Research Stories

Using Natural Language Processing Algorithms
to Track Emotions From 30 Years of K-Pop Hits

An in-depth analysis of over 2,900 K-pop hit songs from 1990 to 2019 revealed increased positive emotional content and decreased negative emotional content embedded within the lyrics, highlighting a generational shift in emotions that many Koreans have valued and preferred over the past 30 years.


  • Using Natural Language Processing Algorithms ▼ to Track Emotions From 30 Years of K-Pop Hits
Scroll Down

Prof. Minue Kim (Department of Psychology) collaborated with Prof. Wonkwang Jo (Graduate School of Public Health, Seoul National University) on a research project entitled, Tracking emotions from song lyrics: analyzing 30 years of K-pop hits.

Emotions that are shared by a large number of people are known to broadly impact affective experiences at the individual level. Music, especially hit songs that have garnered popularity in a society, can be usefully viewed as being reflective of the emotional preferences and experiences of its members. As such, analyzing the musical features of K-pop songs that emerged as hits over time may offer a useful strategy for understanding the emotional characteristics of Koreans at the sociocultural level.

K-pop has risen in popularity in the mainstream music scene and is enjoying a more global reach than before, and thus the emotions expressed in K-pop have implications beyond Korea.

Using text mining and natural language processing algorithms on 30 years of K-pop hit songs, we sought to answer the following research questions: What are the emotions observed in K-pop lyrics, and how do they change over time? Specifically, using morpheme frequency analysis and structural topic modeling on song lyrics, we sought to investigate the changes in the appearance of words and topics conveying positive and negative emotions.

In this work, we used lyrics from songs on Melon’s top 100 list for each year from 1990 to 2019 as our data. Both morpheme frequency analysis and STM yielded converging results: The proportion of adjectives and topics that express positive emotions showed an increasing trend in the past 30 years. Conversely, the proportion of topics that convey negative emotions showed a decreasing trend during the same time period. Surprisingly, this temporal shift is the exact opposite of what prior research has found from popular songs in the United States.

These seemingly conflicting observations may be reconciled by considering cultural differences: that the United States boasts a highly individualistic culture, whereas Korea is a traditionally collectivistic culture. For instance, in an already individualistic culture like the United States, further increases in individualistic traits may exacerbate the adverse aspects of individualism (e.g., a rise in narcissistic traits).

On the other hand, beneficial features of individualism emerge as individualistic traits increase, at least initially, in a collectivistic culture such as Korea. Individualistic values being introduced to a collectivistic society could encourage people to pursue their individual goals and life course, while distancing themselves against the potentially excessive obligation to the community. We also argue that economic growth (e.g., rise in individual purchasing power and gross domestic product) may also have served as the background for increasing feelings of and/or preference for positive emotions.

More generally, our study illustrates a strategy for tracking emotions that people value and prefer from large natural language data, supplementing existing methods such as self-reported surveys and laboratory experiments.

This work is published at ‘Emotion’ (

*Tracking emotions from song lyrics: Analyzing 30 years of K-pop hits (Journal: Emotion, DOI: 10.1037/emo0001185)