diff --git a/src/datasets/news_dataset.py b/src/datasets/news_dataset.py
index fb6515a..753b671 100644
--- a/src/datasets/news_dataset.py
+++ b/src/datasets/news_dataset.py
@@ -1,34 +1,68 @@
 import pandas as pd
-
+from typing import Callable
 from pathlib import Path
 
 
 class NewsDataset:
     def __init__(self, data_dir_path: str = "data", data_lenght: int = None) -> None:
         self.data_dir_path = Path("./" + data_dir_path)
+        self.dataset_dir_path = self.data_dir_path / "dataset"
         self.true_news_path = self.data_dir_path / "True.csv"
         self.fake_news_path = self.data_dir_path / "Fake.csv"
 
-        self.true_news = self.load_news(self.true_news_path, data_lenght)
-        self.fake_news = self.load_news(self.fake_news_path, data_lenght)
+        self.__data_lenght = data_lenght
+
+        self.true_news = self.__load_news(self.true_news_path, self.__data_lenght)
+        self.fake_news = self.__load_news(self.fake_news_path, self.__data_lenght)
 
         self.true_news["label"] = 1
         self.fake_news["label"] = 0
 
-    def load_news(self, file_path: Path, trim: int = None) -> pd.DataFrame:
+        self.train, self.test, self.val = self.__load_train_val_test()
+
+    def __load_news(self, file_path: Path, trim: int = None) -> pd.DataFrame:
         news = pd.read_csv(file_path)
         news = news.drop(columns=["title", "subject", "date"])
 
         return news if not trim else news.head(trim)
 
+    def __load_data(self, file_path: Path) -> pd.DataFrame:
+        df = pd.read_csv(file_path).rename(columns={"Value": "label"})
+        return self.__convert_text(df)
+
+    def __load_train_val_test(self):
+        train_data = self.__load_data(self.dataset_dir_path / "train.csv")
+        test_data = self.__load_data(self.dataset_dir_path / "test.csv")
+        val_data = self.__load_data(self.dataset_dir_path / "val.csv")
+
+        total_data_len = train_data.size + test_data.size + val_data.size
+
+        trim: Callable[[pd.DataFrame], pd.DataFrame] = lambda df: df.head(
+            int(self.__data_lenght * 2 * df.size / total_data_len)
+        )
+        train_data = trim(train_data)
+        test_data = trim(test_data)
+        val_data = trim(val_data)
+        return train_data, test_data, val_data
+
+    def __convert_text(self, df: pd.DataFrame) -> pd.DataFrame:
+        df["text"] = df["text"].str.strip().astype(str)
+        df.dropna(axis=0, how="any", inplace=False, subset=["text"])
+        return df
+
     @property
     def data(self) -> pd.DataFrame:
         dataset = pd.concat([self.true_news, self.fake_news], axis=0)
-        dataset["text"] = dataset["text"].str.strip()
-        dataset.dropna(axis=0, how="any", inplace=False, subset=["text"])
-        return dataset
+        return self.__convert_text(dataset)
 
 
 if __name__ == "__main__":
-    dataset = NewsDataset()
-    print(dataset.data.head(5))
+    dataset = NewsDataset(data_lenght=1000)
+    print("dataset")
+    print(dataset.data.size)
+    print("train")
+    print(dataset.train.size)
+    print("val")
+    print(dataset.test.size)
+    print("test")
+    print(dataset.val.size)
diff --git a/src/main.py b/src/main.py
index 7ee45fa..b5f136f 100644
--- a/src/main.py
+++ b/src/main.py
@@ -1,5 +1,6 @@
-import random
-from sklearn.model_selection import train_test_split
+# import random
+
+# from sklearn.model_selection import train_test_split
 import argparse
 import torch
 from models import BertClassifier, utils
@@ -7,7 +8,7 @@ from datasets import NewsDataset
 from train import train
 from evaluate import evaluate
 
-SEED = 2137
+# SEED = 2137
 
 # Hyperparameters
 
@@ -35,19 +36,23 @@ if __name__ == "__main__":
     # loading & spliting data
     news_dataset = NewsDataset(data_dir_path="data", data_lenght=1000)
 
-    train_val_data, test_data = train_test_split(
-        news_dataset.data,
-        test_size=0.2,
-        shuffle=True,
-        random_state=random.seed(SEED),
-    )
+    train_data = news_dataset.train
+    test_data = news_dataset.test
+    val_data = news_dataset.val
 
-    train_data, val_data = train_test_split(
-        train_val_data,
-        test_size=0.2,
-        shuffle=True,
-        random_state=random.seed(SEED),
-    )
+    # train_val_data, test_data = train_test_split(
+    #     news_dataset.data,
+    #     test_size=0.2,
+    #     shuffle=True,
+    #     random_state=random.seed(SEED),
+    # )
+
+    # train_data, val_data = train_test_split(
+    #     train_val_data,
+    #     test_size=0.2,
+    #     shuffle=True,
+    #     random_state=random.seed(SEED),
+    # )
 
     # trainig model
     if args.train: