Emotions and emotion-related concepts play important roles in communication, natural language processing, and sentiment analysis. Emotion analysis is the extraction and analysis of emotions and opinions from sentences and statements. Emotions can be estimated from text and voice information, such as positive, negative, and neutral opinions, making it useful for SNS monitoring, product review analysis, etc.

　Colexification is a phenomenon in which one word can mean multiple concepts, and analysis allows us to understand how language captures and expresses concepts. We analyzed multiple concepts exhibiting colexification as a network, and investigated the relationships between concepts related to various emotions in multiple languages.

　As a result, they found core concepts that serve as hubs, ``GOOD,'' ``WANT,'' ``BAD,'' and ``LOVE,'' which are highly correlated with many other emotions. Next, we classified these four hubs into six hierarchies based on the high degree of relevance between them and other emotion concepts, and defined the same series of emotion concepts as five communities. As a result, many of the emotional concepts in the hierarchy close to the hub were from the same community as the hub, but there were also concepts such as "HOPE," "HAPPY," and "GLOOMY" that had low relevance to the hub even though they were in the same community. .

　Additionally, six concepts related to the opposite concepts of "GOOD" and "BAD" were identified: "BIG," "SMALL," "ANIMAL," "WHITE," "SICK," and "BRAVE." For example, the English word "BRAVE" includes elements related to "GOOD" and "BAD," such as "respectable" and "reckless."

　It is hoped that the development of this research will contribute to the establishment of innovative natural language processing and sentiment analysis methods.

Paper information:【Scientific Reports】Central emotions and hubs in a colexification network