diff --git a/data/ravdess/exp_ravdess_os_xgb.ini b/data/ravdess/exp_ravdess_os_xgb.ini
index 48d93b98..f77caf93 100644
--- a/data/ravdess/exp_ravdess_os_xgb.ini
+++ b/data/ravdess/exp_ravdess_os_xgb.ini
@@ -1,11 +1,11 @@
 [EXP]
-root = ./
-name = results/exp_ravdess
+root = ./results
+name = exp_ravdess_3
 runs = 1
 epochs = 1
 save = True
 [DATA]
-databases = ['train', 'test', 'dev']
+databases = ['train', 'dev']
 train = ./data/ravdess/ravdess_train.csv
 train.type = csv
 train.absolute_path = False
@@ -13,7 +13,7 @@ train.split_strategy = train
 dev = ./data/ravdess/ravdess_dev.csv
 dev.type = csv
 dev.absolute_path = False
-dev.split_strategy = train
+dev.split_strategy = test
 test = ./data/ravdess/ravdess_test.csv
 test.type = csv
 test.absolute_path = False
@@ -24,4 +24,5 @@ labels = ['angry', 'happy', 'neutral', 'sad']
 type = ['os']
 scale = standard
 [MODEL]
-type = xgb
\ No newline at end of file
+type = xgb
+save = True
\ No newline at end of file
diff --git a/data/turev/README.md b/data/turev/README.md
new file mode 100644
index 00000000..c3b0f0fd
--- /dev/null
+++ b/data/turev/README.md
@@ -0,0 +1,9 @@
+# Nkululeko pre-processing for TurEV-DB  
+
+```bash
+git clone https://github.com/Xeonen/TurEV-DB.git
+python3 process_database.py
+cd ../..
+python3 -m nkululeko.resample --config data/turev/exp.ini
+python3 -m nkululeko.nkululeko --config data/turev/exp.ini
+```
\ No newline at end of file
diff --git a/data/turev/exp.ini b/data/turev/exp.ini
new file mode 100644
index 00000000..9cc5e80a
--- /dev/null
+++ b/data/turev/exp.ini
@@ -0,0 +1,30 @@
+[EXP]
+root = ./results/
+name = exp_turev_os_knn
+[DATA]
+databases = ['train', 'dev', 'test']
+train = ./data/turev/turev_train.csv
+train.type = csv
+train.absolute_path = False
+train.split_strategy = train
+dev = ./data/turev/turev_dev.csv
+dev.type = csv
+dev.absolute_path = False
+dev.split_strategy = train
+test = ./data/turev/turev_test.csv
+test.type = csv
+test.absolute_path = False
+test.split_strategy = test
+target = emotion
+; labels = ['anger', 'neutral', 'fear']
+; get the number of classes from the target column automatically
+[FEATS]
+type = ['os']
+; type = ['hubert-xlarge-ll60k']
+; no_reuse = False
+; scale = standard
+[MODEL]
+type = knn
+; save = True
+[RESAMPLE]
+replace = True
\ No newline at end of file
diff --git a/data/turev/process_database.py b/data/turev/process_database.py
new file mode 100644
index 00000000..2bd203ea
--- /dev/null
+++ b/data/turev/process_database.py
@@ -0,0 +1,52 @@
+# process_database.py for TurEV-DB
+
+import argparse
+from pathlib import Path
+from nkululeko.utils.files import find_files
+import pandas as pd
+
+
+def process_database(data_dir, output_dir):  
+    # check if data_dir exist
+    if not Path(data_dir).exists():
+        raise ValueError(f"data_dir {data_dir} does not exist")
+    
+    # check if output_dir exist, create if not
+    if not Path(output_dir).exists():
+        Path(output_dir).mkdir(parents=True)
+
+    sound_source_dir = Path(data_dir) / "Sound Source"
+    wavs = find_files(sound_source_dir, ext=["wav"], relative=True)
+    print(f"Found {len(wavs)} files in {data_dir}")
+
+    # read basename of each wav file, convert to dataframe
+    basenames = [Path(wav) for wav in wavs]
+    df = pd.DataFrame(basenames, columns=["file"])
+
+    # read emotion from parent dir of basename
+    df["emotion"] = [Path(wav).parent.stem.lower() for wav in wavs]
+
+    # read speaker from basename first string before _
+    df["speaker"] = [Path(wav).stem.split("_")[0] for wav in wavs]
+
+    # split into train, val, test
+    # there are six speakers, use one speaker "6783","1358" as val, test
+    df_test = df[df["speaker"] == "6783"]
+    df_dev = df[df["speaker"] == "1358"]
+
+    # use the rest as train
+    df_train = df[df["speaker"]!= "6783"]
+    df_train = df_train[df_train["speaker"]!= "1358"]
+
+    # save all splits to csv
+    for split in ["train", "dev", "test"]:
+        df_split = eval(f"df_{split}")
+        df_split.to_csv(Path(output_dir) / f"turev_{split}.csv", index=False)
+        print(f"Saved {split} to {output_dir}/turev_{split}.csv with shape {df_split.shape}")
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--data_dir", type=str, default="./TurEV-DB", help="Path to data directory")
+    parser.add_argument("--output_dir", type=str, default=".", help="Path to output directory")
+    args = parser.parse_args()
+    process_database(args.data_dir, args.output_dir)
\ No newline at end of file