add preprocessing of data before predicitons

application/functions/sentiment.py

@@ -1,9 +1,9 @@
 from transformers import AutoTokenizer
 from transformers import pipeline
+import re
 
 model = 'application/models/sentiment_model'
 tokenizer = AutoTokenizer.from_pretrained('application/tokenizers/sentiment_tokenizer')
-# tokenizer = AutoTokenizer.from_pretrained("sdadas/polish-gpt2-small")
 
 def sentiment_prediction(data):
     pipe = pipeline('text-classification', model=model, tokenizer = tokenizer)
@@ -11,6 +11,12 @@ def sentiment_prediction(data):
     
     return result
 
+def clear_data(data):
+    data = [re.sub(r"[^A-Za-zżźćńółęąśŻŹĆĄŚĘŁÓŃ ']+", r"", i) for i in data['sentences']]
+    data = [i.strip() for i in data]
+    data = [i.lower() for i in data]
+    return data
+
 def count_predictions(predictions):
     l0 = 0
     l1 = 0

application/services/sentiment_service.py

@@ -3,14 +3,15 @@ from flask import(
     jsonify, 
     Blueprint,
     )
-from application.functions.sentiment import sentiment_prediction, count_predictions
+from application.functions.sentiment import sentiment_prediction, count_predictions, clear_data
 
 sentiment_service = Blueprint("sentiment_service", __name__)    
 
 @sentiment_service.route("/get_sentiment_data", methods=['POST'])
 def get_data():
     data = request.get_json()
-    predicitons = sentiment_prediction(data['sentences']) #predykcje
+    data_clear = clear_data(data)                       #czyszczenie danych wejsciowych
+    predicitons = sentiment_prediction(data_clear)      #predykcje
     count_labels = count_predictions(predicitons)       #dane do wykresu
 
     for i in range(0, len(predicitons)):