otherWantedaudiopronunciationplace-namesspeech
How locals pronounce the places around them
What they need
Pick a city or region you know well and record 30 different local place names in short sentences. Streets, neighborhoods, stations, landmarks and nearby towns all count. I'm especially interested in names whose local pronunciation isn't obvious from the spelling. One consenting adult speaker is enough. For each recording, include the name as written, a link identifying the place, the exact sentence spoken and the point in the audio where the name occurs. Add a short pronunciation note for at least ten of the names; plain-language explanations are fine, and IPA isn't required. Any language is welcome. Please describe your connection to the area without sharing a home address.
Requirements
formatzip
coverage30 different places in one named city or region, covering at least 2 of these types: streets, neighborhoods, towns, landmarks, transit stops, or natural features. At least 10 names must have a brief note explaining a non-obvious local pronunciation. Any recording language is welcome if used consistently and identified.
acceptance30 distinct verified places, intelligible original speech, accurate transcripts, and target-name timestamps within approximately 0.25 seconds. No repeated names with different carrier sentences to inflate the count. Location links verify place identity; they are not evidence of pronunciation.
deliverablesOne ZIP containing 30 WAV or FLAC clips, places.csv, README.txt, and an anonymous speaker-consent attestation. Each clip should be 3-10 seconds long and contain one target place name in a short natural sentence. Use mono audio at 16 kHz or higher.
verificationUse an official place, municipal, transit, or map reference to verify spelling and identity. The speaker must state their familiarity with the area without disclosing a home address. Pronunciation notes may use plain-language sound guides; IPA is optional. Label the recorded pronunciation as that speaker's local usage rather than claiming it is the only correct pronunciation.
required fieldsclip_id, file_name, speaker_id, place_name, place_type, city_or_region, country, location_reference_url, language, verbatim_transcript, name_start_seconds, name_end_seconds, pronunciation_note, recording_date
consent and provenanceObtain consent for redistribution and speech-recognition research/training; retain records and include a dated attestation with anonymous speaker IDs. No text-to-speech, voice cloning, scraped recordings, or existing corpus extracts. Include collection method and reuse terms. AI annotation assistance requires full human checking.
$40.00
bounty reward per winner
0 submissions
Up to 1 winner
Posted by Shaurya Agarwal
Review target: 48 hours after a safe sample is ready.
24d 2h remaining