Glossary Library Configuration Guide

In specialized fields like medicine, law, engineering, and finance, general speech recognition engines often mishear industry terms, acronyms, or homophones.

DuRT 3.0 provides a dedicated Glossary Library to eliminate these errors before and after transcription.

---

Core Mechanism

flowchart LR
    A["Raw Audio Input"] --> B["ASR Speech Recognition"]
    B --> C["Raw Output Text<br>(e.g. 'Deep Seek', 'dag pipeline')"]
    D["Glossary Mapping Table<br>{ Deep Seek ➔ DeepSeek }<br>{ dag pipeline ➔ DAG Pipeline }"] --> E["Smart Correction Engine"]
    C --> E
    E --> F["Standardized Subtitles<br>(e.g. 'DeepSeek', 'DAG Pipeline')"]

1. Prompt Bias & Post-Processing: Injects specialized vocabulary into model prompts and performs automated string replacements upon transcript completion;
2. Multi-Domain Groups: Maintain distinct glossaries for different subjects (e.g. "Medical Terminology", "Software Architecture");
3. Batch Import & Export: Import and export glossaries via CSV, JSON, or TSV formats.

---

Setup in DuRT

1. Open DuRT, go to Settings ➔ Knowledge & Assets ➔ click Glossary Library;
2. Click "New Glossary";
3. Add mapping entries:
   • Original / Misspelled: Expected common phonetic mistake;
   • Target / Replacement: Correct standardized term;
4. Save the glossary. It will be immediately active for Subtitle Studio batch corrections and DAG workflow steps!