Project 293: US English Conversational Speech Data Collection
Budget: $35 – $36 USD
We are seeking freelancers for a real-world conversational speech data collection project in English (US). Participants must be native speakers of American English and able to work with a partner either in-person or remotely via Zoom, and the pay rate is $36.00 USD/hr per person. This project involves capturing natural two-party conversations across Healthcare, Call Center, and Meeting domains to support advanced language research.
Key Requirements:
- Native speaker of American English (en-US)
- Participants are not required to be in the same room. Recording can be done via Zoom, but strict settings must be applied before recording
- Conversations must be natural, unscripted, and free of long pauses or overlapping speech
- Must be able to speak clearly, including domain-specific terms (medical or professional jargon)
- Participants must be adults (18+)
- Participants will choose preferred conversation length and type when completing the online form (Short 5–7 min, Medium 20–23 min, Long 40–45 min)
- Initial pilot includes three samples from three locales; these will be reviewed within the week. If approved, the project will expand to additional locales, with separate job postings per locale
Technical Details:
- File format: .flac or .wav
- Sample rate: 16 kHz (Healthcare & Meetings), 8 kHz (Call Center)
- Audio channel: Stereo (Speaker 1 left, Speaker 2 right)
- Conversation length: Short 5–7 min, Medium 20–23 min, Long 40–45 min
- Participants must submit metadata for each recording: locale, number of speakers, gender, domain, type of conversation
Environment: Quiet room with minimal background noise; natural sounds allowed but no disruptive noises
Domains:
- Healthcare: Doctor-patient, therapist-patient, dentist-patient, in-person or telemedicine
- Call Center: Sales, marketing, financial services, insurance, healthcare, hospitality, government
- Meetings: Employee-employee, sales, teacher-student, virtual or in-person
Additional Information:
- AI-generated voices are strictly prohibited
- All recordings must be human-made and authentic
- File requirements per domain and length (e.g., Healthcare Long = 8 files, Medium = 32 files, Short = 45)
- Each couple will choose length/type when filling out the online form; immediate selection is not required, but participants must be aware of the options
Rejection Criteria:
Recordings may be rejected if they:
- Do not meet the required audio format or sampling rate
- Are recorded remotely or with separate devices
- Are scripted, robotic, or unnatural
- Contain excessive background noise or distortion
- Misrepresent speaker accents or dialects
Disclaimer: Participants must meet all technical and linguistic requirements. Failure to meet these standards or using AI-generated voices will result in removal from the project.
Key Requirements:
- Native speaker of American English (en-US)
- Participants are not required to be in the same room. Recording can be done via Zoom, but strict settings must be applied before recording
- Conversations must be natural, unscripted, and free of long pauses or overlapping speech
- Must be able to speak clearly, including domain-specific terms (medical or professional jargon)
- Participants must be adults (18+)
- Participants will choose preferred conversation length and type when completing the online form (Short 5–7 min, Medium 20–23 min, Long 40–45 min)
- Initial pilot includes three samples from three locales; these will be reviewed within the week. If approved, the project will expand to additional locales, with separate job postings per locale
Technical Details:
- File format: .flac or .wav
- Sample rate: 16 kHz (Healthcare & Meetings), 8 kHz (Call Center)
- Audio channel: Stereo (Speaker 1 left, Speaker 2 right)
- Conversation length: Short 5–7 min, Medium 20–23 min, Long 40–45 min
- Participants must submit metadata for each recording: locale, number of speakers, gender, domain, type of conversation
Environment: Quiet room with minimal background noise; natural sounds allowed but no disruptive noises
Domains:
- Healthcare: Doctor-patient, therapist-patient, dentist-patient, in-person or telemedicine
- Call Center: Sales, marketing, financial services, insurance, healthcare, hospitality, government
- Meetings: Employee-employee, sales, teacher-student, virtual or in-person
Additional Information:
- AI-generated voices are strictly prohibited
- All recordings must be human-made and authentic
- File requirements per domain and length (e.g., Healthcare Long = 8 files, Medium = 32 files, Short = 45)
- Each couple will choose length/type when filling out the online form; immediate selection is not required, but participants must be aware of the options
Rejection Criteria:
Recordings may be rejected if they:
- Do not meet the required audio format or sampling rate
- Are recorded remotely or with separate devices
- Are scripted, robotic, or unnatural
- Contain excessive background noise or distortion
- Misrepresent speaker accents or dialects
Disclaimer: Participants must meet all technical and linguistic requirements. Failure to meet these standards or using AI-generated voices will result in removal from the project.