Update app.py

app.py

@@ -1837,13 +1837,23 @@ def _wiktionary_find_all_entries(word: str, conn: sqlite3.Connection) -> List[Di
 
 def _wiktionary_format_semantics_block(
     wikt_report: Dict[str, Any], 
-    lemma: str, 
     top_n: int
 ) -> Dict[str, Any]:
     """
-    Combines Wiktionary senses with OdeNet/ConceptNet senses.
+    Combines Wiktionary senses with OdeNet/ConceptNet senses,
+    using the ground-truth lemma from THIS Wiktionary report.
     """
-    # 1. Get Wiktionary senses
+    
+    # --- THIS IS THE FIX ---
+    # The semantic lemma IS the lemma from this specific wiktionary entry.
+    # e.g., for the "heuen" entry, this will be "heuen".
+    # e.g., for the "heute" entry, this will be "heute".
+    semantic_lemma = wikt_report.get("lemma", "")
+    pos_key = _wiktionary_map_pos_key(wikt_report.get("pos"))
+    log(f"Wiktionary Semantics: Building block for lemma='{semantic_lemma}', pos='{pos_key}'")
+    # --- END OF FIX ---
+
+    # 1. Get Wiktionary senses (from the original report)
     wiktionary_senses = []
     for sense in wikt_report.get("senses", []):
         wiktionary_senses.append({
@@ -1851,26 +1861,24 @@ def _wiktionary_format_semantics_block(
             "source": "wiktionary"
         })
 
-    # 2. Get OdeNet senses for this lemma
-    pos_key = _wiktionary_map_pos_key(wikt_report.get("pos"))
+    # 2. Get OdeNet senses for the *semantic_lemma*
     odenet_senses = []
     if WN_AVAILABLE:
         try:
-            # Use the corrected helper from your V18 code
-            senses_by_pos = _get_odenet_senses_by_pos(lemma)
+            senses_by_pos = _get_odenet_senses_by_pos(semantic_lemma)
             odenet_senses_raw = senses_by_pos.get(pos_key, [])
             
             # Filter out placeholder
             if odenet_senses_raw and "info" not in odenet_senses_raw[0]:
                 odenet_senses = odenet_senses_raw
         except Exception as e:
-            log(f"OdeNet lookup failed for {lemma} ({pos_key}): {e}")
+            log(f"OdeNet lookup failed for {semantic_lemma} ({pos_key}): {e}")
 
-    # 3. Get ConceptNet relations
+    # 3. Get ConceptNet relations for the *semantic_lemma*
     conceptnet_relations = []
     if REQUESTS_AVAILABLE:
         try:
-            conceptnet_result = conceptnet_get_relations(lemma, language='de')
+            conceptnet_result = conceptnet_get_relations(semantic_lemma, language='de')
             conceptnet_relations = conceptnet_result.get("relations", [])
         except Exception as e:
             conceptnet_relations = [{"error": str(e)}]
@@ -1883,7 +1891,7 @@ def _wiktionary_format_semantics_block(
         conceptnet_relations = conceptnet_relations[:top_n]
 
     return {
-        "lemma": lemma,
+        "lemma": semantic_lemma, # Return the *correct* lemma for this path
         "wiktionary_senses": wiktionary_senses,
         "odenet_senses": odenet_senses,
         "conceptnet_relations": conceptnet_relations,
@@ -1905,6 +1913,22 @@ def _analyze_word_with_wiktionary(word: str, top_n: int) -> Dict[str, Any]:
     if not conn:
         return {} # Return empty dict to signal failure
         
+    # --- 1. GET SPACY/IWNLP HINT FOR PRIORITIZATION ---
+    spacy_pos_hint = None
+    spacy_lemma_hint = None
+    if IWNLP_AVAILABLE:
+        try:
+            iwnlp = iwnlp_get_pipeline()
+            if iwnlp:
+                doc = iwnlp(word)
+                token = doc[0]
+                spacy_pos_hint = token.pos_.lower()
+                spacy_lemma_hint = token.lemma_
+                log(f"Wiktionary Priority Hint: spaCy POS is '{spacy_pos_hint}', lemma is '{spacy_lemma_hint}'")
+        except Exception as e:
+            log(f"Wiktionary Priority Hint: spaCy/IWNLP failed: {e}")
+
+    # --- 2. FIND ALL WIKTIONARY ENTRIES ---
     try:
         wiktionary_reports = _wiktionary_find_all_entries(word, conn)
     except Exception as e:
@@ -1914,21 +1938,46 @@ def _analyze_word_with_wiktionary(word: str, top_n: int) -> Dict[str, Any]:
     if not wiktionary_reports:
         return {} # No results, signal to fallback
 
+    # --- 3. PRIORITIZE/SORT THE WIKTIONARY ENTRIES ---
+    def get_priority_score(report):
+        wikt_pos = _wiktionary_map_pos_key(report.get("pos"))
+        wikt_lemma = report.get("lemma")
+        
+        # Priority 1: Exact POS match with spaCy hint
+        if spacy_pos_hint and wikt_pos == spacy_pos_hint:
+             # Bonus if lemma also matches
+             if spacy_lemma_hint and wikt_lemma == spacy_lemma_hint:
+                 return 1
+             return 2
+             
+        # Priority 2: Input word is the lemma (e.g., "Haus" -> "Haus")
+        if wikt_lemma.lower() == word.lower():
+            return 3
+            
+        # Priority 3: Other inflected forms (e.g. "gehe" -> "gehen")
+        return 4
+
+    wiktionary_reports.sort(key=get_priority_score)
+    log(f"Wiktionary: Sorted entries: {[r.get('lemma') + ' (' + r.get('pos') + ')' for r in wiktionary_reports]}")
+
+
+    # --- 4. BUILD THE FINAL REPORT (PATH-PURE) ---
     for wikt_report in wiktionary_reports:
         pos_key = _wiktionary_map_pos_key(wikt_report.get("pos"))
         lemma = wikt_report.get("lemma", word)
         
-        # Build the inflection block from Wiktionary data
-        inflections_block = {
+        # --- A. Build Semantics Block (USING WIKT LEMMA) ---
+        # This is the pure path. "heuen" entry looks up "heuen". "heute" entry looks up "heute".
+        semantics_block = _wiktionary_format_semantics_block(wikt_report, top_n)
+        
+        # --- B. Build Wiktionary Inflection Block ---
+        inflections_wikt_block = {
             "base_form": lemma,
             "forms_list": wikt_report.get("forms", []),
             "source": "wiktionary"
         }
         
-        # Build the semantics block
-        semantics_block = _wiktionary_format_semantics_block(wikt_report, lemma, top_n)
-        
-        # Add Pattern.de analysis for comparison/completeness
+        # --- C. Build Pattern Inflection Block (for comparison) ---
         pattern_block = {}
         if PATTERN_DE_AVAILABLE:
             try:
@@ -1938,12 +1987,14 @@ def _analyze_word_with_wiktionary(word: str, top_n: int) -> Dict[str, Any]:
                     pattern_block = pattern_analyze_as_verb(lemma)
                 elif pos_key == "adjective":
                     pattern_block = pattern_analyze_as_adjective(lemma)
-            except Exception:
-                pattern_block = {"error": "Pattern.de analysis failed."}
+                elif pos_key == "adverb":
+                    pattern_block = {"base_form": lemma, "info": "Adverbs are non-inflecting."}
+            except Exception as e:
+                pattern_block = {"error": f"Pattern.de analysis for {pos_key}('{lemma}') failed: {e}"}
 
-        # Build the final report for this POS entry
+        # --- D. Assemble the final report for this entry ---
         pos_entry_report = {
-            "inflections_wiktionary": inflections_block,
+            "inflections_wiktionary": inflections_wikt_block,
             "inflections_pattern": pattern_block,
             "semantics_combined": semantics_block,
             "wiktionary_metadata": {