Japanese Audio Verification & Text Matching – Data Quality Labeling

We are seeking meticulous Native-level Japanese speakers or highly proficient bilinguals to assist in an audio-to-text data verification project. This role requires extreme attention to detail and high-quality listening skills to help train AI speech recognition models. Project Overview: You will be provided with short Japanese audio files and their corresponding text transcripts. Your primary task is to listen to the audio and evaluate whether the text accurately represents what is being said, while also flagging technical audio issues and speaker overlaps. Responsibilities: You will follow a strict labeling guideline for every task: Speech-to-Text Matching: Determine if the audio perfectly matches the text. - You will mark "NO" if there is even a minor discrepancy (e.g., the first/last word is slightly cut off or if you hear a fragment of a different dialogue). - If there is a mismatch, you must categorize the reason (Word cutoff, neighboring dialogue, word count mismatch, or different words used). Speaker Count & Overlap: - Identify if there is more than one speaker (even if the second speaker is barely audible or from a neighboring event). - Assess if the speech overlaps or if multiple speakers have significant dialogue versus just a small portion at the edges. Audio Quality Evaluation: - Rate the quality based on audibility and distortion (Clear quality vs. muted/distorted vs. poor/unclear). Requirements: - Language: Native or near-native Japanese proficiency. - Attention to Detail: Ability to detect "tiny bits" of noise or speech at the start or end of clips. - Equipment: High-quality headphones/earphones are mandatory for this task. - Reliability: Ability to follow strict technical guidelines without deviation. - Experience: Prior experience in Data Labeling, Transcription, or Translation is a plus.