Upload tool

tool.py

@@ -1,7 +1,7 @@
 from smolagents.tools import Tool
-import string
-import json
 import pronouncing
+import json
+import string
 
 class WordPhoneTool(Tool):
     name = "word_phonetic_analyzer"
@@ -94,9 +94,6 @@ class WordPhoneTool(Tool):
     def _calculate_similarity(self, word1, phones1, word2, phones2):
         import pronouncing
     
-        phone_list1 = phones1.split()
-        phone_list2 = phones2.split()
-    
         # Initialize variables before use
         last_vowel1 = None
         last_vowel2 = None
@@ -108,12 +105,20 @@ class WordPhoneTool(Tool):
         end2_clean = []
         i = 0  # Initialize i for loop variable
     
+        phone_list1 = phones1.split()
+        phone_list2 = phones2.split()
+    
         # Get last syllable components
         result1 = self._get_last_syllable(phone_list1)
         result2 = self._get_last_syllable(phone_list2)
         last_vowel1, word1_end = result1
         last_vowel2, word2_end = result2
     
+        # Calculate length similarity score first
+        phone_diff = abs(len(phone_list1) - len(phone_list2))
+        max_phones = max(len(phone_list1), len(phone_list2))
+        length_score = 1.0 if phone_diff == 0 else 1.0 - (phone_diff / max_phones)
+    
         # Calculate rhyme score (most important)
         rhyme_score = 0.0
         if last_vowel1 and last_vowel2:
@@ -122,37 +127,37 @@ class WordPhoneTool(Tool):
                 end2_clean = self._strip_stress(word2_end)
             
                 if end1_clean == end2_clean:
-                    rhyme_score = 1.0  # Perfect rhyme
+                    rhyme_score = 1.0  # Perfect rhyme, capped at 1.0
                 else:
+                    # Partial rhyme based on ending similarity
                     common_length = min(len(end1_clean), len(end2_clean))
                     matched = 0
                     for i in range(common_length):
                         if end1_clean[i] == end2_clean[i]:
                             matched += 1
-                    rhyme_score = 0.7 + (0.3 * (matched / max(len(end1_clean), len(end2_clean))))
-    
-        # Calculate syllable pattern similarity
-        syl1 = pronouncing.syllable_count(phones1)
-        syl2 = pronouncing.syllable_count(phones2)
-        syllable_score = 1.0 if syl1 == syl2 else 1 - (abs(syl1 - syl2) / max(syl1, syl2))
-    
+                    rhyme_score = 0.6 * (matched / max(len(end1_clean), len(end2_clean)))
+
         # Calculate stress pattern similarity
         stress1 = pronouncing.stresses(phones1)
         stress2 = pronouncing.stresses(phones2)
         stress_score = 1.0 if stress1 == stress2 else 0.5
     
-        # Weighted combination prioritizing rhyming
+        # Weighted combination prioritizing rhyming and length
         total_similarity = (
-            (rhyme_score * 0.8) +       # Rhyming is most important (80%)
-            (syllable_score * 0.15) +   # Syllable count has some importance (15%)
-            (stress_score * 0.05)       # Stress pattern has minimal weight (5%)
+            (rhyme_score * 0.6) +       # Rhyming most important (60%)
+            (length_score * 0.3) +      # Length similarity next (30%)
+            (stress_score * 0.1)        # Stress pattern least important (10%)
         )
     
+        # Ensure total similarity is capped at 1.0
+        total_similarity = min(1.0, total_similarity)
+    
         return {
             "similarity": round(total_similarity, 3),
             "rhyme_score": round(rhyme_score, 3),
-            "syllable_score": round(syllable_score, 3),
-            "stress_score": round(stress_score, 3)
+            "length_score": round(length_score, 3),
+            "stress_score": round(stress_score, 3),
+            "phone_length_difference": phone_diff
         }