as ever, with a weak foundation.

[0310] 4. Other potentially confusing neutral sentences

[0311] 1) Attendance at rallies and meetings, such as protests, are not coded as neutral statements of simple facts unless there are specific positive or negative adjectives.

[0312] ex) The fact that even teenagers and mothers with young children are participating in the protests shows that the situation is not simple (Hankyoreh 34).

[0313] * Additional Note: Attending a protest can be an expression of discontent or a show of support, so it must be viewed in the context of the overall situation.

[0314] 2) General announcements, surveys, etc. (public statements, opinion polls, election results, etc.) are considered neutral sentences that are neither positive nor negative and are not coded.

[0315] 3) Appointment and dismissal are considered as mere facts unless the appointing authority is specified, unless it is the person who has resigned (which he denies to himself).

[0316] 5. Related sentences to note

[0317] 1) It is a simple fact, but if the subject clearly expresses affirmation or negation to the object, it is coded as positive or negative. [Reference] See E-2. Positive, E-1. Negative.

[0318] ex) The National Tax Service (9), 10), 11) and the Financial Supervisory Service (12), 13), and 14) have launched an investigation into the Yoo family for suspicions of overseas asset flight, offshore tax evasion, and illegal foreign exchange transactions. (Dong-A Ilbo, 29)

[0319] * Additional explanation: Arrests, indictments, investigations, etc. are objective facts, but they are clearly considered negative actions against the other party, so they are coded as negative.

[0320] 2) Even if it is a simple fact, if the author's subjective opinion is included and an error is pointed out, it is coded as negative.

[0321] ex) The Central Disaster and Safety Countermeasures Headquarters of the Ministry of Public Safety and Security announced at approximately 11:20 AM yesterday that "divers have successfully entered the hull and are currently injecting air." However, after the Korea Coast Guard denied this, they changed their statement four hours later, saying on November 5) that "entry failed."

[0322]

[0323] F. Level of Opinion

[0324] In one embodiment of the present invention, the strength of an opinion is expressed as a number according to a set standard, and if the content of the opinion is positive, it is expressed as +, and if it is negative, it is expressed as -.

[0325]

[0326] F-1. General sentences

[0327] 1. Basically, the opinion strength of each sentence is set to an absolute value of 1. That is, a positive opinion is coded as +1, and a negative opinion is coded as -1 (Liu 2012).

[0328]

[0329] F-2. Title and pictures of the article, etc.

[0330] 1. Pictures and titles are measured by multiplying them by three since they are relatively easy for people to see. Positive comments are coded as +3, and negative comments are coded as -3. (It can be assumed that the influence (impact) of what the photo (or picture) in the document depicts is the same as the title. However, the exact intensity may change depending on the purpose of the coder or researcher's research and analysis.) And descriptions of pictures and the like are considered to have a relatively strong influence, so they are coded by two, that is, positive comments are coded as +2, and negative comments are coded as -2. In addition, the presence or absence of a title and picture can be coded separately in other coding blanks.

[0331]

[0332] F-3. Visualization of the strength of opinions

[0333] The relationship visualization step (S50) calculates the sum, maximum, minimum or average of the strength of opinions expressed in numbers and can display them by differentiating the shape or color of arrows between the subject and object, or by displaying them with a set character near the arrow. It is also possible to visualize the subject and object by inserting them into a drawing or photograph.

[0334] Figure 9 illustrates an example of relationship visualization according to one embodiment of the present invention that reflects the strength of opinions.

[0335]

[0336] G. Time

[0337] 1. Enter the date and time the document was published or posted. Since the time varies for each document, enter the posting time in as much detail as possible. Depending on your analysis objectives, you can use this information to create time series, panels, or other data for further analysis. However, this deviates from the original purpose of the opinion network, so a detailed explanation will be omitted.

[0338] 2. Examples of input methods are as follows, and the input method can be determined by the coder or researcher.

[0339] ex) 2014.08.22 1:30 or 20140822, 13:30, etc.

[0340]

[0341] H. Topic: Topic sentence and other optional additions

[0342] 1. You can enter other necessary information. For example, you can enter a topic and whether or not to include an image, allowing for further analysis of the topic sentence or image later. However, this deviates from the original purpose of the opinion network, so a detailed explanation will be omitted.

[0343] ex) For example, if it is a topic sentence, it can be coded as 1, otherwise 0, or if it is a topic sentence, it can be coded as 1, if it is a picture, it can be coded as 2, etc. Or, if it is a topic sentence, it can be coded as "topic", if it is a picture, it can be coded as "picture", etc., and the researcher or coder can freely set it.

[0344] 2. Other things that can be added to the edge list

[0345] - The most important part of the opinion network and the part required to directly draw the subject and object picture are 3. Senders (subject, starting point of expressing opinion), 4. Receivers (object, target of opinion), 5. Content (content - positive, negative), 6. Level (intensity of opinion). Therefore, researchers or coders can freely add opinions and other necessary information according to the purpose of research and analysis, other than the cells in the Edge List of the table mentioned above. However, since this deviates from the original purpose of the opinion network, a detailed explanation will be omitted.

[0346] a) Type of document

[0347] - You can code the type of document. For example, newspapers can be coded as 1, social media as 2, etc.

[0348] b) Week, Year

[0349] If the analysis unit is too detailed, the date unit can be entered in the previous time period, and can be entered in weekly (1 week, 2 weeks, 3 weeks, etc.), monthly (January, February, March, etc.), or yearly units. Additional analysis can be performed accordingly. However, this deviates from the original purpose of the opinion network, so a detailed explanation will be omitted.

[0350]

[0351] In the above, the first embodiment of the present invention has been examined, and in the following, a method for implementing an opinion network having content or direction according to the second and third embodiments of the present invention will be examined.

[0352] Looking at Figure 10, the method for implementing an opinion network having content or direction according to the second embodiment of the present invention includes a data collection step (S10), an edge list creation step (S30), an inter-coder reliability measurement step (S40), and a relationship visualization step (S50).

[0353] Except for the inter-coder reliability measurement step (S40), the remaining steps are mostly the same as in the first embodiment described above, so the following description will focus on the inter-coder reliability measurement step (S40).

[0354] In the second embodiment of the present invention, the edge list creation step (S30) is performed by a plurality of coders, and the inter-coder reliability measurement step (S40) is a step of measuring intercoder reliability between coders who performed the edge list creation step (S30).

[0355] In addition to 3. Subject (Senders, the starting point for expressing opinions), 4. Object (Receivers, the target of receiving opinions), which are the most important parts of the opinion network and are necessary for drawing a picture of the subject and object directly, intercoder reliability is measured based on 5. Content of the opinion (positive, negative), and 6. Level of the opinion.

[0356] Since the strength (level) of one opinion in a sentence is set to +1, -1 as the default, there is no difference in the strength (level) of the opinion, but this protocol is necessary for the consistency and reliability of the contents of the remaining 3. Subject (Senders, the starting point of expressing the opinion), 4. Object (Receivers, the target of receiving the opinion), and 5. Content (positive, negative).

[0357] The above intercoder reliability is primarily based on content analysis, using two coders. Therefore, it is desirable to achieve a consensus of over 80%.

[0358] The production formula at that time (Holsti, 1969) was as follows.

[0359] Ole Holsti (1969) formula = 2M / (N1+ N2)

[0360] Here, M = Coincident items coding, N1 = The number of coded items by Coder 1 [the number of items (sentence meanings) coded by Coder 1], N2 = The number of coded items by Coder 2 [the number of items (sentence meanings) coded by Coder 2]

[0361]

[0362] Meanwhile, looking at FIG. 11, the method for implementing an opinion network having content or direction according to the third embodiment of the present invention includes a data collection step (S10), a preprocessing step (S20), an edge list creation step (S30), an inter-coder reliability measurement step (S40), and a relationship visualization step (S50).

[0363] Except for the above preprocessing step (S20), the remaining steps are mostly the same as the first and second embodiments described above, so the following description will focus on the preprocessing step (S20).

[0364] The above preprocessing step (S20) is a step for preprocessing data collected in the data collection step (S10) prior to the edge list creation step (S30), and in one embodiment of the present invention, includes a relevance-based extraction step (S21) and a period-based extraction step (S22).

[0365] The above relevance criteria extraction step (S21) is a step for judging and extracting data that is recognized as being related to a specific event or specific topic determined by the user among the data collected in the above data collection step (S10).

[0366] The above relevance criteria extraction step (S21) may be configured to arrange the plurality of data whose relevance is recognized in chronological order, and at this time, the edge list creation step (S30) may be configured to analyze the plurality of data in chronological order to create the edge list.

[0367] In addition, the relevance criterion extraction step (S21) may be configured to determine the relevance based on the paragraph containing the sentence if the relevance is not determined within the sentence.

[0368] Hereinafter, a specific coding method of the above relevance criteria extraction step (S21) will be described in detail with an example.

[0369] When selecting and coding a specific event, the method for setting the scope of relevance to the specific event is as follows.

[0370] 1) When coding documents from multiple sources, code multiple documents in chronological order so that similar content can be coded identically.

[0371] Ex) The Chosun Ilbo and Hankyoreh newspapers must be coded together in chronological order to allow for similar insertion and deletion of the same events.

[0372] 2) Recognize specific events and their relevance within a paragraph, based on context. That is, if sentences are separated but the subject and object can be inferred contextually, infer them within a paragraph. (Otherwise, the number of possible cases would be too large, making coding difficult.) Code positive, negative, and subject / object relationships based on what can be inferred within a paragraph.

[0373] Ex) Multiple incidents can be mentioned in a single text. In this case, mentioning one Sewol incident in a single paragraph and using pronouns such as "Sewol incident" and "Sewol incident" is acceptable. In other words, unless the Sewol incident is specifically mentioned in the next paragraph, it is considered to be mentioned only in the preceding paragraph.

[0374] * Additional explanation: Otherwise, the relevance to the incident becomes unclear, making it difficult to determine the relevance to the incident.

[0375] 3) Mentioning a specific event and then referring to something after that event is also considered related to that specific event. This is because, if it were not related to the specific event, it would not have been explicitly stated as occurring after the specific event. Therefore, it is coded as directly related.

[0376] ex) This problem was solved after the Sewol ferry disaster 1) the demand for tourism decreased 2) chartered buses were mobilized to relieve the burden (Chosun 118).

[0377] ex) The recent Pope's visit and the Sewol ferry sinking 1) have made us look back on our marginalized neighbors (Hankyoreh 167).

[0378] 4) Targets those directly related to a specific incident.

[0379] ex) As a result of the shock of the Sewol ferry disaster, 1) economic activity shrank, and 2) sales at major retailers decreased by 3-5% in April compared to the previous year. With most school trips and various festivals and events canceled, businesses closely related to the economy, such as 3) travel, 4) lodging, 5) distribution, and 6) restaurants, are experiencing significant difficulties (Chosun 18).

[0380] 5) Excluding those that are not directly related to a specific incident.

[0381] ex) Hongdo residents submitted a petition to the Mokpo Coast Guard after the Sewol ferry disaster, claiming that they could not guarantee the safety of the ship, but the accident occurred just five months after receiving permission to operate (Chosun 167).

[0382] 6) The scope and time of document extraction for a specific incident are determined according to the researcher's research interests.

[0383] 7) Even if the time is before the event, if it is directly related to that specific event, include it.

[0384] ex) Yoo Byung-eun's ship structure change, Sewol Special Act deliberation and resolution & search conclusion, Government Organization Act, mentions related to Yoo Byung-eun Act, Ministry of Public Safety and Security, Sewol-related laws deliberation and resolution in plenary session, disbandment of National Fire Agency and Coast Guard and related matters, establishment of Ministry of Public Safety and Security and related matters, considering the impact of Sewol Ferry incident on local elections, local elections are also included. Sewol-related local elections and superintendent of education election on June 4, by-election on July 30, Sewol Ferry state investigation & Sewol Special Act, declaration of disbandment of Coast Guard, Yoo Byung-eun investigation and Yoo Byung-eun Act, Prime Minister Chung Hong-won's resignation and reinstatement (Prime Minister Chung Hong-won's reinstatement is directly related to Sewol Ferry)

[0385] * Additional explanation: Although this happened before the Sewol ferry disaster, it is coded as it is mentioned as a direct cause of the disaster.

[0386] ex) The resignation of Presidential Secretary for Education and Culture Song Kwang-yong on September 20 is considered to be unrelated to the Sewol incident.

[0387] 8) Any other missing or ambiguous events are included through agreement between coders.

[0388] 9) Anything derived secondary to a specific event (i.e., not directly) is treated as a simple fact and is not coded. Anything that is mentioned as having a direct impact on a specific event is coded.

[0389] Ex) The legal deadline for processing the 2013 budget, worth 349 trillion won, was the 31st of last month, but due to the opposition party's extra-parliamentary protests, this deadline was missed again this year. The state audit of some ministries, originally scheduled for the last week of August, has also been postponed. Currently, over 8,000 bills are pending in the National Assembly. Yet, lawmakers continue to receive monthly allowances exceeding 10 million won (Chosun 145).

[0390] * Additional explanation: The struggles and debates related to the Sewol Ferry Special Act are included in the coding, but secondary matters thereafter are simply considered facts and are not coded.

[0391] Figure 12 illustrates an example of implementing an opinion network having content or directionality through a relevance criteria extraction step (S21) in a preprocessing step (S20) according to the third embodiment of the present invention.

[0392] Looking at Figure 12, it can be seen that an opinion network with grouped display of issues has been implemented in the entire data through the relevance criteria extraction step (S21).

[0393] Meanwhile, each point in Fig. 12 may display the name or photo of a subject or object, but this has been omitted in Fig. 12 for ease of viewing, and a photo may be used instead of a point.

[0394] The above period-based extraction step (S22) is a step of extracting data corresponding to the period range set by the user from among the data collected in the above data collection step (S10).

[0395] Figure 13 illustrates an example diagram of an opinion network having content or directionality implemented through a period-based extraction step in a preprocessing step according to the third embodiment of the present invention.

[0396] Looking at Figure 13, it can be confirmed that through the period-based extraction step (S22), it is possible to analyze who receives positive and negative opinions about a specific node for a specific period.

[0397]

[0398] [Appendix 1]

[0399] Positive and Negative Opinions (some)

[0400] - The case here is just one example, and positive and negative opinions may vary depending on the context.

[0401] Positive, Negative, Hope, Failure, Treatment, Encouragement, Thank You, Inspiration, Responsibility, Righteousness, Mistake, Love, Distrust, Disappointment, Indignation, Union, Baseness, Apology, Mori, Pear, Hope, Illegibility, Cleanliness, Incompetence, Integrity, Anger, Uprightness, Confusion, Courage, Hesitation, Consideration, Slow-wittedness, Response, Cooperation, Disappointment, Respect, Shame, Self-Portrait, Ancient, Guilt, Remorse, Prayer, Abandonment, Comfort, Weak

[0402] [Appendix 2]

[0403] Examples of common node names DATA (partial)

[0404] - The principle is to code exactly what is in the "original text", but since what is referred to in a sentence varies depending on the author, situation, and context, putting it in according to the "general node name" allows for more accurate distinction of the same subject and object (person, event, thing, etc.).

[0405] - For example, as shown in the table below, New Politics Alliance for Democracy - Representative Moon Hee-sang can be expressed differently as "New Politics Alliance for Democracy Emergency Response Committee Chairman Moon Hee-sang", "New Politics Alliance for Democracy Emergency Response Committee Chairman Moon Hee-sang", "New Politics Alliance for Democracy New Emergency Response Committee Chairman Moon Hee-sang", etc., and can be expressed differently as "Emergency Response Committee Chairman" by using pronouns or abbreviating the title. Therefore, to prevent confusion in node coding due to these different expressions, if a "general node name" is set, the analysis can be more accurate and data can be stored without omission.

[0406] - At this time, if you add a large category as an additional explanation or write it separately, it can be easy to classify it later.

[0407] Original text General node name New Politics Alliance for Democracy Party leadership New Politics Alliance for Democracy - Leadership New Politics Alliance for Democracy Moon Hee-sang, Emergency Response Committee Chairperson New Politics Alliance for Democracy - Assemblyman Moon Hee-sang New Politics Alliance for Democracy Park Young-sun, Emergency Response Committee Chairperson and Floor Leader New Politics Alliance for Democracy - Assemblyman Park Young-sun New Politics Alliance for Democracy Park Young-sun, Floor Leader New Politics Alliance for Democracy - Assemblyman Park Young-sun New Politics Alliance for Democracy Park Young-sun, Floor Leader New Politics Alliance for Democracy - Assemblyman Park Young-sun New Politics Alliance for Democracy Assemblyman Baek Gun-ki New Politics Alliance for Democracy - Assemblyman Baek Gun-ki New Politics Alliance for Democracy Members New Politics Alliance for Democracy - Members New Politics Alliance for Democracy Ahn Cheol-soo, Co-Chairperson New Politics Alliance for Democracy - Ahn Cheol-soo New Politics Alliance for Democracy Members New Politics Alliance for Democracy - Members New Politics Alliance for Democracy 15 members New Politics Alliance for Democracy - 15 members New Politics Alliance for Democracy Hardline struggle outside the National Assembly New Politics Alliance for Democracy - Hardline struggle outside the National Assembly StruggleNew Politics Alliance for Democracy LeadershipNew Politics Alliance for Democracy - LeadershipNew Politics Alliance for DemocracyNew Politics Alliance for DemocracyNew Politics Alliance for Democracy Representative Kim Han-gilNew Politics Alliance for Democracy - Representative Kim Han-gilPro-Roh forces within the New Politics Alliance for DemocracyNew Politics Alliance for Democracy - Pro-Roh forcesNew Politics Alliance for Democracy Emergency Response Committee Chairman Moon Hee-sangNew Politics Alliance for Democracy - Representative Moon Hee-sangNew Politics Alliance for Democracy Moon Hee-sang New Emergency Response Committee ChairmanNew Politics Alliance for Democracy - Representative Moon Hee-sang OmittedEconomy for the common peopleEconomy - Economy for the common peopleAgenda directly related to the lives of the common peopleEconomy - Agenda directly related to the lives of the common peopleEconomy - Economy for the common peopleSeo Byung-soo, Mayor of BusanBusan City - Seo Byung-soo, Mayor of BusanFostering the service industryEconomy - Fostering the service industry OmittedCrewSewol Ferry Crew 10 crewSewol Ferry Crew - 10 crew 15 crewSewol Ferry CrewCaptainSewol Ferry CaptainCaptain 4 people including Sewol crew - Captain Lee Jun-seok, Captain of Sewol Omitted Sewol Family Countermeasures Committee Spokesperson Sewol Family Countermeasures Committee - Spokesperson Sewol Family Countermeasures Committee Yoo Kyung-geun Spokesperson Sewol Family Countermeasures Committee - Spokesperson Sewol Family Countermeasures Committee - Yoo Kyung-geun Spokesperson Sewol Family Countermeasures Committee - Spread of false information Sewol Family Countermeasures Committee - Spokesperson Yoo Kyung-geun - Spread of false information Sewol Management Freedom Byeong-eon - Sewol Management Ministries related to Sewol Government - Ministries related to Sewol Extra-parliamentary struggle related to Sewol Opposition hardliners - Extra-parliamentary struggle related to Sewol Omitted Sewol ProblemSewol AccidentSewol IncidentSewol AccidentSewol AccidentSewol AccidentSewol AccidentMany people who truly mourned the National Investigation Agency's "Sewol" incident - Many people who truly mourned the "Sewol" incident The prolonged "Sewol" incident - The prolonged "Sewol" incident The ship's crew of the "Sewol" - The crew of the "Sewol" 15 crew members of the "Sewol" - 15 crew members of the "Sewol" The passengers of the "Sewol" - The passengers of the "Sewol" The original captain of the "Sewol", Shin Mo - The original captain of the "Sewol", Shin Mo The Family Response Committee of the "Sewol" (Chairman Kim Byung-kyun) - The Family Response Committee of the "Sewol" - Chairman Kim Byung-kyun The representative of the families of the "Sewol" - The Family Response Committee of the "Sewol" - Representative The Family Response Committee of the "Sewol" - The Family Response Committee of the "Sewol" The forces creating a stir around the families of the "Sewol" - The opposition forces The politicians creating a stir around the families of the "Sewol" - The opposition politicians People who say that not only the families of the "Sewol" but also the nation should be seen - People who say that not only the families of the "Sewol" but also the nation should be seen The representatives of the families of the "Sewol" survivors - The representatives of the families of the "Sewol" survivors Ordinary victims' families (students & ordinary people) - The representatives of the families of the "Sewol" survivors Other citizens who are not families of the "Sewol" - Other citizens who are not families of the "Sewol" Omitted The "Sewol" tragedy - The "Sewol" incident The "Sewol" incident The executive department of the Family Response Committee for the "Sewol" tragedy - The Family Response Committee of the "Sewol" - Executive department The National Response Conference for the "Sewol" tragedy - The National Response Conference for the "Sewol" tragedy Other media that shielded power in the coverage of the "Sewol" tragedy - Other media that shielded power in the coverage of the "Sewol" tragedy The Truth Investigation Committee for the "Sewol" tragedy - The Truth Investigation Committee for the "Sewol" tragedy The Truth Investigation Committee for the "Sewol" tragedy Actions to commemorate the victims of the "Sewol" tragedy - Actions to commemorate the victims of the "Sewol" tragedy The "Sewol" tragedy (tragic incident) - The "Sewol" incident Omitted Citizens - Citizens Members of parliament from civic groups - Members of parliament from civic groups (opposition) Dozens of citizens - Dozens of citizens One or two citizens - One or two citizens Civic groups - Civic groups Civic groups - The balloon containing a call for the enactment of a special law for the "Sewol" Citizens - Citizens Civil society - Civil society Civil society organizations - Civic groups - Civil society organizations The processions of citizens - Citizens Protesters - Protesters Omitted The ruling party, the Saenuri Party - The ruling party, the Saenuri Party The Saenuri Party's special committee for the National Investigation of the "Sewol" - The Saenuri Party's special committee for the National Investigation of the "Sewol" The internal affairs of the Saenuri Party - The internal affairs of the Saenuri Party The representative of the ruling party within the party - The representative of the ruling party within the party Members of parliament of the Saenuri Party - Members of parliament of the Saenuri Party Some members of parliament of the Saenuri Party - Some members of parliament of the Saenuri PartyPoliticiansSaenuri Party - PoliticiansPublic opinionPublic opinion?Citizens - ControversyPublic opinion - Public opinionRuling and opposition partiesSecond agreementSecond agreement on the Sewol Ferry Special ActPublic opinion to renegotiate the second agreementPublic opinion - Public opinion to renegotiate the second agreementRuling and opposition partiesThird negotiation process of the ruling and opposition partiesThird negotiation process of the Sewol Ferry Special ActRuling and opposition parties confrontation - Media research Public opinion pollSewol Ferry incident prolonged - Ruling and opposition parties confrontation - Media research Public opinion pollAgreement on the floor leaders of the ruling and opposition partiesFirst agreement on the Sewol Ferry Special Act - 8.7Agreement on the floor leaders of the ruling and opposition parties - 8.7First agreement on the Sewol Ferry Special Act - 8.7Ruling and opposition parties meetingRuling and opposition parties - Floor leaders meetingRuling and opposition parties leadershipRuling and opposition parties - LeadershipRuling and opposition parties agreement - Sewol Ferry Special Act 8.7First agreement on the Sewol Ferry Special Act - 8.7Agreement on the floor leaders of the ruling and opposition parties - Sewol Ferry Special Act 8.8First agreement on the Sewol Ferry Special Act - 8.7Public opinion to at least process the livelihood bills agreed upon by the ruling and opposition partiesCitizens - Public opinion that at least the livelihood bills agreed upon by the ruling and opposition parties should be processed

[0408] The opinion network and its implementation method, having the content or directionality described above and illustrated in the drawings, are merely one embodiment for implementing the present invention and should not be construed as limiting the technical spirit of the present invention. The protection scope of the present invention is defined solely by the matters set forth in the following claims, and embodiments that are improved and modified without departing from the gist of the present invention are deemed to fall within the protection scope of the present invention, provided that such modifications would be obvious to a person skilled in the art to which the present invention pertains.

[0409] [Explanation of symbols]

[0410] S10 data collection stage

[0411] S20 preprocessing stage

[0412] S21 Relevance Criteria Extraction Step

[0413] S22 Period-based extraction stage

[0414] S30 Edge List Creation Steps

[0415] S31 Opinion Existence Determination Stage

[0416] S32 Opinion Content Judgment Stage

[0417] S33 Opinion Direction Decision Stage

[0418] S40 Inter-coder Reliability Measurement Step

[0419] S50 Relationship Visualization Step

Claims

1. Data collection stage, which collects data containing information on the subject of interest; An edge list creation step in which the collected data is coded according to a set protocol to create an edge list for the target of interest; and A relationship visualization step that visually displays the relationship between the object of interest and other objects based on the edge list written above; including, The above edge list creation step is: An opinion existence judgment step that determines whether there is an opinion on the subject of interest in the collected data; If the above opinion exists, an opinion content judgment step to determine whether the opinion content is positive or negative; and It includes an opinion direction judgment step for extracting the subject that is the object expressing the opinion and the object that is the object receiving the opinion from the sentence and judging the direction of the opinion; The above relationship visualization step is, A method for implementing an opinion network having content or directionality, characterized in that the direction of the opinion is indicated using the direction of an arrow between the subject and the object, and the content of the opinion is indicated by changing the shape or color of the arrow between the subject and the object or by indicating it with a set character near the arrow.

2. In paragraph 1, The above edge list creation step is: A method for implementing an opinion network having content or direction, characterized in that the edge list is created by creating a list based on sentences having opinions in the collected data and entering a set of writing items into each of the created lists.

3. In paragraph 1, The above edge list creation step is: The above edge list includes the subject, object, content of the opinion and intensity of the opinion. The strength of the above opinion is expressed in numbers according to established standards. If the content of the above opinion is positive, it is expressed as +, and if it is negative, it is expressed as -. The above relationship visualization step is, A method for implementing an opinion network having content or directionality, characterized in that the total, maximum, minimum or average value of the strength of the opinions indicated by the numbers is calculated and displayed by changing the shape or color of the arrow between the subject and the object or by displaying it with a set character near the arrow.

4. In paragraph 1, The above edge list creation step is: It is performed by multiple coders, A method for implementing an opinion network having the above content or direction, A method for implementing an opinion network having content or direction, characterized in that it further includes an inter-coder reliability measurement step for measuring the reliability between coders who performed the above edge list creation step.

5. In paragraph 1, The above edge list creation step is: A method for implementing an opinion network having content or direction characterized by being performed by an AI coder using artificial intelligence technology.

6. In paragraph 1, A method for implementing an opinion network having the above content or direction, Further comprising a preprocessing step of preprocessing the collected data prior to the edge list creation step; The above preprocessing step is, A method for implementing an opinion network having content or direction, characterized by including a relevance criteria extraction step for determining and extracting data recognized as being related to a specific event or specific topic determined by a user among the collected data.

7. In paragraph 6, The above relevance criteria extraction step is: Arrange multiple pieces of data with recognized relevance in chronological order, The above edge list creation step is: A method for implementing an opinion network having content or direction, characterized in that the edge list is created by analyzing the above plurality of data in chronological order.

8. In paragraph 6, The above relevance criteria extraction step is: A method for implementing an opinion network having content or direction, characterized in that the relevance is determined based on a paragraph containing the sentence when the relevance is not determined within the sentence.

9. In paragraph 6, The above preprocessing step is, A method for implementing an opinion network having content or direction, characterized in that it further includes a period-based extraction step for extracting data corresponding to a period range set by a user from among the collected data.

10. In paragraph 2, The above edge list creation step is: Generate a list based on sentences that have opinions, If the sentence is composed of a complex sentence, compound sentence, or multiple sentence, the sentence is divided into multiple short sentences and a list is created by repeating the number of short sentences. If there are multiple opinions in a sentence, a list is created by repeating the number of opinions. A method for implementing an opinion network having content or direction, characterized in that when there are multiple subjects or objects in a sentence, a list is created by repeating the number of products of the subjects and objects.

11. In paragraph 3, The above edge list creation step is: A method for implementing an opinion network having content or direction, characterized in that, when the subject, the object, the content of the opinion or the intensity of the opinion included in the items of the above edge list are not judged within a sentence, the subject, the object, the content of the opinion or the intensity of the opinion is judged based on the paragraph including the sentence.

12. In paragraph 1, The above edge list creation step is: A method for implementing an opinion network having content or direction, characterized by assigning a representative name that can represent a name or expression having identity and creating an edge list based on the assigned representative name.

13. In paragraph 3, The above edge list creation step is: A method for implementing an opinion network having content or direction, characterized in that the above edge list writing items further include at least one of a document number assigned to each of the collected data, a sentence number within the document assigned the document number, the date and time the collected data was published or posted, and whether it is a topic sentence and picture.

14. An opinion network having content or direction implemented by any one of the methods in paragraphs 1 to 13.

Images