Conversation
Press releases are a distinct document type and should not be counted alongside judgments. Before this fix, build_pages_dashboard.py processed every record in the JSONL regardless of document_type, causing press releases to inflate total_cases, violation rates, country rankings, article breakdowns, chamber counts, outcome distributions, and every other metric that is meant to describe judicial outcomes only. Changes: - Add is_press_release() helper (mirrors search-app.js logic: document_type contains "press release", case-insensitive). - Skip press releases at the top of the build_payload() loop so they do not touch any judgment counter. - Track press release count separately; expose it as summary.total_press_releases so the KPI card in index.html can still display it. - total_cases now reflects judgments only; input_record_count preserves the original total (judgments + press releases) for data-provenance purposes. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Rebuilt from full dataset (18,429 records): 13,480 judgments, 4,949 press releases excluded. Violation rate now 77.8% on judgments only; total_press_releases added to summary. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add docs/about.html with authors (Łukasz Szoszkiewicz, Sebastian Marcisz), data sources table with direct link to ECTHR-PCR dataset on HuggingFace, dataset field reference, methodology, and acknowledgements. - Add scripts/merge_ecthr_pcr.py to enrich the ECHR JSONL with structured citation-network data from RashidHaddad/ECTHR-PCR (15,729 cases, joined on appno). Adds pcr_citations, pcr_cited_by, pcr_citation_count, pcr_cited_by_count fields. Original dataset is left untouched; output goes to a new file. - Add "About" nav link to index.html and analytics.html. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
lszoszk
added a commit
that referenced
this pull request
May 11, 2026
Searching for a case name (e.g. "KRASNOSHAPKA") returned 38 cases that
cite the source case before the actual KRASNOSHAPKA v UKRAINE judgment
itself. Cause: SQLite FTS5 BM25's field-length normalisation lets
repeated body matches from Pop C committee cases (each citing the case
once) outscore the short-title source case even with title weight 5.0.
The existing comment in main.py already noted the same pathology on
"Hirst" — max*ln dampens it but isn't enough for proper-name queries.
Fix:
- `ranking.title_matches_query(title, query_terms)` — case-insensitive
whole-word check, ignores stopwords and tokens shorter than 4 chars
("L.P." case-name searches stay unboosted because both letters are
too short to anchor on).
- `ranking.TITLE_MATCH_BOOST = 3.0` multiplicative boost applied in
`compute_final_score` when any query term hits the case title.
- Search SQL now selects `c.title` in the per-case aggregate so the
rerank pass can tag each candidate with `title_match=True/False`.
Verified on VM (post `docker compose build`):
Query Before After
───────── ────────────────────────────── ──────────────────────────────
KRASNOSHAPKA KRASNOSHAPKA v UKRAINE @ #39 KRASNOSHAPKA v UKRAINE @ #1 (38.7)
Hirst HIRST v UK (No. 2) @ #1 already HIRST v UK (No. 2) @ #1 (60.1)
Mirzoyan MIRZOYAN v ARMENIA @ #1 already MIRZOYAN v ARMENIA @ #1 (41.4)
Grzęda GRZĘDA v POLAND @ #1 GRZĘDA v POLAND @ #1 (68.8)
torture no title-match cases, unchanged no regression (substantive on top)
margin of … no title-match cases, unchanged no regression
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
lszoszk
added a commit
that referenced
this pull request
May 17, 2026
…n metrics #1 — HUDOC fidelity. The Case Note context window now groups consecutive physical rows that share a logical paragraph (a numbered ¶ plus its indented quote / continuation rows) into one block: the "¶ N" number shows once, with quote rows rendered as an indented sub-block — the way HUDOC presents a paragraph with a quotation inside it. Previously a paragraph that quoted e.g. Article 8 rendered as three separate "¶ 74" rows, which read as a numbering error. getCaseNoteContext now windows in logical paragraphs; "show earlier/later" steps by logical ¶. #3 — trust. The citation graph (P29) has partial coverage, so a 0 is ambiguous ("uncited" vs "no data"). The INFLUENCE bars (cites / cited by) and the Case Note "Citations" field now render 0 as a dimmed "—" with an explanatory tooltip, instead of asserting a misleading zero on landmark cases. The hit count stays a real number. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.