Browse categories
Explore
Fiverr Pro
English
$
USD
I will transcribe and annotate mandarin audio for ai datasets
About this Gig
Prepare Mandarin audio for AI datasets with human-reviewed transcripts and structured labels.
I am a native Mandarin speaker and a Big Data and AI Engineer, combining Chinese-language understanding with structured data preparation.
I transcribe clear Mandarin, label up to 2 speakers, add turn-level timestamps, and mark unclear or overlapping speech without guessing. Deliverables: TXT and CSV; JSON in Standard and Premium.
Basic: 2 audio minutes, up to 100 segments. Standard: 5 minutes, up to 200 segments, plus up to 5 client-defined intent labels. Premium: 10 minutes, up to 300 segments, plus observable AI dialogue issues such as irrelevant replies or clear transcript errors. Both duration and segment limits apply.
Provide audio, labeling guidelines and terminology. Send a sample first for noise, accents or specialist topics. Word-level alignment, translation, voice recording and commercial dubbing are not included.
Revisions correct work within the original audio and agreed rules. New files or changed labels need a new scope. Share only authorized audio and remove confidential details.
Technique:
Manual
Tagging type:
Text
•
Audio
FAQ
Do you record or dub voices?
This Gig annotates your existing audio. Human voice recording and commercial dubbing are not included.
How precise are timestamps?
Timestamps mark speaker turns. Word-level forced alignment is not included. Both audio duration and segment limits apply.
Can you label intent or review AI replies?
Standard and Premium support up to 5 agreed intent labels. Premium adds issues observable from supplied audio and context; it does not guarantee complete fact verification.
Can you handle noisy recordings or dialects?
Send a short sample first. Packages cover clear Mandarin with up to 2 speakers. Unclear sections are marked rather than invented.
