Update services/web_sources.py

services/web_sources.py

@@ -8,6 +8,56 @@ from .chain_filters import is_chain
 UA = {"User-Agent": "HF-Space-Trip-Planner/1.0 (+web-sources)"}
 OVERPASS = "https://overpass-api.de/api/interpreter"
 
+# ------------ ヘルパ（非レジャー/閉館検出） ------------
+
+_NON_LEISURE_NEG = {
+    # 行政・公共
+    "区役所","市役所","県庁","都庁","役所","庁舎","合同庁舎","裁判所","税務署","警察署","消防署",
+    # オフィス/ビジネス
+    "本社","オフィス","事務所","本部","支社","支店","株式会社","財団法人","社屋","企業",
+    # 学術・医療（観光目的外）
+    "大学","研究所","病院","クリニック","高校","中学校","小学校",
+    # コンベンション/展示
+    "コンベンション","会議","展示場","見本市","フォーラム","国際フォーラム","カンファレンス","conference","exhibition"
+}
+_LEISURE_POS = {
+    # レジャー/商業の肯定表現
+    "美術館","博物館","水族館","動物園","公園","庭園","展望","遊園地","テーマパーク","市場","商店街",
+    "温泉","銭湯","スパ","サウナ","ミュージアム","ギャラリー","ショッピング","アミューズメント","アート",
+    "神社","寺","城","史跡","遺跡","mall","shopping","market","museum","gallery","aquarium","zoo","park","theme park",
+}
+def _text_has_any(s: str, keys: set[str]) -> bool:
+    s = (s or "").lower()
+    return any(k.lower() in s for k in keys)
+
+def _is_closed_text(s: str) -> bool:
+    # Wikipedia 抜粋の閉館/閉業/休館検出
+    return _text_has_any(s, {"閉館","閉業","休館中","廃止","permanently closed","closed in","defunct","abolished"})
+
+def _is_non_leisure_wiki(title: str, extract: str) -> bool:
+    # 非レジャー語が含まれ、かつレジャー肯定語が無い場合に除外
+    t = f"{title} {extract}"
+    return (_text_has_any(t, _NON_LEISURE_NEG) and not _text_has_any(t, _LEISURE_POS))
+
+def _is_osm_non_leisure(tags: dict) -> bool:
+    # OSMタグから非レジャー/オフィス系を弾く
+    if not isinstance(tags, dict):
+        return False
+    if tags.get("office"):  # office=*
+        return True
+    amenity = (tags.get("amenity") or "").lower()
+    building = (tags.get("building") or "").lower()
+    banned_amenity = {
+        "townhall","police","fire_station","embassy","courthouse","clinic","hospital",
+        "university","school","college","kindergarten","bank","post_office","conference_centre"
+    }
+    banned_building = {"office","public","civic","government"}
+    if amenity in banned_amenity:
+        return True
+    if building in banned_building:
+        return True
+    return False
+
 # ========== OSM（Overpass） ==========
 
 def _overpass(q: str) -> dict:
@@ -30,6 +80,9 @@ def _doc_from_osm(e: dict, tags: dict, extra_tags: List[str]) -> dict | None:
     name = tags.get("name") or tags.get("name:ja") or tags.get("name:en")
     if not name:
         return None
+    if _is_osm_non_leisure(tags):
+        return None
+    # チェーン/ファストフードは上位で判定（飲食系のみ）
     lat, lon = c
     title = name
     text = ""
@@ -39,7 +92,10 @@ def _doc_from_osm(e: dict, tags: dict, extra_tags: List[str]) -> dict | None:
         "source": "osm",
         "tags": extra_tags
     }
-    # opening_hours はそのままメモ。厳密解釈は別途（今回は昼/夜の枠で調整）
+    # 参考タグ（チェーン検出用）
+    for k in ("brand","operator","network","brand:wikidata","brand:wikipedia","amenity","building"):
+        if k in tags:
+            meta[k] = tags[k]
     if "opening_hours" in tags:
         meta["hours"] = str(tags["opening_hours"])
     return {"id": meta["osm_id"], "text": text, "meta": meta}
@@ -61,7 +117,6 @@ out center tags {limit};
         out = []
         for e in j.get("elements", []):
             tags = e.get("tags", {}) or {}
-            # fast_food を除外
             if tags.get("amenity") == "fast_food":
                 continue
             name = tags.get("name") or ""
@@ -74,21 +129,19 @@ out center tags {limit};
     except Exception:
         return []
 
-def _osm_sights(center: Tuple[float, float], radius_km: float, limit: int = 60) -> List[dict]:
+def _osm_sights(center: Tuple[float, float], radius_km: float, limit: int = 80) -> List[dict]:
     lat, lon = center
     radius_m = int(radius_km * 1000)
-    # ミュージアム/ギャラリー/神社仏閣/公園/展望/歴史/アート関連など
+    # レジャー/商業寄りを拡充（aquarium/zoo/theme_park/marketplace/stadium など）
     q = f"""
 [out:json][timeout:25];
 (
-  node["tourism"~"museum|gallery|attraction"](around:{radius_m},{lat},{lon});
-  way["tourism"~"museum|gallery|attraction"](around:{radius_m},{lat},{lon});
-  node["leisure"="park"](around:{radius_m},{lat},{lon});
-  way["leisure"="park"](around:{radius_m},{lat},{lon});
-  node["amenity"="place_of_worship"](around:{radius_m},{lat},{lon});
-  way["amenity"="place_of_worship"](around:{radius_m},{lat},{lon});
-  node["amenity"="arts_centre"](around:{radius_m},{lat},{lon});
-  node["amenity"="theatre"](around:{radius_m},{lat},{lon});
+  node["tourism"~"museum|gallery|attraction|aquarium|zoo|theme_park"](around:{radius_m},{lat},{lon});
+  way["tourism"~"museum|gallery|attraction|aquarium|zoo|theme_park"](around:{radius_m},{lat},{lon});
+  node["leisure"~"park|stadium|water_park|amusement_arcade|garden"](around:{radius_m},{lat},{lon});
+  way["leisure"~"park|stadium|water_park|amusement_arcade|garden"](around:{radius_m},{lat},{lon});
+  node["amenity"~"place_of_worship|arts_centre|marketplace|spa|public_bath"](around:{radius_m},{lat},{lon});
+  way["amenity"~"place_of_worship|arts_centre|marketplace|spa|public_bath"](around:{radius_m},{lat},{lon});
   node["historic"](around:{radius_m},{lat},{lon});
   way["historic"](around:{radius_m},{lat},{lon});
 );
@@ -99,22 +152,24 @@ out center tags {limit};
         out = []
         for e in j.get("elements", []):
             tags = e.get("tags", {}) or {}
+            if _is_osm_non_leisure(tags):
+                continue
             name = tags.get("name") or ""
-            # 劇場/映画館は候補に入れるが後段で上限をかける
             ex_tags: List[str] = []
-            tourism = tags.get("tourism")
-            amenity = tags.get("amenity")
-            leisure = tags.get("leisure")
-            historic = tags.get("historic")
+            tourism = tags.get("tourism"); amenity = tags.get("amenity"); leisure = tags.get("leisure"); historic = tags.get("historic")
 
             if tourism in ("museum","gallery"):
                 ex_tags = ["Culture","indoor","museum"]
-            elif leisure == "park":
-                ex_tags = ["Nature","outdoor","park"]
+            elif tourism in ("aquarium","zoo","theme_park"):
+                ex_tags = ["Leisure","outdoor",tourism]
+            elif leisure in ("park","garden"):
+                ex_tags = ["Nature","outdoor",leisure]
+            elif leisure in ("stadium","water_park","amusement_arcade"):
+                ex_tags = ["Leisure","outdoor",leisure]
             elif amenity == "place_of_worship":
                 ex_tags = ["Heritage","outdoor","temple"]
-            elif amenity in ("arts_centre","theatre","cinema"):
-                ex_tags = ["PerformingArts","indoor","performing_arts"]
+            elif amenity in ("arts_centre","marketplace","spa","public_bath"):
+                ex_tags = ["Leisure","indoor",amenity]
             elif tourism == "attraction" or historic:
                 ex_tags = ["Sightseeing","outdoor","attraction"]
             else:
@@ -133,7 +188,6 @@ def _wikipedia_places(center: Tuple[float, float], radius_km: float, limit: int
     lat, lon = center
     radius_m = int(radius_km * 1000)
     try:
-        # 近傍ページ検索
         r = requests.get(
             "https://ja.wikipedia.org/w/api.php",
             params={
@@ -147,7 +201,6 @@ def _wikipedia_places(center: Tuple[float, float], radius_km: float, limit: int
         if not gs:
             return []
         page_ids = [str(x["pageid"]) for x in gs if "pageid" in x]
-        # 抜粋取得
         r2 = requests.get(
             "https://ja.wikipedia.org/w/api.php",
             params={
@@ -164,8 +217,15 @@ def _wikipedia_places(center: Tuple[float, float], radius_km: float, limit: int
             if not p:
                 continue
             title = p.get("title") or "Unknown"
-            extract = (p.get("extract") or "")[:500]
-            # 座標は geosearch の方に入っている
+            extract = (p.get("extract") or "")[:700]
+
+            # 閉館/閉業/休館の除外
+            if _is_closed_text(extract):
+                continue
+            # 非レジャー（オフィス/行政/コンベンション等）の除外
+            if _is_non_leisure_wiki(title, extract):
+                continue
+
             match = next((x for x in gs if str(x.get("pageid")) == gid), None)
             if not match:
                 continue
@@ -182,17 +242,14 @@ def gather_web_docs(center: Tuple[float, float], interests: List[str], date: str
     """
     Web only / Hybrid 用の外部ソース収集。
     - OSM/Overpass: 独立系レストラン（チェーン除外）
-    - OSM/Overpass: 観光POI（博物館/公園/神社仏閣/歴史/劇場ほか）
-    - Wikipedia: 近傍の観光スポット解説
+    - OSM/Overpass: 観光/レジャーPOI（商業・娯楽を拡充）
+    - Wikipedia: 近傍の観光スポット解説（オフィス/行政/閉館ページは除外）
     """
-    # OSM
-    sights = _osm_sights(center, radius_km, limit=80)
-    restos = _osm_restaurants(center, radius_km, limit=50)
-
-    # Wikipedia（軽め）
-    wiki = _wikipedia_places(center, min(radius_km, 8.0), limit=20)
+    sights = _osm_sights(center, radius_km, limit=100)
+    restos = _osm_restaurants(center, radius_km, limit=60)
+    wiki = _wikipedia_places(center, min(radius_km, 8.0), limit=24)
 
-    # 連結＋重複排除（titleベース、小文字）
+    # 連結＋重複排除（title��文字キー）
     def key(d): return (d.get("meta", {}).get("title") or "").strip().lower()
     seen = set(); merged: List[dict] = []
     for d in sights + restos + wiki:
@@ -201,6 +258,4 @@ def gather_web_docs(center: Tuple[float, float], interests: List[str], date: str
             continue
         seen.add(k)
         merged.append(d)
-
-    # 興味があれば軽くブースト…は planner 側スコアに任せるためここではそのまま返す
     return merged