{
"$comment": "Maps heterogeneous Hugging Face audio-dataset columns onto ngano's canonical schema. Heuristic aliases are applied to every dataset; repo overrides win where present. Runtime feature inspection always takes precedence over guesswork, so an unknown dataset still loads. Language identity is carried by three canonical fields: `language` keeps the source's own wording, `language_iso` the bare ISO 639-3 code, and `language_tag` the full BCP 47 tag. Column claiming is table-driven: an SDK must not add its own heuristics beyond this file, so that Python, JavaScript and Rust produce byte-identical mappings. Datasets-server or datasets feature types MAY be consulted only to confirm that a column already claimed as audio really is audio, never to claim a new one.",
"version": 1,
"canonical": {
"audio": "Audio handle. Lazy by default: {url, path, bytes, sampling_rate}. Never decoded unless the caller asks.",
"transcript": "Reference text for the utterance.",
"language": "Language name as the source labels it.",
"language_iso": "ISO 639-1/3 code where the source provides one.",
"country": "ISO 3166-1 alpha-2, from the source row or the catalogue entry.",
"speaker_id": "Stable speaker identifier within the dataset.",
"gender": "Speaker gender as stated by the source. Never inferred.",
"age": "Speaker age or age band as stated by the source.",
"duration_s": "Utterance duration in seconds.",
"sampling_rate": "Sample rate in Hz.",
"domain": "Recording domain, e.g. read, broadcast, clinical.",
"split": "Source split name.",
"dataset_id": "ngano catalogue id.",
"hf_repo": "Hugging Face repo id.",
"licence": "Licence string from the catalogue.",
"source_url": "Canonical URL for the dataset.",
"language_tag": "BCP 47 tag for this row's language; the primary subtag is always an ISO 639-3 code and an optional region subtag marks a country-specific variety, for example eng-NG. Resolved from the row's own language value where it states one, otherwise from the catalogue record when that record names exactly one language."
},
"aliases": {
"transcript": [
"sentence",
"text",
"transcription",
"transcript",
"normalized_text",
"normalised_text",
"sentence_norm",
"raw_transcription",
"utterance",
"utt",
"transcript_text",
"orthographic",
"orthographic_transcription",
"target_text",
"target",
"label_text",
"ref",
"reference",
"content",
"phrase",
"words",
"sentence_text",
"clean_text",
"corrected_text",
"gold_transcript",
"transcript_clean",
"text_clean",
"sentence_clean",
"annotation",
"caption",
"script",
"prompt",
"source_text"
],
"audio": [
"audio",
"speech",
"wav",
"path",
"file",
"audio_filepath",
"audio_path",
"file_name",
"filename",
"filepath",
"audio_file",
"waveform",
"input_values",
"wav_path",
"wavpath",
"audio_url",
"audiourl",
"audio_file_path",
"sound",
"sound_path",
"recording",
"recording_path",
"clip",
"clip_path",
"mp3",
"flac",
"ogg",
"opus",
"audio_bytes",
"raw_audio",
"signal",
"utterance_audio",
"voice",
"media",
"media_url"
],
"language": [
"language",
"locale",
"lang",
"language_name",
"lang_id",
"locale_id",
"language_code",
"lang_code",
"langid",
"iso_language"
],
"language_iso": [
"language_iso",
"iso_639_3",
"iso639_3",
"iso_code",
"lang_iso"
],
"speaker_id": [
"client_id",
"speaker_id",
"speaker",
"spk_id",
"speakerid",
"spkid",
"user_id",
"voice_id",
"narrator",
"speaker_name"
],
"gender": [
"gender",
"sex",
"speaker_gender",
"speaker_sex",
"voice_gender"
],
"age": [
"age",
"age_group",
"speaker_age",
"age_band"
],
"duration_s": [
"duration",
"duration_s",
"duration_sec",
"duration_seconds",
"length",
"audio_duration",
"dur",
"seconds"
],
"duration_ms": [
"duration_ms",
"duration_millis",
"length_ms"
],
"sampling_rate": [
"sampling_rate",
"sample_rate",
"sr",
"samplerate",
"frequency"
],
"country": [
"country",
"accent_country",
"nationality",
"country_code",
"region_country",
"country_iso",
"iso_country",
"accent",
"origin_country"
],
"domain": [
"domain",
"topic",
"category",
"genre",
"style",
"subject"
],
"split": [
"split",
"subset",
"partition"
],
"language_tag": [
"language_tag",
"lang_tag",
"bcp47",
"bcp_47",
"bcp47_tag",
"locale_tag",
"language_bcp47",
"ietf_tag",
"langtag"
]
},
"drop": [
"up_votes",
"down_votes",
"variant",
"segment",
"__index_level_0__",
"audio_id",
"id",
"idx",
"index",
"path_orig",
"is_gold_transcript",
"num_samples"
],
"unit_hints": {
"duration_ms": {
"canonical": "duration_s",
"multiply": 0.001
},
"duration_frames_16k": {
"canonical": "duration_s",
"multiply": 6.25e-05
}
},
"overrides": {
"mozilla-foundation/common_voice_17_0": {
"transcript": "sentence",
"speaker_id": "client_id",
"language": "locale",
"audio": "audio",
"verified": true
},
"mozilla-foundation/common_voice_16_1": {
"transcript": "sentence",
"speaker_id": "client_id",
"language": "locale",
"audio": "audio",
"verified": true
},
"mozilla-foundation/common_voice_13_0": {
"transcript": "sentence",
"speaker_id": "client_id",
"language": "locale",
"audio": "audio",
"verified": true
},
"mozilla-foundation/common_voice_11_0": {
"transcript": "sentence",
"speaker_id": "client_id",
"language": "locale",
"audio": "audio",
"verified": true
},
"google/fleurs": {
"transcript": "transcription",
"language": "language",
"language_iso": "lang_id",
"audio": "audio",
"gender": "gender",
"verified": true
},
"google/xtreme_s": {
"transcript": "transcription",
"language": "lang_id",
"audio": "audio",
"verified": true
},
"intronhealth/afrispeech-200": {
"transcript": "transcript",
"speaker_id": "speaker_id",
"audio": "audio",
"age": "age_group",
"gender": "gender",
"domain": "domain",
"verified": true
}
}
}