yore-vocabulary-llm-filter

What problem does it solve?

This Skill ensures that the Whisper speech-to-text model accurately transcribes specialized vocabulary, preventing common misinterpretations and improving downstream LLM comprehension.

Core Features & Use Cases

• Domain-Specific Vocabulary Enrichment: Identifies and isolates terms crucial for accurate transcription in technical or project-specific contexts.
• LLM-Aware Filtering: Differentiates between terms that are phonetically ambiguous for speech-to-text versus those that are merely formatting variations understandable by an LLM.
• Use Case: When transcribing technical discussions about AI models like 'bm25' or project names like 'yore', this skill ensures Whisper outputs the correct terms instead of common English words.

Quick Start

Run the yore-vocabulary-llm-filter skill to classify terms from a harvest file and review the results before updating the Whisper vocabulary.