`press_releases`

Appropriations press releases

One row per item in one appropriations committee press-release feed capture. bill_id is an inferred candidate when a pattern matches a published bill identifier in the title or description. All columns are stored as VARCHAR.

Coverage. Window. Releases from the House and Senate Appropriations committees' feeds as captured since September 19, 2026, including items that later rotated out of a feed; bill links are inferred from release text. Not all committees, earlier history or release-page bodies. See press-release-qualification/ and press-link-repair/ in the fork execution receipts. (measured 2026-09-28)

Data quality. The bill linkage is a pattern match over the release's own words, not a publisher statement. An earlier local run falsely matched s 2027 inside a reference to the President's 2027 budget to Senate bill 2027. The corrected matcher rejects possessive suffixes while preserving quoted bill citations; wider matching quality remains unmeasured. matched_field and matched_text carry what was matched and where: a false positive is readable from the row rather than only by re-running the matcher. One pattern is compiled per published bill in the Congresses in scope, bounded on both sides and built from the bill's own type, so a bare number matches nothing. Three states are distinct and must not be read as one: a bill_id with match_rule = bill_number_in_title or bill_number_in_excerpt is a match; match_rule = unmatched with a NULL bill_id means the matcher ran and found no pattern match; an all-NULL match_rule means no matching pass ran at all, because no congress_bills table was published when the rollup ran, or none was published for that release's Congress, or the release states no readable pub_date to scope by. The Senate feed carries no item description, so a Senate release can only ever match on its title. One consequence worth knowing before relying on these four columns: a run that cannot read congress_bills republishes the whole live window with NULL match columns, because this table merges row-wise and a fresh row wins whole. The next run that can read it re-derives them for every release still in the feed, so the gap normally lasts one cycle — but a release that rotates off the feed during that cycle keeps the NULL permanently. The other publisher columns are unaffected; only these four degrade.

Parquet file: press_releases.parquet
MCP query_sql support: Configured; requires an available artifact.
Publication status: Not established by this schema page or its measurement date.
Row count: Not stated here; the MCP describe_table reply gives the live count under publication.

Column	Type	Description
`release_id`	`VARCHAR`	Digest of the chamber and the item's full link; replaces a truncated-URL unique key.
`chamber`	`VARCHAR`	Which committee feed this item came from.
`feed_url`	`VARCHAR`	The feed address the capture requested.
`channel_title`	`VARCHAR`	The channel title, which is also the proof the requested feed answered.
`channel_link`	`VARCHAR`	The channel link, checked against the publisher's own host.
`channel_description`	`VARCHAR`	The channel description as the publisher wrote it.
`channel_language`	`VARCHAR`	The channel language tag.
`channel_copyright`	`VARCHAR`	The channel copyright line, which only the Senate feed states.
`channel_docs`	`VARCHAR`	The channel docs URL, which only the Senate feed states.
`channel_last_build_date`	`VARCHAR`	The channel lastBuildDate, which only the Senate feed states.
`channel_ttl`	`VARCHAR`	The channel ttl: the publisher's own polling contract.
`channel_skip_days`	`VARCHAR`	The channel skipDays, unit-separator joined; NULL where the channel states none.
`channel_skip_hours`	`VARCHAR`	The channel skipHours, unit-separator joined; NULL where the channel states none.
`item_index`	`VARCHAR`	Zero-based position of this item in the captured channel.
`title`	`VARCHAR`	The item title in full, never truncated to a display length.
`link`	`VARCHAR`	The item's full link.
`guid`	`VARCHAR`	The item's guid, where it carries one.
`guid_is_permalink`	`VARCHAR`	The guid's isPermaLink, defaulting to true when the attribute is absent.
`description`	`VARCHAR`	The item's raw HTML description exactly as the publisher wrote it, untruncated.
`description_text`	`VARCHAR`	That description read as plain text, not a truncated excerpt.
`description_chars`	`VARCHAR`	Character length of description_text.
`pub_date`	`VARCHAR`	The item's pubDate exactly as the publisher spelled it, wrong zone abbreviation included.
`pub_date_instant`	`VARCHAR`	That pubDate parsed to an instant under RFC 822's fixed abbreviation table. A zone label is taken at its word: the Senate feed stamps `EST` all year, so a summer release reads -05:00, an hour off its clock time.
`author`	`VARCHAR`	The item's author, which the Senate feed states as a shared mailbox.
`creator`	`VARCHAR`	The item's dc:creator, which the House feed states as a named staffer.
`categories_json`	`VARCHAR`	Every category the item lists, as a JSON array.
`enclosure_url`	`VARCHAR`	The enclosure URL, where the item carries one.
`enclosure_length`	`VARCHAR`	The enclosure length in bytes, where the item states one.
`enclosure_type`	`VARCHAR`	The enclosure media type, where the item states one.
`observed_at`	`VARCHAR`	When the feed was captured; the merge prefers the larger value.
`bill_id`	`VARCHAR`	The bill this release names, where a bill-number pattern matched.
`match_rule`	`VARCHAR`	Which release-matching rule fired: a rule's name where it found a bill, `unmatched` where the matching pass ran and found none; NULL where no matching pass ran, which says nothing about whether the release names a bill.
`matched_field`	`VARCHAR`	Which field the mention was found in; the Senate feed can only ever match on title.
`matched_text`	`VARCHAR`	The exact text that matched, so a false positive is readable from the row.