Nexa

Description:  

Participants will perform speech transcription and segmentation annotation. They convert audio signals into accurate textual representations to support the training of machine intelligence transcription systems.  

Transcription tasks include: 

• Transcribing audio signals into text accurately and consistently, following defined transcription guidelines.  

• Transcribing not only speech data but also relevant non-speech sounds such as ringing phones, background music, and other audible events where required.  

• Ensuring high-quality transcription suitable for training AI and machine learning models.  

• Applying correct speaker labels and language labels for speech segments. 

Segmentation tasks include:  

• Manually segmenting audio files by timestamping according to defined structural boundaries.  

• Identifying and creating segments based on the four primary segment types (Speech, Babble, Music, Noise) 

• Marking conversational turns, utterances, phrases, and sound-type boundaries accurately.  

• Segmenting entire audio files prior to transcription where applicable, to improve efficiency and accuracy.  

• Ensuring segmentation enables manageable listening chunks for effective transcription. 

Purpose:  

To create high-quality training data for speech technology systems by converting spoken content into accurately formatted written transcripts that capture all linguistic nuances and conventions.   

Main Requirements: 

• Fluency in any one of the 5 languages: Akan, Hmong, Lao, Lu Mien and Mizo with strong listening and writing skills.  

• Excellent attention to detail, grammar, vocabulary, and contextual understanding. 

• Strong attention to detail 

• Minimum availability of 10 hours per week. 

• Ability to work independently and manage deliverables within deadlines. 

• Ability to adhere strictly to project parameters, guidelines, and quality standards.  

• Ability to maintain consistency across transcription and segmentation tasks.  

• Ability to review work to ensure accuracy and completeness before submission. 

• Prior experience in transcription, speech data annotation, or AI data labeling is a plus. 

• Familiarity with speaker labeling and timestamp-based segmentation is a plus.⁠  

Important:

If you refer someone who speaks Iu Mien, Hmong, Lao or Mizo, and they successfully apply to the study, you can earn an extra $10.  

Use the following link to start referring your friends and family: https://oneforma.aidaform.com/referral-form-nexa-project