diff --git a/handlers/messages_routing.py b/handlers/messages_routing.py
index 256a792..118b626 100644
--- a/handlers/messages_routing.py
+++ b/handlers/messages_routing.py
@@ -5,6 +5,7 @@ from state.scan import get_average_pattern
 from bot.api import telegram_api, download_file
 from bot.config import FEEDBACK_CHAT_ID
 from handlers.handle_private import handle_private
+from nlp.segment_text import segment_text
 from nlp.toxicity_detector import detector
 from nlp.normalize import normalize
 from nlp.ocr import ocr_recognize
@@ -16,22 +17,9 @@ async def messages_routing(msg, state):
     cid = msg["chat"]["id"]
     uid = msg["from"]["id"]
     text = msg.get("caption", msg.get("text", ""))
-
-    for photo in msg.get("photo", []):
-        file_id = photo.get("file_id")
-        if file_id:
-            async for temp_file_path in download_file(file_id):
-                text += ocr_recognize(temp_file_path)
-                text += '\n'
-
     reply_msg = msg.get("reply_to_message")
 
-    if cid == uid:
-        # сообщения в личке с ботом
-        logger.info("private chat message: ", msg)
-        await handle_private(msg, state)
-
-    elif str(cid) == FEEDBACK_CHAT_ID:
+    if str(cid) == FEEDBACK_CHAT_ID:
         # сообщения из группы обратной связи
         logger.info("feedback chat message: ", msg)
         logger.debug(msg)
@@ -44,8 +32,15 @@ async def messages_routing(msg, state):
                     text=text,
                     reply_to_message_id=reply_msg.get("message_id"),
                 )
+                return
 
-    elif bool(text):
+    elif cid == uid:
+        # сообщения в личке с ботом
+        logger.info("private chat message: ", msg)
+        await handle_private(msg, state)
+        return
+
+    elif bool(text) or msg.get("photo"):
         mid = msg.get("message_id")
         if text == "/toxic@welcomecenter_bot":
             # latest in chat
@@ -106,8 +101,17 @@ async def messages_routing(msg, state):
             except Exception:
                 pass
         else:
+            # on screen recognition
+            for photo in msg.get("photo", []):
+                file_id = photo.get("file_id")
+                if file_id:
+                    async for temp_file_path in download_file(file_id):
+                        text += ocr_recognize(temp_file_path)
+                        text += '\n'
+
             normalized_text = normalize(text)
-            toxic_score = detector(normalized_text)
+            segmented_text = segment_text(normalized_text)
+            toxic_score = detector(segmented_text)
             toxic_perc = math.floor(toxic_score * 100)
 
             if toxic_perc > 49:
diff --git a/nlp/segment_text.py b/nlp/segment_text.py
index 67cbea8..56af820 100644
--- a/nlp/segment_text.py
+++ b/nlp/segment_text.py
@@ -1,25 +1,17 @@
-import torch
-from transformers import ByT5Tokenizer, T5ForConditionalGeneration
-
-# Use ByT5 for the ByT5 model
-tokenizer = ByT5Tokenizer.from_pretrained("google/byt5-small")
-model = T5ForConditionalGeneration.from_pretrained("google/byt5-small")
-
-
+import spacy
 
+# Load the Russian language model
+nlp = spacy.load("ru_core_news_sm")
 
 def segment_text(text):
     """
-    Use a neural network model to segment text into words.
+    Use SpaCy to segment text into words.
     """
-    # Encode the input text for the model as UTF-8 bytes
-    inputs = tokenizer.encode("segment: " + text, return_tensors="pt")
+    # Process the text with SpaCy
+    doc = nlp(text)
 
-    # Generate predictions
-    with torch.no_grad():
-        outputs = model.generate(inputs)
-
-    # Decode the generated tokens back to text
-    segmented_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    # Extract words from the processed document
+    segmented_text = ' '.join([token.text for token in doc])
 
     return segmented_text
+
diff --git a/requirements.txt b/requirements.txt
index 59ed4f7..c4d9f2b 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,7 +1,7 @@
 redis[hiredis]
 aiohttp
 aiofiles
-torch
+spacy
 transformers
 easyocr
 # protobuf