Sundanese–Indonesian Code-Switching Conversational Speech Corpus — First-of-its-Kind, Consented

The language of West Java is not a language — it is two, braided sentence by sentence. This corpus captures spontaneous conversational code-switching between Sundanese (su) and Indonesian (id) as it actually happens among native speakers: mid-sentence alternations, borrowing, register shifts — not scripted prompts, not lab elicitation. Recorded with informed consent in Bandung, West Java, Indonesia. WHY IT'S FIRST-OF-ITS-KIND Code-switch corpora are research milestones (cf. SEAME for Mandarin–English), but for the Sundanese–Indonesian pair a consented, transcribed spontaneous corpus effectively does not exist commercially. Our Sundanese corpus work in this region has been supported and validated through Mozilla's community funding program; this listing is its production-grade continuation. WHAT'S INSIDE • Recordings 48 kHz / 16-bit WAV, quiet indoor environments • Verbatim transcripts with timestamps • Speaker metadata: age band, gender, domicile, dialectal area (e.g., Priangan) • Per-recording consent forms and provenance documentation — consent covers commercial AI training, sub-licensing, and cross-border transfer (structured under Indonesian Law No. 27/2022 on Personal Data Protection) QUALITY CONTROLS • SNR-checked recordings (target ≥ 25 dB), no clipping • Human QC pass; published grading rubric (A/B/C, open thresholds) • Documented collection SOP; consistent device & room standards SUITABLE FOR • Code-switch ASR and multilingual speech models • Code-switch point detect