Update ner_logic.py

ner_logic.py

@@ -1,29 +1,39 @@
 import torch
+import wikipedia  # Hata almamak için pip install wikipedia yapılmalı
 import re
 import json
 import spaces
 from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
 from labels_config import CUSTOM_LABELS
-from prompts import SYSTEM_COT_PROMPT  # Harici prompt modulu
+from prompts import SYSTEM_COT_PROMPT
 
-# Modellerin yuklenmesi (XLMR ve Qwen)
+# --- MODELLER ---
 ner_pipe = pipeline("ner", model="xlm-roberta-large-finetuned-conll03-english", aggregation_strategy="simple", device=-1)
 model_id = "Qwen/Qwen2.5-1.5B-Instruct"
 tokenizer = AutoTokenizer.from_pretrained(model_id)
 llm_model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.bfloat16, device_map="auto", trust_remote_code=True)
 
+def get_wiki_summary(term):
+    """Wikipedia'dan bilgi çeken kritik fonksiyon."""
+    try:
+        wikipedia.set_lang("tr")
+        search_results = wikipedia.search(term)
+        if not search_results: 
+            return "Wikipedia'da bu terim için sonuç bulunamadı."
+        # İlk sonucun özetini al
+        summary = wikipedia.summary(search_results[0], sentences=2, auto_suggest=False)
+        return summary
+    except Exception as e:
+        return f"Wikipedia sorgu hatası: {str(e)}"
+
 @spaces.GPU
 def refine_labels_batch(misc_items, full_sentence):
     if not misc_items: return []
     device = "cuda" if torch.cuda.is_available() else "cpu"
     
-    # Etiket kurallari metni oluşturma
     label_rules = "".join([f"### {k}: {v['tanim']}\n" for k, v in CUSTOM_LABELS.items()])
-    
-    # Varlik ve Wiki metni oluşturma
     targets_text = "".join([f"- VARLIK: {item['word']} | WIKI: {item['wiki']}\n" for item in misc_items])
 
-    # Harici promptu doldurma
     final_prompt = SYSTEM_COT_PROMPT.format(
         label_rules=label_rules,
         full_sentence=full_sentence,
@@ -41,7 +51,6 @@ def refine_labels_batch(misc_items, full_sentence):
         json_match = re.search(r'\[\s*\{.*\}\s*\]', output, re.DOTALL)
         if json_match:
             results = json.loads(json_match.group(0))
-            # Karar alaninda True/False temizligi
             for r in results:
                 if not isinstance(r.get("karar"), str) or r["karar"] in ["True", "False"]:
                     r["karar"] = "MISC"