French Speech Annotation & Evaluation Specialist (TTS Direction)

Position Overview

We are looking for a meticulous and dedicated French Speech Annotation & Evaluation Specialist to join our team in developing next-generation high-quality French Text-to-Speech (TTS) large models. You will be responsible for text alignment, phonetic and prosodic annotation of French audio data, and quality evaluation of synthesized speech. Your work will directly impact the naturalness and emotional expressiveness of our French TTS models.

Key Responsibilities

1. Audio Data Annotation (Data Construction Phase)

• Text Verification & Refinement: Listen to professional French recordings (or synthesized audio), verify and correct mismatches between text and audio, and accurately annotate spelling, punctuation, and special symbols.
• Prosodic & Pause Annotation: Based on auditory perception, annotate pauses, stress, intonation patterns, and speech rate variations within the audio.

2. Audio Data Evaluation (Model Assessment Phase)

• Subjective Listening Evaluation: Conduct multi-dimensional subjective scoring (e.g., standard MOS evaluation) on synthesized French audio generated by the model, assessing pronunciation accuracy, naturalness, fluency, and emotional expression appropriateness.
• Defect Analysis & Categorization (Bad Case Analysis): Accurately identify defects in synthesized audio—such as missing phonemes, electronic noise/artifacts, unnatural intonation, hallucinated sounds—and tag and categorize them according to specifications, providing clear optimization feedback to algorithm engineers.

Qualifications

1. Language & Professional Background

• Language Proficiency: Native French speaker, or graduate of a French language program in China with TFS-8 (Test de Français Spécialisé - Niveau 8) or European C1/C2 certification. Standard pronunciation with minimal regional accent (Metropolitan French/Parisian French preferred).
• Professional Foundation: Background in Linguistics, French Language, Acoustics, or Phonetics preferred. Familiarity with French pronunciation rules, with deep theoretical understanding and keen auditory perception of sandhi phenomena (liaison, elision, etc.).

2. Skills & Tools

• Phonetic Transcription: Ability to accurately perform phoneme-level error correction.
• Tool Proficiency: Proficient in standard office software. Experience with audio processing tools (e.g., Audacity, Praat) preferred (on-the-job training available for candidates with strong learning ability and no prior experience).

3. Work Ethic & Environment

• Work Attitude: Exceptional patience, meticulousness, and concentration. Annotation work is highly repetitive, requiring sustained high accuracy over extended periods.
• Auditory Sensitivity: Acute hearing with high sensitivity to audio details (e.g., subtle electronic noise, missing phonemes, background noise).
• Work Equipment: For remote/part-time positions, candidates must have a quiet work environment and must provide their own high-quality studio-monitor headphones (over-ear preferred; in-ear or standard noise-canceling headphones are not acceptable, as they may mask subtle audio defects).

Preferred Qualifications

• Previous experience in TTS (Text-to-Speech) or ASR (Automatic Speech Recognition) data annotation.
• Previous experience in subjective MOS evaluation for speech synthesis.

Work Arrangements

• Work Mode: Remote / Part-time / Freelancer / Full-time on-site options available.
• Working Hours: Flexible working hours; compensation is settled in phases based on delivery quality and progress. xuezdbg
• Training Support: Professional training on TTS annotation standards and evaluation criteria will be provided, with technical team support available in real time.