goon-foss/goon
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

fix(hqporner): require ALL query tokens in slug — stop performer over-attribution

Browse source 
hqporner search post-filter kept a scene if its slug contained ANY query token
(>=3 chars). For multi-word performer names this matched on a single common token
(e.g. "anna","mia"), so the performer-driven ingest attributed the scene to EVERY
performer sharing that token — scenes accumulated up to 503 wrong performers
(hqporner = 5659 of 5897 scenes with >30 performers; bug-reports 2026-06-07).

Switch ANY->ALL: the slug must contain every query token, requiring a full name
match before attribution. Single-word names still work. Precision over recall —
144 wrong performers is far worse than missing a few loose matches.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
jtrzupek 2026-06-08 09:28:18 +02:00
parent bc72515227
commit cd257740be
1 changed files with 7 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

10
app/connectors/direct_scrapers/hqporner.py
View file

@ -47,8 +47,12 @@ class HQPornerScraper(BaseDirectTubeScraper):
log.debug("hqporner search %s status=%d", url, r.status_code) log.debug("hqporner search %s status=%d", url, r.status_code)
return return
# Filtr: slug musi zawierać przynajmniej jedno z słów query (case-insensitive) # Filtr: slug musi zawierać WSZYSTKIE słowa query (≥3 znaki), case-insensitive.
# Eliminuje totalnie niezwiązane wyniki gdy fuzzy search szumi. # Wcześniej `any` (≥1 token) → przy 2-słownych nazwach match na jednym pospolitym
# tokenie (np. "anna"/"mia") atrybutował scenę do KAŻDEGO performera dzielącego ten
# token → sceny z setkami błędnych aktorek (do 503; hqporner = 5659/5897 takich scen,
# bug-report 2026-06-07). `all` wymaga pełnego dopasowania nazwy → precyzja.
# Pojedyncze nazwy ("Belladonna") nadal działają (jeden token musi być).
query_tokens = {tok for tok in query.lower().split() if len(tok) >= 3} query_tokens = {tok for tok in query.lower().split() if len(tok) >= 3}
seen_urls: set[str] = set() seen_urls: set[str] = set()
@ -63,7 +67,7 @@ class HQPornerScraper(BaseDirectTubeScraper):
# Title-token filter # Title-token filter
slug_lower = slug_part.lower() slug_lower = slug_part.lower()
if query_tokens and not any(tok in slug_lower for tok in query_tokens): if query_tokens and not all(tok in slug_lower for tok in query_tokens):
continue continue
title = slug_part.replace("_", " ").replace("-", " ").strip() title = slug_part.replace("_", " ").replace("-", " ").strip()