Update location hierarchy to include new Overture L2 categories like Farmers Markets. Ensure that all OSM and OM points of interest are being categorized, even if only to a catch-all category.

Author: njhenry

config.yaml

@@ -5,8 +5,8 @@ versions:
   snapshot_osm: "20260313"
   snapshot_overture: "20260313"
   snapshot_foursquare: "20260313"
-  aws: "20260313"
-  conflation: "20260316"
+  aws: "20260318"
+  conflation: "20260318"
 
 # Settings for downloading data
 download:

scripts/build_taxonomy.py

@@ -40,21 +40,29 @@ def osm_cell(group):
 
 
 def overture_cell(group):
-    """Format Overture categories for one shared_label, grouped by l0."""
-    by_l0 = {}
+    """Format Overture categories for one shared_label."""
+    parts = []
     for _, row in group.iterrows():
         l0 = row["overture_l0"]
-        l1 = row["overture_l1"] if pd.notna(row["overture_l1"]) else None
-        by_l0.setdefault(l0, set())
-        if l1:
-            by_l0[l0].add(l1)
-    parts = []
-    for l0, l1s in by_l0.items():
-        if l1s:
-            l1_str = ", ".join(sorted(l1s))
-            parts.append(f'<span class="tx-key">{l0}:</span> {l1_str}')
+        l1 = row.get("overture_l1", "")
+        l2 = row.get("overture_l2", "")
+        if l1 and l2:
+            parts.append(
+                f'<span class="tx-key">{l0} &rsaquo;'
+                f' {l1}:</span> {l2}'
+            )
+        elif l1:
+            parts.append(
+                f'<span class="tx-key">{l0}:</span> {l1}'
+            )
+        elif l2:
+            parts.append(
+                f'<span class="tx-key">{l0}:</span> {l2}'
+            )
         else:
-            parts.append(f'<span class="tx-key">{l0}</span>')
+            parts.append(
+                f'<span class="tx-key">{l0}</span>'
+            )
     return "<br>".join(parts)
 
 

scripts/conflation/README.md

@@ -36,8 +36,8 @@ OSM POIs are assigned categories using the filter key priority order (shop >
 healthcare > leisure > amenity). If the specific tag value is not in the
 crosswalk, a wildcard (`*`) fallback for that key is used (default 50m radius).
 
-Overture POIs are matched by (taxonomy_l0, taxonomy_l1), falling back to
-l0-only match when l1 is missing.
+Overture POIs are matched using a 4-tier cascade from most to least specific:
+(L0, L1, L2), then (L0, L2), then (L0, L1), then L0-only.
 
 ### 2. Spatial Candidate Search
 

scripts/conflation/conflate.py

@@ -98,7 +98,7 @@
     "conf_mean", "conf_lower", "conf_upper", "geometry",
 ]
 OVERTURE_MATCH_COLS = [
-    "overture_id", "taxonomy_l0", "taxonomy_l1",
+    "overture_id", "taxonomy_l0", "taxonomy_l1", "taxonomy_l2",
     "overture_name", "brand_name", "confidence", "geometry",
 ]
 
@@ -204,7 +204,7 @@ def _load_gdf(
     ]:
         if col in osm_gdf.columns:
             osm_gdf.drop(columns = col, inplace = True)
-    for col in ["taxonomy_l0", "taxonomy_l1"]:
+    for col in ["taxonomy_l0", "taxonomy_l1", "taxonomy_l2"]:
         if col in overture_gdf.columns:
             overture_gdf.drop(columns = col, inplace = True)
     gc.collect()

scripts/overture/download.py

@@ -23,7 +23,7 @@
 Output file:
     overture_snapshot.parquet — GeoParquet with ~7.2M US POIs
         Columns: overture_id, overture_name, taxonomy_l0, taxonomy_l1,
-        brand_name, confidence, geometry, source
+        taxonomy_l2, brand_name, confidence, geometry, source
 """
 from config_versioned import Config
 from openpois.io.overture import download_overture_snapshot

site/src/constants.js

@@ -66,7 +66,6 @@ export const CONFLATED_LABELS = [
   'Alternative Medicine',
   'Arcade',
   'Arts Venue',
-  'Assisted Living',
   'Bakery',
   'Bank',
   'Bar',
@@ -91,8 +90,8 @@ export const CONFLATED_LABELS = [
   'Discount Store',
   'Dog Park',
   'Dry Cleaning',
-  'Event Venue',
   'Eye Care',
+  'Farmers Market',
   'Fast Food',
   'Fitness Center',
   'Florist',

src/openpois/conflation/data/match_radii.csv

@@ -2,7 +2,6 @@
 Alternative Medicine,50
 Arcade,50
 Arts Venue,100
-Assisted Living,100
 Bakery,50
 Bank,50
 Bar,50
@@ -27,8 +26,8 @@ Dessert Shop,50
 Discount Store,50
 Dog Park,200
 Dry Cleaning,50
-Event Venue,50
 Eye Care,50
+Farmers Market,100
 Fast Food,50
 Fitness Center,50
 Florist,50

src/openpois/conflation/data/taxonomy_crosswalk_openstreetmap.csv

@@ -77,7 +77,7 @@ amenity,hospital,Clinic
 amenity,ice_cream,Dessert Shop
 amenity,kindergarten,Kindergarten
 amenity,library,Library
-amenity,marketplace,Other Shop
+amenity,marketplace,Market
 amenity,museum,Museum
 amenity,nightclub,Nightclub
 amenity,pharmacy,Pharmacy