# Southwest Mushrooms Cultivation Corpus

Spoken-word transcripts from the Southwest Mushrooms YouTube channel, a working
commercial gourmet and medicinal mushroom operation in Phoenix, Arizona, run by
Michael Crowe. Covers lab work, grain spawn, substrate preparation,
sterilization, fruiting conditions, harvest, and the business of growing.

This is field material, not textbook material. It is what was actually said
while doing the work.

## Contents

| File | Records | Description |
|---|---|---|
| `corpus.jsonl` | 68 | One record per video, full transcript |
| `chunks.jsonl` | 600 | ~1200 character passages with deep-link timestamps |
| `corpus.csv` | 68 | Flattened, for spreadsheets |

## Statistics

- Videos: **68**
- Words: **131,872**
- Runtime: **15.4 hours**
- Combined views on source videos: **10,052,363**
- Date span: **20200607 to 20260721**

## Schema, corpus.jsonl

`video_id`, `url`, `title`, `upload_date`, `duration_sec`, `view_count`,
`like_count`, `description`, `tags`, `captions_source`, `word_count`, `transcript`

## Schema, chunks.jsonl

`chunk_id`, `video_id`, `title`, `start_sec`, `url`, `text`

Each chunk `url` deep-links to the exact second in the source video, so any
answer built on this corpus can cite the moment it came from.

## Caveats

- `captions_source` is measured, not assumed: 50 records are automatic
  speech recognition and 18 are human-written. Classification uses
  sentence-mark density on the parsed text, because filenames do not
  distinguish the two and raw VTT timestamps contain periods that corrupt the
  measurement. ASR records make errors and are least reliable on species names
  and Latin binomials, so treat those spellings as approximate.
- Timestamps come from the caption track and are accurate to the cue, not the word.
- Rolling-buffer duplication in auto-captions has been removed. Without that
  step the text is roughly twice as long and repeats every line.

## Provenance and license

Source: https://www.youtube.com/@southwestmushrooms
Compiled: 2026-07-26

Transcripts are of original spoken content owned by Southwest Mushrooms.
Set the license deliberately before publishing this anywhere public.