[Collecting Pansori Dataset]#1 Introduction

Limitations of previous work

• The largest dataset has only 378 tunes of Pansori song
• Mean length of the tune in the dataset is 14 seconds for each track
• Metadata doesn't contain information specific for Pansori ex)유파, singer, sub-title of madang, cross-version

판소리(Pansori)

from Korean tradional song broadcasing, youtube, link

Description

Pansori (Korean: 판소리) is a Korean genre of musical storytelling performed by a singer and a drummer.

It's a unique form of narrative music that combines 창(singing), 발림(gestures), 아니리(dramatic storytelling) and music.

A Pansori performance typically involves two performers: a singer (sorikkun) and a drummer (gosu). The singer not only sings but also acts out the story through gestures and expressions. The drummer accompanies the singer with rhythms on the drum called buk, while also providing verbal encouragement known as chuimsae.

History

Pansori is thought to have originated in the late 17th century during the Joseon Dynasty. The earliest performers of pansori were most likely shamans and street performers, and their audiences were lower-class people.

During the 18th century, 12 song cycles, or madang (Hangul: 마당), were established as the repertoire of pansori stories.

There were originally twelve pansori works but only five complete works have survived to this day: "Simcheongga," "Heungbuga," "Sugungga," "Jeokbyeokga" and "Chunhyangga." Each work can take several hours(3 to 8 hours) to perform in full.

Uniqueness in Pansori

Inherited by oral lesson only

from the movie 서편제(Seopyeonje)

the genealogy line in Simcheongga, from the wikipedia, link

Pansori is traditionally passed down orally from master to student

However, in recent years there have been efforts to transcribe pansori into musical notation to help preserve it for future generations. These transcriptions can provide a useful reference but they cannot capture all of the nuances of a live performance.

Variation in Styles

three big styles in Pansori, from the wikipedia, link

Since Pansori is an oral tradition without a fixed musical notation, it allows for a great deal of variation and improvisation. This can result in different interpretations and performances of the same song, even by the same performer.

No Sheet music

Because Pansori is an oral tradition and performances can vary greatly, collecting audio data is extremely important for preserving the art form and studying its variations.

Contributions

• Preservation: Recordings can preserve performances of master singers for future generations to learn from and enjoy. This is especially important as some styles or songs may become less common over time.
• Study(Musicology): Scholars can study these recordings to analyze different performance styles, vocal techniques, interpretations of stories, and other aspects of Pansori.
• Data Analysis: With advances in technology such as machine learning and AI, collected audio data could potentially be used to analyze patterns or trends in Pansori music on a larger scale than previously possible.
• Follow-up studies: Apply data into desired future purposes ex) Vocal synthesis, Melody or music(audio) generation, Pose detection

Pansori is evolving

from 이날치(Leenalchi), youtube, link

The rise of modern or fusion Pansori bands like Leenalchi demonstrates the potential for innovation within this traditional art form.
we can leverage the power and beauty liying in Pansori by collecting and analyzing the data.

reference

wikipedia: https://en.wikipedia.org/wiki/Pansori

Dataset

Collect(Crawling) videos

Identify Sources

• singer, band
• title of Pansori song, subtitle of Pansori song
• Youtube Channels

Data Collection

• using pytube

Collecting metadata

Youtube metadata

Title of the song, Singer, Lyric, Transcribed Lyric

Optical Character Recognition(OCR)

can extract Title of the song, Subtitle(대목) of the song, Singer, Lyric

Organizing & Cleaning

Organizing Data

This could involve creating a database with metadata about each performance (e.g., performer name, performance date) and storing audio files in an organized manner.

Audio Detection (cleaning)

delete noise: not Singing part(Explanation, Applause sound)

Ethics and Privacy

Respect privacy and consent when collecting data from online sources like YouTube - this includes not only copyright considerations but also considerations around collecting user-generated content like comments.

Storage & Backup

Audio and video files can take up significant storage space so make sure you have enough storage capacity with proper backup mechanism in place.