Sea Shanties - A computational musicology project

---

A bit about Sea Shanties

Sea shanties are maritime work songs historically sung by sailors to coordinate labor, typically characterized by:

• call-and-response
• steady rhythms, and
• repeated refrains.

Kinds of Shanties:

• Capstan Shanty: Long, Continuous Tasks
• Halyard / Long-haul shanties: Coordinated hauling
• Short-drag shanties: Quick pulls.

---

Project Goals:

This project looks to identify certain features in sea shanties that may be relevant in understanding the use of those shanties.

Clustering & Classification: Do shanties of different types exhibit distinct melodic/rhythmic/lyrical features? Can we classify shanty type from these features?

Musical & Lyrical Comparison: Do sea shanties that share similar melodic and rhythmic traits (as measured from their musical scores) also exhibit similarities in their lyrics and/or textual themes?

---

Feature Extraction:

1. Pitch and Interval Features

• Pitch Range: Measure the distance between the highest and lowest pitches.

• Average Interval: Calculate the mean distance between consecutive pitches.

• Interval Complexity: Quantify the diversity in intervals — for example, using the entropy of the distribution of intervals.

• Leap Frequency: Count the proportion of intervals that exceed a step (e.g., intervals larger than a major second) compared to stepwise motion.

• Contour Directionality: Measure the percentage of upward versus downward movements.

• Melodic Contour Complexity: Compute metrics such as the number of directional changes or "turns" in the melody to capture the smoothness versus jaggedness of the line.

---

2. Rhythmic and Duration Features

• Average Note Duration: Compute the mean length of notes across the score.

• Rhythm Complexity: Quantify the diversity in note lengths — for example, using the entropy of the distribution of note durations.

• Syncopation Count/Index: Count the number of syncopated events or compute a syncopation score based on the degree of off-beat placements.

• Note Count per Bar: Calculate the number of note onsets per bar, accounting for meter variations.

• Note Count per Bar Variability: Determine the variance of the note count in each bar, accounting for meter variations.

• Rest Frequency: Count the ratio of rests to notes, contributing to the overall rhythmic pattern.

---

3. Structural and Statistical Complexity Features

• Score Length in Bars: Simply count the total number of measures in the score.

• Repetition: Rhythmic Pattern Recurrence: Quantify how often rhythmic patterns repeat throughout the score.

• Repetition: Melodic Pattern Recurrence: Quantify how often melodic patterns repeat throughout the score.

• Entropy of the Pitch Sequence: Use information theory (e.g., Shannon entropy) to gauge the unpredictability or variability in the sequence of pitches.

• Variance in Note Density: Measure changes in note density across the score to detect fluctuations between zones of activity.

---

Running the scripts:

Set up the virtual environment of your choice, with python version 3.10 installed. Once activated, run

pip install -r requirements.txt

You can then run the main script on your database of midi files. You do it like so:

python src/main.py <dataset_folder>