Add OCR text for scanned House PTR filings

trades/politician/README.md

@@ -51,7 +51,7 @@ One row per transaction, parsed from 5,966 electronically filed House PTR PDFs (
 | source_pdf_url | Original PDF |
 | ocr, ocr_confidence | False / blank for e-filed rows (reserved for scanned filings) |
 
-The 2,457 scanned paper filings (document IDs starting with 8 or 9, mostly 2014–2019) are not in the trade table yet. They're listed in `house_ptr_unparsed_filings.parquet`, and their OCR text will be added in a later update.
+The 2,457 scanned paper filings (document IDs starting with 8 or 9, mostly 2014–2019) are not in the trade table yet. They're listed in `house_ptr_unparsed_filings.parquet`, and their machine OCR text (one row per filing, with `ocr=true` and a mean `ocr_confidence`) is in `house_ptr_paper_filings_ocr_text.parquet`. Scanned forms were handwritten or typed on rotated grids, so they are not split into trade rows.
 
 ## congress_trades.parquet (66,865 rows)
 Senate and House trades combined, ready for backtests. Columns are the same as above (`member_name` replaces `senator_name`), and the `chamber` column tells the two apart.

trades/politician/house_ptr_paper_filings_ocr_text.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84d150d71c20714600cae23b0cef245187bf4cb8d4638491aa298ffa12e59a27
+size 1810497

trades/politician/house_ptr_unparsed_filings.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:38d38bede83bf71d3578cc2a404c99f1a33d61f768f948252fbd2c8349018d7b
-size 39188
+oid sha256:4c6df812b1ea2d89bc5e299dfc0a2caeda48b076b074b82a6c43a8e4909847bf
+size 39135