Text Corpus Analysis

Skills for analyzing large text corpora — topic modeling (BERTopic with temporal evolution), NER, categorization into fixed taxonomies, bottom-up category derivation, multi-level taxonomy design, word frequency, synonym clustering for voice-note/STT corpora, parametric stats, and metadata↔content correlation. Three execution lanes (classical NLP, local LLM via Ollama, cloud LLM via OpenRouter) with explicit cost-awareness: mandatory pre-run estimates for >1k-doc LLM passes, two-pass cheap→premium pattern, embeddings+clustering preferred over pairwise LLM comparison.