Kren - Analyze Korean-English Language Songs

A small script that aims to visualize and understand K-Pop songs.

I have always noticed that K-Pop songs tend to end lines in English, or so I thought. One day, when thinking about this phenomenon, I thought it could be understood through some rudimentary processing. Thus, Kren was born.

(Below is the output of lyrics_violin)

Installation

Kren is not developed as a formal package. Try to evaluate Kren.R.

Packages required:

• Unicode
• syllable (requires manual installation)
• ggplot2
• dplyr

Getting Started

Search for your favorite songs, specifically its Genius page in original (unromanized & untranslated) lyrics.

Copy only the lyrics into a file, see ./examples/dummy_lyrics.txt for the correct format. Make sure there is an empty line in between each section, and to keep section headers.

The data produced by lyrics_tree_data is the main interface for this program.

# Object must be in this NAME._.TITLE format for plots to show proper title
Artist._.Song <- lyrics_tree_data("dummy_lyrics.txt")
lyrics_violin(Artist._.Song)
lyrics_smry(Artist._.Song)

See ./examples/usage.R for a full list of interface examples.

Using Containers (Docker/Podman)

Optionally, run:

podman build -t kren .
podman run -v ./examples:/home/src/examples kren

I prefer podman, but docker should work too.

Methodology Details

See METHODOLOGY.pdf for examples.

• Words
  • In both languages, words are defined as space-delimited chunks of Unicode.
  • Ex: “대통령은 of the great (truth)” has 5 words
• Syllables
  • Korean
    • Each block is counted as one syllable
    • Further work can be done on condensing slurred-vowels such as “다음”
  • English
    • Syllable counting is done through the syllable library, which uses poetrysoup.com
    • Further work can be done using Knuth-Liang Hyphenation
      • Such expansion would allow for analysis of many more languages (mostly latin script-based)
• Special Characters, such as æ
  • Processed as 0 syllables, as they are not in the Korean or English Unicode sets
• Ad-Libs
  • Ad-Libs are considered any lyric encapsulated by parentheses
  • Can be ignored through lyrics_tree_data(fileName, rm.adlibs = TRUE)
  • Example
    • Original: normal (this is an adlib) lyrics
    • With rm.adlibs = TRUE: normal lyrics
• Punctuation is ignored

Available Analyses

• lyrics_violin
  • Create a violin-esque plot
  • Plots syllable count and language across lines and sections
  • Meant to be read top-down, left to right
• lyrics_chist
  • Creates a centered histogram plot
  • Plots language word proportion across total line count
• lyrics_series
  • Creates a time series-esque plot
  • Plots language word proportion across total line count
  • Meant to be more artistic, think as a wallpaper
• lyrics_smry
  • Returns a df with sections and respective smry_f(stat)
    • Default is mean(kr_word_prop)
  • smry_f can be and one-dimensional statistic
  • stat is any of lyrics_STATS$line
• lyrics_begend
  • Analyses occurrence of language at beginning and end of lines
  • Computes McNemar’s test, with a null that the two marginal probabilities for kr_beg and kr_end are the same
    • Rejection means that there is a difference in language frequency between the beginning and end of lines
• lyrics_comp for a stat across line_tot
  • Compare two sets of lyrics on a stat
    • Default is kr_word_prop
    • stat is any of lyrics_STATS$line
  • Computes a two-sample Kolmogorov-Smirnov test
    • Effectively tells if the maximum distance between the two empirical cumulative distribution (ecdf) is large
    • Rejection means that the two sets do not come (“sample”) from a similar generator

Gallery

lyrics_chist

lyrics_series