split speech sentences process

embedder.py

@@ -27,6 +27,7 @@ from pathlib import Path
 
 # NLP and ML libraries
 from sentence_transformers import SentenceTransformer
+from transformers import AutoTokenizer
 from sklearn.decomposition import PCA
 from sklearn.manifold import TSNE
 from sklearn.metrics.pairwise import cosine_similarity
@@ -108,7 +109,24 @@ class PersianVectorAnalyzer:
         except Exception as e:
             logger.error(f"Error loading model: {e}")
             raise
-
+    def split_sentence(self, sentence:str):
+        sentences = []
+        sentence_len = len(self.tokenize_sentence(sentence))
+        if sentence_len < 512:
+            sentences.append(sentence)
+        else:
+            temp_sentences = str(sentence).split('.')
+            for sent in temp_sentences:
+                sent_len = len(self.tokenize_sentence(sent))
+                if sent_len > 512:
+                    temp_sentences_2 = str(sentence).split('،')
+                    for snt in temp_sentences_2:
+                        sentences.append(snt)
+                else:
+                    sentences.append(sent)
+                    
+        return sentences
+        
     def load_json_data(self, file_path: str) -> List[str]:
         """
         Load Persian sentences from JSON file.
@@ -136,11 +154,11 @@ class PersianVectorAnalyzer:
                         # Extract sentences from different possible keys
                         for key in ['persian_translate']:
                             if key in item and item[key]:
-                                splited_sentences = str(item[key]).split('.')
+                                splited_sentences = self.split_sentence(item[key])
                                 for sent in splited_sentences:
                                     sentences.append(sent)
                     elif isinstance(item, str):
-                        splited_sentences = str(item).split('.')
+                        splited_sentences = self.split_sentence(item[key])
                         for sent in splited_sentences:
                             sentences.append(sent)
             elif isinstance(data, dict):
@@ -181,6 +199,18 @@ class PersianVectorAnalyzer:
 
         return text.strip()
 
+    def tokenize_sentence(self, sentence:str):
+        
+        try:
+            tokenizer = AutoTokenizer.from_pretrained(self.model_name)
+            print(self.model_name)
+            tokens = tokenizer.tokenize(sentence)
+            return tokens
+        except:
+            error = "An exception occurred in tokenizer : " + model_checkpoint
+            #file.write( error + '\n' )
+            return []
+
     def extract_words(self, sentences: List[str]) -> List[str]:
         """
         Extract all words from sentences.
@@ -464,8 +494,8 @@ class PersianVectorAnalyzer:
         Run the complete processing pipeline.
 
         Args:
-            input_file: Path to input JSON file
-            output_dir: Output directory for results
+            input_file(str): Path to input JSON file
+            output_dir(str): Output directory for results
         """
         # Create output directory
         Path(output_dir).mkdir(exist_ok=True)
@@ -475,6 +505,10 @@ class PersianVectorAnalyzer:
         # Step 1: Load data
         sentences = self.load_json_data(input_file)
 
+        for s in sentences:
+            s_len = len(self.tokenize_sentence(s))
+            if s_len > 512:
+                print(s)
         # Step 2: Extract words
         # all_words = self.extract_words(sentences)
 
@@ -530,8 +564,8 @@ def main():
     analyzer = PersianVectorAnalyzer()
 
     # Define input and output paths
-    input_file = "./data/final_wisdom.json"
-    output_dir = "output"
+    input_file = "./out/nahj_speeches.json"
+    output_dir = "output-speechs"
 
     # Run the complete pipeline
     analyzer.process_pipeline(input_file, output_dir)