Jazz improvisation pedagogy has long emphasized vocabulary development, yet empirical data on how developing improvisers use musical patterns remains limited. This study introduces the Developing Jazz Improvisers Database (DJID), a corpus of jazz solos by middle and high school students, and compares their improvisational patterns to those of advanced professionals in the Weimar Jazz Database (WJD).We employed Sonic Visualizer (SV), an open-source audio analysis application, to conduct comprehensive annotations of 16 DJID solo recordings. The SV interface enabled detailed multi-layered analysis including chord progressions, melodic content, beat structure, meter, phrase boundaries, and mid-level unit (MLU) categorization following the Jazzomat Research Project's established protocols. Each audio file was systematically annotated across these dimensions, with all analyzed files subsequently validated using an algorithm that provides detailed procedural feedback to ensure data integrity. The validated SV files were then imported into MeloSpyGUI, a computational analysis program developed by the WJD research team, which extracted features related to chordal diatonic pitch class (CDPCX) and mid-level analysis (MLA). CDPCX features quantified the proportion of chord tones, diatonic non-chord tones, and chromatic tones relative to annotated harmonic progressions, while MLA features classified phrases into nine categories (lick, line, melody, theme, rhythm, void, quote, expressive, and fragment) and measured their duration and frequency. This multi-stage analytical pipeline combining manual annotation in SV with automated computational analysis in MeloSpyGUI enabled direct comparison with the WJD corpus using identical analytical frameworks.Statistical analyses revealed significant differences between developing and advanced improvisers. Chi-square analysis showed developing improvisers used slightly more small intervals (60.8% vs. 59.4%, χ² = 7.408, p = 0.006), though with negligible effect size (Cramer's V = .006). Pitch class analysis demonstrated that developing improvisers relied more on chord tones (64% vs. 51%) while using fewer chromatic tones (9% vs. 22%), with chromatic usage showing moderate positive correlation with improvisation experience (R² = .44). Independent samples t-tests revealed that developing improvisers performed significantly longer melody phrases (M = 3.48s vs. 2.95s; t = 2.86, p = .004), while lick durations showed no significant difference (p = .782). MLU distribution analysis indicated developing improvisers utilized only 5 of 9 categories, predominantly lick (39.5%) and melody (38.2%).This methodological approach demonstrates how computational music analysis tools, when systematically applied to developing improvisers' recordings, can generate pedagogically relevant insights about vocabulary development and provide empirical foundations for jazz education curriculum.