Principal Speech Data Linguist

This is a fully remote position, open to applicants in United States.

📋 Description

• Take ownership of the linguistic framework for Innodata's segmentation and transcription tasks across various languages, domains, and applications.

• Establish standards for transcription and segmentation, including style guides and annotation conventions.

• Implement and manage quality frameworks, such as rubrics, error taxonomies, adjudication processes, inter-annotator agreement, and large-scale human quality assurance.

• Oversee the complete quality lifecycle for transcription and segmentation outputs.

• Create workflows that involve human oversight for review, correction, and adjudication as the quality of Automatic Speech Recognition (ASR) progresses.

• Address challenging cases involving accented and dialectal speech, low-resource and multilingual audio, overlapping speech, domain-specific jargon, and challenging acoustic environments.

• Collaborate with the Speech & Audio Research Scientist to convert model objectives into actionable specifications and evaluate the impact of transcription methods on ASR, Text-to-Speech (TTS), and speech/content-understanding models.

• Train, calibrate, and mentor expert transcribers and reviewers.

• Represent Innodata's approach to transcription and segmentation to clients and leading research labs.

• Contribute to the development of methodologies and best-practice documentation.

⛳️ Requirements

• Significant industry experience (typically over 8 years) in transcription, segmentation, and speech-data quality assurance.

• A Bachelor's degree in linguistics, phonetics, computational linguistics, or a closely related discipline is mandatory.

• Strong knowledge base in phonetics, phonology, and sociolinguistics.

• Comprehensive understanding of how choices in transcription and segmentation influence speech and content-understanding models.

• Proficiency in phonetic transcription and the International Phonetic Alphabet (IPA).

• Practical experience with acoustic and phonetic analysis, including spectrograms, formants, pitch, prosody, and segment boundaries.

• Extensive experience in audio segmentation, utterance and turn boundaries, timestamping, speaker labeling, and diarization.

• Proficient with tools such as Whisper, commercial ASR engines like AssemblyAI, Deepgram, Rev, and Speechmatics, Montreal Forced Aligner, and ELAN.

• Competency in Python scripting for batch processing, quality assurance, and metrics such as inter-annotator agreement and Word Error Rate (WER).

• Familiarity with regular expressions and Praat scripting.

• Solid data-management skills encompassing preprocessing, quality checks, post-processing, validation, and report preparation.

• Multilingual proficiency and hands-on experience with accented, dialectal, and code-switched speech.

• Excellent written and verbal communication skills.

• Comfortable collaborating with research scientists and clients.

• Bonus: Awareness of responsible AI considerations in speech, including accent and dialect bias, privacy, and consent regarding voice data.

🏝️ Benefits

• Competitive salary and performance-based bonuses.

• Opportunities for professional growth and development.

• Comprehensive health and wellness benefits.

• Flexible work arrangements, including remote work options.

• Collaborative and innovative work environment.