Ultimate Outline

明白了，我将开始调研英文文献中对 J-core（Japanese Hardcore）音乐的定义，并查找适用于 Python 统计分析的 MP4/ASV 数据来源，同时筛选所有与该课题相关的英文文献，包括计算机音乐发展、J-core 音乐特征、基于数据的音乐风格分析等内容，帮助你撰写论文各章节。完成后我会整理好文献引用并附上APA格式。

另外，以下是你可填写的作者信息模板：

Author information:

• Name: [Your Name Here]

• School: [University Name Here]

• City: [City of University]

• Country: [Country]

• Email: [School Email or Family/Company Domain Email / Gmail / Outlook]

稍后我会带来详细的研究文献汇总和对应章节建议。

Feature Analysis of J-Core Music: A Literature Review

Computer-Based Music and Algorithmic Analysis: Historical Development

Computer-based music composition and analysis have roots stretching back to the mid-20th century. Calculation and formal algorithms have influenced composition for centuries, but the digital era brought new possibilities (Algorithmic Composition – Communications of the ACM). One early milestone was Lejaren Hiller and Leonard Isaacson’s Illiac Suite (1956) – the first known computer-assisted composition, generated on the ILLIAC I computer (Algorithmic Composition – Communications of the ACM). Around the same time, pioneers like Iannis Xenakis began applying mathematical models to music. Xenakis, an engineer-composer, used stochastic processes on early computers (e.g. IBM 7090) to compose pieces such as Pithoprakta (1956) (Algorithmic Composition – Communications of the ACM) (Algorithmic Composition – Communications of the ACM). His work exemplified algorithmic composition by treating the computer as a “pilot” to explore musical structures, generating material which the composer could then accept or modify (Algorithmic Composition – Communications of the ACM) (Algorithmic Composition – Communications of the ACM).

Another often-cited landmark is Alan Turing’s 1951 experiments, where a Ferranti Mark I computer was programmed to play tunes like “God Save the King.” This occurred years before the more famous 1957 Bell Labs synthesis experiments, establishing Turing’s overlooked but leading role in early computer music (Turing and the History of Computer Music | SpringerLink) (Turing and the History of Computer Music | SpringerLink). Throughout the 1960s and 1970s, algorithmic and computer music research expanded: composers and researchers developed formalism for composition (e.g. Hiller’s later works, Max Mathews at Bell Labs) and began using computers to analyze music as well, anticipating today’s data-driven musicology. By the late 20th century, the field had grown into what we now call computational musicology – encompassing both algorithmic composition and algorithmic analysis of music. Comprehensive overviews, such as the Oxford Handbook of Algorithmic Music (2018), document this history and its lineage from the avant-garde classical tradition into modern computing.

Methodologically, early work relied on custom algorithms and nascent hardware. Hiller’s Illiac Suite, for example, used Markov chains and probability rules coded in FORTRAN (Algorithmic Composition – Communications of the ACM). Xenakis developed the Stochastic Music Program, translating statistical formulas into scores (Algorithmic Composition – Communications of the ACM). These efforts laid the groundwork for later statistical analysis of music: treating compositions as data. Over time, as computing power grew, focus expanded beyond composition to analysis: identifying patterns in melody, harmony, and rhythm across large corpora. This evolution set the stage for modern data-driven music analysis techniques, many of which leverage programming (e.g. Python) and statistical models to dissect musical features at scale.

J-Core (Japanese Hardcore): Evolution and Musical Characteristics

J-Core, short for Japanese hardcore, is a high-speed electronic music style that emerged in Japan in the late 1990s. It is essentially the Japanese take on hardcore techno, heavily influenced by the European gabber/happy hardcore scene but blended with Japanese pop culture. J-Core is characterized by an extreme tempo, frenetic energy, and a distinct otaku aesthetic. It commonly features fast breakbeats and kicks (often ~180–200 BPM) and liberally samples media from video games and anime, incorporating “cute” or kawaii elements into an otherwise hardcore sound (J-core - Wikipedia). In fact, it is marked by the use of high-pitched “chipmunk” voices (often anime or game character samples) and bright chiptune-style melodies over aggressive beats (A Kick in the Kawaii: Inside the World of J-Core | Red Bull Music Academy Daily). Visually, album art and imagery are filled with colorful anime-inspired graphics (J-core - Wikipedia), reflecting the genre’s deep roots in otaku subculture and the doujin (indie fan-made) music scene.

History and development: J-Core took shape in the late 1990s during Japan’s rave and game culture boom (J-core - Wikipedia). Early innovators like DJ Sharpnel (Jea + LeveL), DJ Technorch, m1dy, and REDALiCE began splicing UK/US rave and Rotterdam gabber sounds with J-pop, anime soundtrack snippets, Vocaloid vocals, and game sound effects (A Kick in the Kawaii: Inside the World of J-Core | Red Bull Music Academy Daily). The resulting style was faster and more sample-heavy than its Western counterparts. For example, DJ Sharpnel’s 1998 track “Project Gabbangelion” (riffing on Neon Genesis Evangelion) set a template: relentless 200+ BPM kickdrums, abrupt cuts and tempo shifts, and an eclectic collage of melodic material from anime and retro games (A Kick in the Kawaii: Inside the World of J-Core | Red Bull Music Academy Daily) (A Kick in the Kawaii: Inside the World of J-Core | Red Bull Music Academy Daily). Unlike typical club-oriented hardcore, J-Core tracks often lack a predictable dancefloor structure, opting for constant variation—“surgically sharp cuts and turns” with sudden breakdowns and insertions of whimsical melodies (A Kick in the Kawaii: Inside the World of J-Core | Red Bull Music Academy Daily). This unpredictable arrangement pattern is a hallmark of J-Core, making it feel closer to breakcore in form, yet usually more melodic and “bright” in tone.

Throughout the 2000s, J-Core grew via underground communities and the Internet. The term “J-Core” itself came into popular use around 2006, and netlabels like HARDCORE TANO*C (founded 2003) and Sharpnelsound spearheaded the scene (J-core - Wikipedia) (J-core - Wikipedia). J-Core became closely tied to rhythm games (e.g. Beatmania IIDX), which featured many J-Core tracks and helped increase BPMs to extreme levels to challenge players (ICCM2012—Abstract) (ICCM2012—Abstract). In fact, a recent comparative analysis of Japanese vs. British hardcore notes that Japanese hardcore (J-Core) tracks average around 186–187 BPM, significantly faster than the ~165 BPM of UK hardcore, and are more often in major keys (giving a “brighter” feel) (ICCM2012—Abstract). This difference is attributed to J-Core’s link with arcade music games and anime events, in contrast to Western hardcore’s club/rave orientation (ICCM2012—Abstract) (ICCM2012—Abstract). J-Core tracks frequently use simple, catchy chord progressions (often I–V–VI–IV or similar, common in J-Pop) but played at breakneck speed and supplemented by rapid arpeggios and octave-running melodies. The melodic tendencies lean toward either very “cute” major-key tunes or, sometimes, intentionally “disturbing” dissonant riffs for shock effect (A Kick in the Kawaii: Inside the World of J-Core | Red Bull Music Academy Daily) – a duality also noted by scene observers.

In summary, musicologically J-Core can be defined by:

• Tempo and Rhythm: Extremely fast BPM (typically 170–200+). Rhythms include gabber-style four-on-the-floor kicks, breakbeats, and sudden tempo switches.

• Harmony and Melody: Tends toward major keys and upbeat, consonant tonal material (to evoke happy/kawaii feelings) (ICCM2012—Abstract), though minor-key and darker tracks exist. Melodies often mimic chiptune or anime theme songs, and are sometimes pitch-shifted up (chipmunk effect) for a cute sound (A Kick in the Kawaii: Inside the World of J-Core | Red Bull Music Academy Daily).

• Textures and Sampling: Heavy use of samples from video games, anime, or pop culture. Layering of rapid oscillating synth leads, bright pads, and 8-bit video game sounds is common. Vocals (if present) are often sped-up anime dialogue or Vocaloid singing.

• Arrangement: Nonlinear structure with frequent breakdowns, tempo changes, and playful interludes (e.g. inserting a recognizable game melody) mid-track. Unlike mainstream EDM, J-Core is less about DJ-friendly structure and more about maximalist expression over a short span (many tracks are 3–5 minutes of constant variation).

• Cultural context: Firmly rooted in Japanese doujin circles – sold at comic conventions, game expos, or released on Niconico/YouTube – and linked to rhythm games and anime fandom (J-core - Wikipedia) (ICCM2012—Abstract).

These features distinguish J-Core within the broader hardcore techno umbrella. Understanding them is crucial for any statistical analysis of the genre’s musical features, as one would expect J-Core’s data (MIDI or audio features) to show signatures like higher average tempos, certain chord intervals (e.g. lots of perfect fifths and simple triads in the MIDI data reflecting its J-Pop influence), and spectral patterns corresponding to its dense, treble-heavy sound.

Data-Driven Music Analysis: Python and Statistical Approaches

Contemporary research frequently applies Python-based tools and statistical methods to analyze music features across large datasets, a practice often termed data-driven musicology or computational music analysis. Python has become a lingua franca in this field due to its rich ecosystem of libraries for music and audio processing. Studies in this domain range from computational music theory analyses (e.g. melodic or harmonic pattern discovery) to machine learning models for genre classification.

A prominent example is the Information Dynamics of Music (IDyOM) model, originally implemented in Lisp, which analyzes melodic expectation using statistical learning. Recently, researchers introduced IDyOMpy, a Python re-implementation of this model that computes information content and entropy for each note in a melody, after training on a corpus (IDyOMpy: A new Python-based model for the statistical analysis of musical expectations - PubMed). Such a tool enables quantitative analysis of melodies – for instance, measuring how unpredictable a J-Core melody is compared to other genres. IDyOMpy’s Python base makes it accessible and extensible, illustrating how Python facilitates complex musicological computations (like entropy of note sequences) that were once confined to specialized languages (IDyOMpy: A new Python-based model for the statistical analysis of musical expectations - PubMed) (IDyOMpy: A new Python-based model for the statistical analysis of musical expectations - PubMed).

Beyond specific models, general frameworks for music feature extraction have emerged. Llorens et al. (2023) present musif, a Python package that automatically extracts a large array of features from symbolic music (e.g. MIDI or MusicXML) (musif: a Python package for symbolic music feature extraction). This includes features curated by musicologists – from melodic interval patterns and scale usage to chord frequencies and rhythm syncopation measures. The package supports common symbolic formats (MusicXML, MIDI, etc.) and allows custom feature creation using standard Python libraries (musif: a Python package for symbolic music feature extraction) (musif: a Python package for symbolic music feature extraction). Such a tool could be directly applied to a collection of J-Core tracks (if MIDI transcriptions are available) to statistically profile their melodic and harmonic tendencies. For example, one might use musif to compute the distribution of chords (e.g. how often power chords or major triads appear in J-Core), or average note density (notes per second) which would likely be high given the fast tempos.

There are also tools aimed at making music analysis approachable for researchers and students. Eck (2023) developed the Interactive Music Analysis Tool (I-MaT) built on Python libraries like music21 (a toolkit for music theory analysis in Python) and MidiTok ((PDF) Interactive Music Analysis Tool (I-MaT)) ((PDF) Interactive Music Analysis Tool (I-MaT)). I-MaT provides a user-friendly interface (CLI) to perform statistical analysis and visualization on music data without requiring advanced programming skills ((PDF) Interactive Music Analysis Tool (I-MaT)). It can tokenize music into analytical units and compute patterns, serving as a bridge for musicologists to leverage Python’s power. This trend of creating accessible analysis tools means that even without extensive coding, one can conduct, say, a statistical survey of J-Core chord progressions or rhythm patterns and visualize the results.

Machine learning and AI techniques are also heavily used. Kishi et al. (2024), in a study aimed at classifying J-Core vs UK hardcore, leveraged Python-based libraries for audio processing and modeling (ICCM2012—Abstract). They converted 30-second WAV clips into mel-spectrograms and used frameworks (e.g. a Python implementation of music2vec and a CNN model Soundnet) to learn feature representations (ICCM2012—Abstract) (ICCM2012—Abstract). The models, trained on a dataset of J-Core and UK hardcore tracks, identified differences like the aforementioned BPM and key distinctions. The use of Python in this context (for data handling, model training, etc.) underscores its role in experimental musicology. Python’s ecosystem (NumPy, SciPy, librosa for audio, music21 for symbolic, scikit-learn or PyTorch/TensorFlow for modeling) allows researchers to combine signal processing with statistical analysis seamlessly. For instance, one could calculate the average spectral centroid of J-Core songs (a proxy for brightness) using librosa, or analyze chord transitions using music21, all within one environment.

In summary, methodologies in recent literature emphasize:

• Feature Extraction: Tools like musif and music21 to obtain symbolic features (scales, intervals, chords) and librosa for audio features (spectral properties, tempo, etc.).

• Statistical Analysis: Information-theoretic measures (entropy, surprise as in IDyOMpy (IDyOMpy: A new Python-based model for the statistical analysis of musical expectations - PubMed)), frequency counts, distributions and visualizations (histograms of interval sizes, chord type frequencies, etc.).

• Machine Learning: Classification (genre, composer identification) and clustering using feature vectors. For J-Core, classification can distinguish it from other EDM by its feature profile (ICCM2012—Abstract) (ICCM2012—Abstract). Unsupervised learning might reveal groupings of tracks by sub-style (e.g. speedcore vs happy J-Core).

• Visualization: Plotting results in meaningful ways (tonal plots, self-organizing maps of songs, PCA projections of feature space, etc.) to interpret the statistical patterns in musical terms.

Notably, all these approaches can be conducted in Python, making it a unifying platform for the project at hand.

Open Datasets for J-Core and Related Genres

One challenge in analyzing J-Core is obtaining a suitable dataset. Ideally, we want collections of J-Core or similar Japanese electronic hardcore tracks with accessible audio (or MIDI). While J-Core is niche, several broader open datasets and resources can be leveraged:

• Free Music Archive (FMA): A large open dataset of 106,574 tracks spanning 161 genres (ICCM2012—Abstract). It includes many electronic tracks under Creative Commons and could contain some hardcore techno relevant to J-Core. FMA provides audio (mostly MP3) and metadata; researchers have used it for genre classification tasks (ICCM2012—Abstract). Although J-Core might not be an explicit tag, subgenres like gabber, happy hardcore, chiptune, or game music in FMA could be proxied to study aspects of J-Core. The size of FMA allows robust statistical analysis, and one could filter it to a subset of high-BPM electronic tracks to approximate a J-Core-like sample.

• MTG-Jamendo Dataset: An open dataset for music auto-tagging with over 55,000 full audio tracks, tagged with genres, moods, and instruments (The MTG-Jamendo Dataset for Automatic Music Tagging) (The MTG-Jamendo Dataset for Automatic Music Tagging). While Jamendo (a platform for independent music) may not label something “J-Core,” it has tags for hardcore, j-pop, anime etc., and one could retrieve tracks that match a combination (e.g. genre:“Hardcore” + tag:“Japan” or similar). All tracks are Creative Commons and provided as high-quality audio (The MTG-Jamendo Dataset for Automatic Music Tagging). This dataset is valuable for training statistical models or doing tag-based analysis, and its large scale could help identify J-Core-adjacent music for comparison.

• Lakh MIDI Dataset: A collection of 176,581 MIDI files, with 45,129 of them aligned to corresponding songs in the Million Song Dataset (MSD) (Machine Learning Datasets | Papers With Code). While this dataset is predominantly Western pop/rock, it is useful for symbolic analysis techniques. If MIDI files for J-Core songs aren’t readily available, one could still use Lakh to develop analysis methods (for example, prototyping chord recognition or motif analysis on known MIDI songs) and then apply those methods to any transcribed J-Core MIDI. Additionally, some anime or game music MIDIs in Lakh might capture the kind of melodies J-Core artists sample. MIDI data allows detailed feature extraction (scales, motifs, rhythm patterns) without the complexity of audio signal processing.

• Custom Genre Datasets: The study by Kishi et al. (2024) essentially created a small specialized dataset: 180 songs (30-second clips) each of J-Core and UK hardcore (ICCM2012—Abstract) (ICCM2012—Abstract). While not publicly released, this demonstrates the approach of curating a genre-specific corpus. Similarly, one might compile J-Core tracks from netlabels (many J-Core artists release free tracks on SoundCloud or Bandcamp). These could be converted to consistent format (e.g. 30-second WAV segments at 22,050 Hz as Kishi et al. did (ICCM2012—Abstract)) to feed into analysis pipelines. Community resources like Vocaloid database or Touhou doujin music archives might also provide leads on J-Core tracks (since some J-Core is linked to Touhou arrangements or Vocaloid remixes).

• Other open EDM datasets: There are genre datasets focusing on electronic music that, while not Japan-specific, cover fast hardcore styles. For instance, the GTZAN genre dataset (a classic small benchmark) includes a “metal” and could be extended with “techno” but is limited. More relevant is the ISMAS (ISMIR 2004 dance music dataset) or others that include “hardcore” or “gabber” categories. Additionally, the EDM genre dataset used in some MIR research (if available) might have a hardcore category. The recently released EDM genre classification dataset (EDM32) includes various EDM subgenres and could potentially have a category overlapping with J-Core style.

When seeking audio+MIDI format, we might consider video game music datasets: since J-Core often draws from game music, any dataset of game OSTs in MIDI (e.g. Bemani rhythm game songs, if available) would be relevant. While not openly published, MIDI collections for rhythm games or fan-made transcriptions of J-Core tracks (some fans do create MIDI files of favorite songs) could be sources of symbolic data. Given J-Core’s ties to doujin culture, some compilation albums (e.g. at Comiket) might include a mix of audio tracks that could be manually assembled into a research dataset with proper tagging.

Summary of dataset options: For a statistically significant analysis, one may use large datasets like FMA and MTG-Jamendo to establish general features of hardcore techno, and then supplement with curated J-Core examples. FMA provides broad coverage and diversity (ICCM2012—Abstract), MTG-Jamendo provides multi-tagged content for targeted searches (The MTG-Jamendo Dataset for Automatic Music Tagging), and Lakh MIDI offers millions of notes to test melodic analysis methods (Machine Learning Datasets | Papers With Code). The key is ensuring the J-Core content is represented. If needed, a new dataset can be compiled from legally available J-Core tracks (many artists release free MP3s). This would mirror the approach of Kishi et al., who prepared specific training and testing sets for J-Core vs UK hardcore (ICCM2012—Abstract) (ICCM2012—Abstract). In doing so, they were able to derive quantitative differences (tempo, key, etc.) as shown in their results tables and figures.

Suggested Figures and Tables

• Figure 1: Timeline of Computer Music Milestones – A visual timeline from 1951 to present, highlighting key events (e.g., 1951 Turing’s computer tune, 1956 Illiac Suite, 1960s Xenakis’s stochastic compositions, 1980s MIDI invention, 2000s MIR evolution). This would contextualize the historical development of algorithmic music analysis.

• Figure 2: J-Core vs UK Hardcore Feature Comparison – A bar chart or infographic comparing average BPM, % of tracks in major key, and other features between J-Core and UK hardcore. Data could be drawn from Kishi et al. 2024 (who found J-Core ~186.7 BPM vs UK ~165.5) (ICCM2012—Abstract). This figure would illustrate the statistical differences in melody/harmony (key mode) and tempo that characterize J-Core.

• Figure 3: Example Spectrogram and Waveform of a J-Core Track – A spectrogram showing the frequency content of a short J-Core excerpt, alongside a waveform. This could reveal the dense frequency spectrum (strong highs from cymbals and chip-tune leads, and booming lows from kick drums). It would serve as a visual example of the audio features being analyzed (e.g., one can point out the rapid kicks and complex high-frequency patterns visible on the spectrogram).

• Table 1: Feature Extraction Summary – A table listing which musical features will be analyzed and their definitions. For instance: Melodic interval distribution (frequency of intervals 1, 2, … in semitones), Chord type frequency (% major, minor, diminished, power chord, etc.), Rhythmic density (notes per bar at given tempo), Average BPM, Spectral centroid (Hz), etc. The table can have a column for “Expected J-Core tendency” (e.g., high BPM (~180-200), preference for major chords and pentatonic scales, high note density). This would be informed by the literature review findings.

• Table 2: Datasets for Analysis – A comparison table of potential datasets: listing FMA, MTG-Jamendo, Kishi’s custom dataset, Lakh MIDI, etc., with columns for Size (tracks), Format (audio/MIDI), Genres, and relevance to J-Core. For example, FMA – 106k tracks, MP3 audio, 161 genres (could filter Hardcore), Open license; Kishi2024 – 200 tracks, WAV audio, J-Core vs UK hardcore, private; Lakh MIDI – 176k MIDI files, many genres, symbolic, etc. This provides a quick reference for what data sources are available and their characteristics.

By including such figures and tables, the paper can more effectively communicate the context, methodology, and findings of the J-Core feature analysis. They would complement the textual analysis with visual evidence and organized data, aiding reader comprehension of both historical background and technical results.

Annotated Bibliography (APA format)

1. Edwards, M. (2011). Algorithmic Composition: Computational Thinking in Music. Communications of the ACM, 54(7), 58–65. – This article provides an accessible history of algorithmic composition, tracing it from pre-computer techniques to modern computer-based methods. Edwards discusses key figures (like Hiller and Xenakis) and how the formalization of musical processes led to the use of algorithms in composition (Algorithmic Composition – Communications of the ACM) (Algorithmic Composition – Communications of the ACM). It offers insight into the evolution of computer-based music, emphasizing that computational approaches to music have roots going back at least a millennium in concept (Algorithmic Composition – Communications of the ACM). This source is useful for understanding the historical context and philosophical motivations behind algorithmic music analysis.

2. Copeland, J., Long, J., & Carpenter, B. (2017). Turing and the History of Computer Music. In J. Copeland et al. (Eds.), The Turing Guide (pp. 219–230). Oxford University Press. – An open-access book chapter detailing Alan Turing’s pioneering work in creating computer-generated music (Turing and the History of Computer Music | SpringerLink). It dispels the myth that computer music started at Bell Labs in 1957 by documenting Turing’s 1951 programming of a computer to play melodies (Turing and the History of Computer Music | SpringerLink). The authors employed a form of “digital archaeology” to reconstruct and even restore the original recordings of the computer’s output (Turing and the History of Computer Music | SpringerLink) (Turing and the History of Computer Music | SpringerLink). This chapter underscores Turing’s role and provides technical context for how the first computer music was achieved, illustrating early intersections of computing and music that predate formal algorithmic composition.

3. Kishi, N., Shioya, R., & Nakabayashi, Y. (2024). Development of a system for classifying J-Core and UK Hardcore music genres using music2vec. Proceedings of the ICCM 2024 Conference (International Conference on Computational Mechanics) (ICCM2012—Abstract) (ICCM2012—Abstract). – This conference paper (in English) is one of the few academic studies focusing on J-Core. The authors apply AI techniques (music2vec embeddings and a SoundNet CNN) to distinguish Japanese hardcore (J-Core) from UK hardcore. They detail their dataset of J-Core vs UK hardcore tracks, the feature extraction via mel-spectrograms, and classification results. Notably, the paper presents statistical comparisons: e.g., Japanese Hardcore’s higher average BPM and greater tendency toward major keys vs British Hardcore (ICCM2012—Abstract) (ICCM2012—Abstract). It demonstrates a practical methodology using Python-based tools (they explicitly mention using Python for the implementation) (ICCM2012—Abstract). This source is invaluable for understanding characteristic musical features of J-Core in a data-driven way and as a reference for techniques to analyze genre-specific audio features.

4. Host, V. (2015, January 19). A Kick in the Kawaii: Inside the World of J-Core. Red Bull Music Academy Daily (A Kick in the Kawaii: Inside the World of J-Core | Red Bull Music Academy Daily) (A Kick in the Kawaii: Inside the World of J-Core | Red Bull Music Academy Daily). – A journalistic feature that explores the J-Core scene and sound. Vivian Host interviews key artists and describes J-Core’s formation in late-90s Japan, painting a vivid picture of its musical traits. The article highlights how producers merged Western rave/hardcore with anime and game music, emphasizing J-Core’s fast tempos, lack of conventional structure, and use of “cute” samples (A Kick in the Kawaii: Inside the World of J-Core | Red Bull Music Academy Daily). It also recounts historical anecdotes, such as DJ Sharpnel’s early releases. While not peer-reviewed, this piece is well-researched and often cited in Wikipedia (J-core - Wikipedia) (J-core - Wikipedia); it provides cultural context and qualitative detail on J-Core’s aesthetic, complementing academic sources with first-hand genre insights.

5. Jenkins, D. (2018, April 26). Beyond J-Core: An Introduction to the Real Sound of Japanese Hardcore. Bandcamp Daily (J-core - Wikipedia) (J-core - Wikipedia). – Another scene report that serves as an introduction to J-Core for English-language readers. Dave Jenkins outlines the genre’s definition, noting its origin (initially known as “Japcore”) and key characteristics like otaku culture influences, colorful kawaii imagery, and prevalence in rhythm games (J-core - Wikipedia). The article also points readers to notable labels and artists, thus functioning as both a historical overview and a guide to listening. It confirms and expands on features such as J-Core’s ties to the doujin music market and the high-energy, sample-driven style. This source is helpful for a musicological perspective on what defines J-Core and how it evolved, written in an accessible yet informative style.

6. Marion, G., Gao, F., Gold, B. P., Di Liberto, G. M., & Shamma, S. (2025). IDyOMpy: A new Python-based model for the statistical analysis of musical expectations. Journal of Neuroscience Methods, 415, 110347 (IDyOMpy: A new Python-based model for the statistical analysis of musical expectations - PubMed). – This peer-reviewed article introduces IDyOMpy, a Python tool for computing information-theoretic features (like entropy and information content) of melodies. The authors re-implemented the established IDyOM model in Python to make it more accessible to researchers. The paper provides details of the model’s validation and its application to neural and behavioral music studies. For our purposes, IDyOMpy represents a cutting-edge method to quantify melodic structure and predictability. The ability to train on a corpus and then measure each note’s surprise is directly applicable to analyzing melodic tendencies in J-Core – for example, determining if J-Core melodies are more or less predictable than those in other genres. This source is technical but important for methodology, showcasing how Python can be used to bridge music theory, statistics, and cognition in music analysis.

7. Llorens, A., Simonetta, F., Serrano, M., & Torrente, Á. (2023). musif: A Python package for symbolic music feature extraction. arXiv preprint arXiv:2307.01120 (musif: a Python package for symbolic music feature extraction). – This open-access paper (also presented at a music technology conference) introduces musif, a comprehensive feature extraction library for symbolic music (scores/MIDI). The authors enumerate a wide range of implemented features covering melody, harmony, rhythm, and texture, designed with input from musicologists (musif: a Python package for symbolic music feature extraction). They emphasize the tool’s extensibility and ease of use with common formats (musif: a Python package for symbolic music feature extraction). The paper includes use-case examples and validation on music corpora. For a project analyzing J-Core, musif could drastically speed up the gathering of statistical data from any MIDI transcriptions of songs. The annotation of this reference underlines that it’s a bridge between music theory and data science, enabling complex analyses (like counting how often certain scales or chords appear) without starting from scratch. It’s a prime example of the state-of-the-art resources available in Python for music analysis.

8. Eck, S. O. (2023). Interactive Music Analysis Tool (I-MaT). In Proceedings of JADH 2023 – Digital Humanities Conference (Tokyo, Japan) ((PDF) Interactive Music Analysis Tool (I-MaT)) ((PDF) Interactive Music Analysis Tool (I-MaT)). – This conference paper (in the digital humanities context) describes I-MaT, a Python-based tool aimed at making music analysis interactive and accessible for educational settings. It uses libraries like music21 for backend processing but presents a simpler interface for users. The abstract highlights that I-MaT allows statistical analysis, visualization, and tokenization of music data without steep learning curves ((PDF) Interactive Music Analysis Tool (I-MaT)). This is relevant as it exemplifies efforts to democratize computational musicology. For the J-Core analysis project, I-MaT or similar tools could be used to prototype analysis workflows or even involve students in exploring the dataset. The reference is annotated to show how user-friendly tools can still carry out rigorous analysis, indicating that one doesn’t always need to code everything from scratch – a beneficial consideration for planning the analysis methodology.

9. Defferrard, M., Benzi, K., Vandergheynst, P., & Bresson, X. (2017). FMA: A Dataset for Music Analysis. In Proc. of the 18th International Society for Music Information Retrieval Conference (ISMIR) (pp. 316-323) (ICCM2012—Abstract). – This paper introduces the Free Music Archive (FMA) dataset, one of the largest publicly available audio collections for research. It details the dataset’s composition (≈106k tracks, 161 genres) and provides baseline analyses and tasks (like genre classification). The authors discuss how the data is organized and the potential research opportunities it affords. Citing this is useful to justify using FMA: it is a well-documented, open benchmark dataset in MIR. For our purposes, FMA can serve as a source of hardcore/electronic tracks under Creative Commons, and this reference provides the authoritative description of its scale and genre taxonomy (ICCM2012—Abstract). The annotation here notes FMA’s relevance as a foundation for statistical genre studies and highlights that it has been used in prior music feature research, lending credibility to including it in the project.

10. Bogdanov, D., Won, M., Tovstogan, P., Porter, A., & Serra, X. (2019). The MTG-Jamendo Dataset for Automatic Music Tagging. In Proc. of the Machine Learning for Music Discovery Workshop, ICML 2019 (The MTG-Jamendo Dataset for Automatic Music Tagging). – This reference describes the creation of the MTG-Jamendo dataset, which comprises over 55k full audio tracks with 195 tags covering genres, moods, and instruments (The MTG-Jamendo Dataset for Automatic Music Tagging). The authors outline how they collected Creative Commons music from Jamendo and curated tags. They also provide baseline results for auto-tagging algorithms on this dataset. For the J-Core paper, this source is included to emphasize the availability of large tag-labeled datasets and how they can be mined for niche genres. The annotation indicates that MTG-Jamendo’s rich tagging system might help filter tracks relevant to J-Core (e.g., tracks tagged “Hardcore”, “Anime”, etc.), and its open nature allows integration into our analysis. It supports the point that open datasets exist even for less mainstream music, and using them can bolster the statistical power of genre studies.

11. Raffel, C., & Ellis, D. P. W. (2016). Large-scale analysis of MIDI data: The Lakh MIDI Dataset. In Proc. of the 17th International Society for Music Information Retrieval Conference (ISMIR). – This paper (implicitly referenced via dataset description) introduced the Lakh MIDI Dataset, aligning thousands of MIDI files with the Million Song Dataset. It discusses how the dataset was assembled and the potential for symbolic music analysis at scale. We cite the dataset’s key stats (Machine Learning Datasets | Papers With Code): 176k MIDI files, with many matched to actual songs, enabling studies that connect MIDI features to audio or metadata. The annotation for this reference clarifies that while Lakh is not specific to J-Core, it provides a huge testbed for developing analysis techniques (which can then be applied to J-Core music). It’s evidence of the data-driven approach in musicology, showing how researchers can leverage big data. Including this reference demonstrates awareness of symbolic data resources and justifies any approach that involves MIDI-based feature analysis or machine learning in the project.

Each of these references contributes a piece to the puzzle: from historical context and genre-specific knowledge (refs 1–5) to methodologies and tools (refs 6–8) to data sources (refs 9–11). Together, they form a comprehensive foundation for analyzing J-Core music features with statistical and computational methods using Python. Each entry above is annotated with its relevance, ensuring this bibliography not only lists sources but also explains how they inform the research.