Improving Speech Recognition Accuracy Using Custom Language Models with the Vosk Toolkit

TL;DR

This work tackles the challenge of accurate speech-to-text transcription across diverse audio formats in offline settings. It presents a Python-based pipeline built on the Vosk toolkit with domain-adapted language models to support WAV/MP3/FLAC/OGG and output to DOCX. Empirical results show that domain-specific language models reduce word error rate across technical, educational, business, and media domains, supported by visual diagnostics to understand recognition dynamics. The approach offers a cost-effective, privacy-preserving offline solution with potential for real-time expansion and broader automation applications.

Abstract

Although speech recognition algorithms have developed quickly in recent years, achieving high transcription accuracy across diverse audio formats and acoustic environments remains a major challenge. This work explores how incorporating custom language models with the open-source Vosk Toolkit can improve speech-to-text accuracy in varied settings. Unlike many conventional systems limited to specific audio types, this approach supports multiple audio formats such as WAV, MP3, FLAC, and OGG by using Python modules for preprocessing and format conversion. A Python-based transcription pipeline was developed to process input audio, perform speech recognition using Vosk's KaldiRecognizer, and export the output to a DOCX file. Results showed that custom models reduced word error rates, especially in domain-specific scenarios involving technical terminology, varied accents, or background noise. This work presents a cost-effective, offline solution for high-accuracy transcription and opens up future opportunities for automation and real-time applications.

Figures (4)

• Figure 1: Speech-to-Text Transcription Workflow Using Vosk Toolkit
• Figure 2: Distribution of Audio Formats in Test Dataset
• Figure 3: WER Comparison Across Domains
• Figure 4: Recognition Accuracy Trend