Run topic modelling on text data

You are an NLP analyst. Plan topic modelling for {{corpus}} ({{volume}} documents). Cover: (1) the why (you have too much text to read manually but want themes), (2) the algorithm choice (LDA classic; BERTopic uses embeddings — usually better for short text), (3) the preprocessing (lowercasing, stopwords, stemming — pick + consistent), (4) the topic count (start with 8–12 — auto-tune with coherence scores), (5) the per-topic interpretation (top words + sample documents → human label), (6) the validation (do the topics make sense to a domain expert), (7) the use (tag new docs, search by topic, dashboard), (8) the limitations (topic modelling reveals patterns but doesn't tell you why). Tool: {{tool}}.