Document analysis system using artificial intelligence to identify the emotional state and keywords of the document writer and determine relationships among group members

1 . A document analysis system for analyzing documents using artificial intelligence to determine an emotional state and keywords of a document writer and identify relationships among group members, the system comprising:

a communication interface;

a memory storing an emotion list and a comparison reference table including a keyword similarity threshold, an emotion score threshold, an emotion similarity threshold, and a predefined tracking period; and

at least one processor configured to execute computer-readable instructions stored in the memory,

wherein the at least one processor is configured to:

(a) receive document data from member terminals via the communication interface, the document data including a context of daily documents commonly written by group members who regularly interact as a group in everyday life;

(b) analyze the context of each document by loading and executing an AI-based emotion analysis algorithm stored in the memory, extract a preset number of a plurality of primary emotions of the document writer from among a plurality of emotions listed in the emotion list stored in the memory, calculate an emotion score for each extracted emotion, and compute an overall emotion score of the document writer by performing an aggregate operation on the emotion scores;

(c) analyze the context of each document and extract a preset number of primary keywords from the context of the document by performing morpheme analysis, wherein the at least one processor is configured to process a semantic context of the document using a morpheme analyzer to decompose a text into morphemes, extract meaningful morphemes, remove stopwords, and apply TF-IDF or TextRank to automatically select the preset number of primary keywords;

(d) calculate the similarity between a plurality of keywords extracted from documents authored by group members on the same date, and, if the calculated similarity exceeds the keyword similarity threshold stored in the comparison reference table, connect the corresponding group members with a relationship line;

(e) for the group members connected by the relationship line on the same date, determine that the connected group members share emotions when at least one type of emotion extracted from the documents of the connected group members on the same date matches and a difference between their overall emotion scores is equal to or less than the emotion score threshold stored in the comparison reference table;

(f) for each relationship line between group members connected on the same date, when the connected group members have been determined to share emotions in step (e) and the difference between their overall emotion scores is equal to or less than the emotion similarity threshold stored in the comparison reference table, display the relationship line as being emotionally connected, either positively or negatively according to a type of the shared emotion, and, when the connected group members are not determined to share emotions in step (e) or the difference between their overall emotion scores exceeds the emotion similarity threshold, display the relationship line as indicating an emotionally unrelated connection;

(g) for each document generated on each date within the predefined tracking period, additionally display the relationship line between group members based on the keyword similarity analysis of step (d) and the emotion sharing determination of step (e) for that specific date, so that multiple relationship lines can be displayed between the same group members;

(h) calculate a centrality score for each group member by aggregating the number and types of relationship lines connected to each member;

(i) collect and accumulate daily centrality scores calculated for each member and store the data in the memory;

(j) construct a time-series dataset representing changes in each member's centrality over the predefined tracking period;

(k) track each member's centrality score on a daily basis and analyze trends over time; and

(l) if, during the predefined tracking period stored in the comparison reference table, the centrality score is determined to continuously decrease, determine that the member is gradually becoming isolated, and if the member is determined to be isolated, output an individual alert or a behavioral recommendation message to an administrator terminal or a corresponding member terminal.

2 . The document analysis system according to claim 1 ,

wherein the at least one processor is further configured to:

classify each emotion into one of positive, neutral, or negative categories;

refer to an emotion classification score table stored in the memory to assign a score corresponding to the category of each emotion; and

calculate an overall emotion score of the document writer by summing the assigned scores,

whereby the system analyzes a document using artificial intelligence to identify the emotional state and keywords of the document writer and to determine relational characteristics among group members.

3 . The document analysis system according to claim 1 ,

wherein the comparison reference table is stored in the memory and

includes a keyword similarity threshold, an emotion score threshold, a group identification threshold, a negative emotion threshold, an anomaly deviation threshold, an anomalous emotion frequency threshold, a cumulative negative threshold, an emotion similarity threshold, an insider threshold, an outsider threshold, and a risk judgment threshold,

which are used as reference criteria to determine relationships among group members.

4 . The document analysis system according to claim 1 ,

wherein the at least one processor is further configured to determine that specific members are associated as a group or as close companions when the number of days during which the members satisfy both a keyword similarity condition and an emotion similarity condition exceeds a group identification threshold stored in the comparison reference table,

and determines that a member is isolated when the member is not associated with any other member and the member's negative emotion score over a recent predetermined period exceeds the negative emotion threshold stored in the comparison reference table.

5 . The document analysis system according to claim 1 ,

wherein the at least one processor is further configured to:

calculate similarities among a plurality of keywords from documents written by group members on the same date, and to determine that the members share an event when the calculated similarity exceeds a keyword similarity threshold stored in the comparison reference table;

determine that the members share emotions when at least one type of emotion extracted from the documents of the group members on the same date matches, and the difference between comprehensive emotion scores calculated for the documents on the same date is less than or equal to an emotion score threshold stored in the comparison reference table; and

determine that a group of three or more members belong to the same group when the number of cases in which both events and emotions are shared among them over a recent period exceeds a group identification threshold stored in the comparison reference table,

to determine that two members are close companions when the number of cases in which they share both events and emotions exceeds the group identification threshold, and

to determine that a member is isolated when the member is not part of any group or close companionship and the member's negative emotion score over a recent period exceeds the negative emotion threshold stored in the comparison reference table.

6 . The document analysis system according to claim 5 ,

wherein the at least one processor is further configured to:

analyze daily comprehensive emotion scores of group members over a recent period,

and determine that a member exhibits an anomaly when the number of days on which the emotion score deviates from the average or median by more than an anomaly deviation threshold stored in the comparison reference table exceeds an anomalous emotion frequency threshold stored in the comparison reference table;

and determine that a member is a high-risk member when the member is not part of any group or close companionship, the member's negative emotion score over a recent period exceeds a cumulative negative threshold stored in the comparison reference table, and the member is determined to exhibit anomalies.

7 . The document analysis system according to claim 5 ,

wherein the at least one processor is further configured to:

based on a connection status or isolation determination of each group member,

retrieve from memory the keyword and emotion information that served as the basis for such determination,

analyze the retrieved information using an artificial intelligence algorithm,

and generate and output individualized support messages reflecting each member's connection status or isolation and corresponding keywords and emotions.

8 . The document analysis system according to claim 1 ,

wherein the at least one processor is further configured to,

determine a member as an insider if the member's centrality score is equal to or greater than an insider threshold stored in the comparison reference table,

determine a member as an outsider if the centrality score is equal to or lower than an outsider threshold, and

determine a member as a high-risk member if the ratio of relationship lines connected by negative emotions exceeds a risk judgment threshold stored in the comparison reference table.