Loading the data dictionary from the Hub…
Before you use a column
- Check coverage. Several fields are well under 50 % populated and the gaps are not random — they follow eras of the show. Empty for the back catalogue means missing, not zero.
- Null is not zero. Where a null carries meaning, the null means line says what.
- Two models are ~93 % of the corpus. “Over time” and “by model” are nearly the same axis and mostly cannot be separated.
- Episode length moved — ~3,000 words (2025) to ~4,200 (2026). Most repetition metrics move with length; control for it before reading a trend.
- The dialogue is synthetic. A record of what the show said — not human speech, not a factual reference. Listener prompts are not published.