ASR-Zhuoseveral Technology Co., Ltd.

Project Requirement Checklist

Language and accent

• Audio collection method

• Specific accent or region

Resource Requirements

• IP Region

• Personnel background

• Single-person output efficiency, etc.

Application Scenarios

• Scenarios in which AI products are specifically used

Audio Length

• Short audio (8-15 seconds)

• Long audio (30s-2min)

Unit of Measurement

• The audio length unit can be in hour, minute, or second.

• Audio length divided by raw length and effective length

Transcribe Rules

• The initial version will be updated as resources encounter more.

Annotation Platform

• Hardware equipment requirements

• Platform account issues, lagging, and other issues

• Statistical process performance dashboard

Audio Quality

• Audio source

• Single speaker or multi-speaker audio

• Valid audio length percentage

Acceptance

• Review frequency

• Acceptance criteria (WER)

• Batch data volume

• Rework process

Rules in Different Use Cases

Domain	Terminology	Speaker Identification	Formatting Guidelines	Non-verbal Cues	Punctuation	Verbatim Transcription	Background Noise
Call Centers and Customer Service	Industry-specific	Required	Standardized	Not emphasized	Standardized	Optional	Minimal impact
Transcription Services	Client-specific	Optional	Flexible	Client-specific	Flexible	Optional	Varied impact
Voice Assistants and Virtual Agents	Context-specific	Required	Flexible	Not applicable	Standardized	Optional	Minimal impact
Language Learning and Education	Pronunciation focus	Optional	Flexible	Not emphasized	Flexible	Optional	Minimal impact
Accessibility and Assistive Technologies	Inclusive language	Optional	Standardized	Captured as necessary	Standardized	Optional	Minimal impact
Market Research and Sentiment Analysis	Domain-specific	Optional	Flexible	Not emphasized	Flexible	Optional	Varied impact
Broadcast and Media	Industry-specific	Optional	Standardized	Optional	Standardized	Optional	Varied impact
Legal and Law Enforcement	Legal terminology	Required	Standardized	Not emphasized	Standardized	Optional	Minimal impact
Research and Data Analysis	Research-specific	Optional	Flexible	Not emphasized	Flexible	Optional	Varied impact
Conference and Event Transcription	Event-specific	Required	Flexible	Optional	Flexible	Optional	Varied impact
Speech-to-Text Applications	Application-specific	Optional	Flexible	Not applicable	Flexible	Optional	Varied impact

ASR

Audio annotation overview

We have extensive experience in providing ASR data annotation services for AI applications. Our experts have worked on numerous projects, accurately annotating large volumes of audio data. We transcribe speech into text with meticulous attention to detail, ensuring accuracy and consistency. With rigorous quality assurance measures, we verify the annotations to maintain precision. Our experience includes voice assistants, call center automation, and speech-to-text conversion, across multiple languages and accents. Leveraging advanced annotation tools, we ensure efficient and timely delivery of annotated ASR data. Our expertise in ASR data annotation enables us to meet clients' unique requirements for training and enhancing ASR models effectively.

ABOUT US

Typical audio annotation

Speech transcription

Speaker identification

Sound event detection and classification

Sentiment analysis

Emotion analysis

Language identification

Named entity recognition

Noise identification

Code

Switching