acuity-framework 3.0.0__tar.gz → 3.1.0__tar.gz

{acuity_framework-3.0.0 → acuity_framework-3.1.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: acuity-framework
-Version: 3.0.0
+Version: 3.1.0
 Summary: ACUITY — A machine learning framework for extracting, verifying, and recommending local micro-enterprise profiles from unstructured community posts.
 Author: ACUITY Research Team
 License: MIT

{acuity_framework-3.0.0 → acuity_framework-3.1.0}/acuity/utils.py

@@ -82,3 +82,51 @@ def levenshtein_details(s1: str, s2: str) -> dict:
         return {"score": 1.0, "edits": 0, "max_len": 0}
     edits = distance[len(s1)][len(s2)]
     return {"score": 1.0 - (edits / max_len), "edits": edits, "max_len": max_len}
+
+import re
+
+def _tokenize(s: str) -> list[str]:
+    return re.findall(r'\w+', str(s).lower())
+
+def token_sort_ratio(s1: str, s2: str) -> float:
+    t1 = _tokenize(s1)
+    t2 = _tokenize(s2)
+    t1.sort()
+    t2.sort()
+    return levenshtein_ratio(' '.join(t1), ' '.join(t2))
+
+def token_set_ratio(s1: str, s2: str) -> float:
+    t1 = set(_tokenize(s1))
+    t2 = set(_tokenize(s2))
+    
+    intersection = sorted(list(t1.intersection(t2)))
+    diff1 = sorted(list(t1.difference(t2)))
+    diff2 = sorted(list(t2.difference(t1)))
+    
+    str_intersection = ' '.join(intersection)
+    str1 = ' '.join(intersection + diff1).strip()
+    str2 = ' '.join(intersection + diff2).strip()
+    
+    score1 = levenshtein_ratio(str1, str2)
+    score2 = levenshtein_ratio(str_intersection, str1) if str_intersection else 0.0
+    score3 = levenshtein_ratio(str_intersection, str2) if str_intersection else 0.0
+    
+    return max(score1, score2, score3)
+
+def hybrid_fuzzy_match(s1: str, s2: str) -> float:
+    if not s1 or not s2:
+        return 0.0
+        
+    plain_score = levenshtein_ratio(s1.lower(), s2.lower())
+    sort_score = token_sort_ratio(s1, s2)
+    set_score = token_set_ratio(s1, s2)
+    
+    # Apply penalty to Token-Set if length disparity is massive (to prevent short acronym false positives)
+    len1, len2 = len(s1), len(s2)
+    if len1 > 0 and len2 > 0:
+        ratio = min(len1, len2) / max(len1, len2)
+        if ratio < 0.35:
+            # Heavily penalize the Token-Set score
+            set_score = set_score * ratio
+            
+    return max(plain_score, sort_score, set_score)

{acuity_framework-3.0.0 → acuity_framework-3.1.0}/acuity/verification/bplo.py

@@ -14,7 +14,7 @@ import csv
 from typing import Any
 
 from ..config import AcuityConfig
-from ..utils import levenshtein_ratio
+from ..utils import levenshtein_ratio, hybrid_fuzzy_match
 
 
 class BPLOVerifier:
@@ -80,7 +80,7 @@ class BPLOVerifier:
             if not bplo_name:
                 continue
 
-            score = levenshtein_ratio(name_lower, bplo_name)
+            score = hybrid_fuzzy_match(name_lower, bplo_name)
             if score > best_score:
                 best_score = score
                 best_match = entry

{acuity_framework-3.0.0 → acuity_framework-3.1.0}/acuity_framework.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: acuity-framework
-Version: 3.0.0
+Version: 3.1.0
 Summary: ACUITY — A machine learning framework for extracting, verifying, and recommending local micro-enterprise profiles from unstructured community posts.
 Author: ACUITY Research Team
 License: MIT

{acuity_framework-3.0.0 → acuity_framework-3.1.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "acuity-framework"
-version = "3.0.0"
+version = "3.1.0"
 description = "ACUITY — A machine learning framework for extracting, verifying, and recommending local micro-enterprise profiles from unstructured community posts."
 readme = "README.md"
 license = {text = "MIT"}

{acuity_framework-3.0.0 → acuity_framework-3.1.0}/setup.py

@@ -84,7 +84,7 @@ class AcuityBuild(build_py):
     def run(self):
         if _is_fun_enabled():
             print(_color(ANSI_BOLD + BANNER, ANSI_CYAN))
-            print(_color("  ACUITY installer v1.0.0 — extracting local visibility...", ANSI_YELLOW))
+            print(_color("  ACUITY installer v3.1.0 — extracting local visibility...", ANSI_YELLOW))
             print()
             _animate()
             print(_color("  Installing ACUITY Framework...", ANSI_YELLOW))

{acuity_framework-3.0.0 → acuity_framework-3.1.0}/tests/test_verification.py

@@ -4,7 +4,7 @@ Tests for the ACUITY Verification module.
 import pytest
 
 from acuity.verification import BPLOVerifier
-from acuity.utils import levenshtein_ratio, levenshtein_details
+from acuity.utils import levenshtein_ratio, levenshtein_details, token_sort_ratio, token_set_ratio, hybrid_fuzzy_match
 from acuity.config import AcuityConfig
 
 
@@ -104,3 +104,30 @@ class TestBPLOVerifier:
         ])
         result = verifier.verify("Test Biz")
         assert result["status"] == "Verified"
+
+# ── Hybrid Match Tests ─────────────────────────────────────────────────────
+
+class TestHybridFuzzyMatch:
+    def test_token_sort_ratio(self):
+        # Order shouldn't matter
+        assert token_sort_ratio("bakeshop juan", "juan bakeshop") == 1.0
+        assert token_sort_ratio("juan bakeshop", "bakeshop juan") == 1.0
+
+    def test_token_set_ratio(self):
+        # Extra words shouldn't ruin the score completely
+        assert token_set_ratio("juan bakeshop in mamatid", "juan bakeshop") == 1.0
+        
+    def test_hybrid_match_takes_max(self):
+        plain = levenshtein_ratio("bakeshop juan", "juan bakeshop") # Will be low
+        sort = token_sort_ratio("bakeshop juan", "juan bakeshop") # Will be 1.0
+        
+        hybrid = hybrid_fuzzy_match("bakeshop juan", "juan bakeshop")
+        assert hybrid == 1.0
+        assert hybrid > plain
+
+    def test_hybrid_match_penalty(self):
+        # "jb" is an acronym for "juan bakeshop". The length ratio is 2 / 13 = 0.15 (which is < 0.35).
+        # Token set ratio might normally score it too high if it thinks they share tokens, 
+        # but with penalty, it should be lowered to avoid false positives.
+        score = hybrid_fuzzy_match("jb", "juan bakeshop in the city")
+        assert score < 0.5