diff --git a/README.md b/README.md
index a79cb01..880d305 100644
--- a/README.md
+++ b/README.md
@@ -3,7 +3,36 @@
 1. Create conda env: `conda create -n pos-bert python=3.7`
 2. Install dependencies: `pip install -r requirements.txt`
 3. Download UD: Use [this link](https://lindat.mff.cuni.cz/repository/xmlui/handle/11234/1-3226) to download all treebanks, unzip and place them inside `data` folder
-4. Rename the files in the concerned treebanks to `train/dev/test.conllu`
-5. Meta-train pos-tagger: `python meta_trainer.py ewt partut` (can take arbitrary number of datasets)
-6. Evaluate pos-tagger: `python test.py ewt partut --split=test`
+4. TODO: Make a script to create the 'x' folder inside `data`
+5. Create new configs in the `configs` folder and use them to start training
 
+Trainer usage
+```
+usage: trainer.py [-h] --config_path CONFIG_PATH [--train_type {meta,mtl}]
+
+Train a classifier
+
+optional arguments:
+  -h, --help            show this help message and exit
+  --config_path CONFIG_PATH
+                        Path of the config containing training params
+  --train_type {meta,mtl}
+                        Whether to perform MTL or meta-training
+```
+
+Tester usage
+```
+usage: tester.py [-h] --test_path TEST_PATH --model_path MODEL_PATH
+                 [-e {meta,full,both}]
+
+Test POS tagging on various UD datasets
+
+optional arguments:
+  -h, --help            show this help message and exit
+  --test_path TEST_PATH
+                        Datasets to test on
+  --model_path MODEL_PATH
+                        Path of the model to load
+  -e {meta,full,both}, --eval_type {meta,full,both}
+                        Type of evaluation to perform
+```
\ No newline at end of file
diff --git a/tester.py b/tester.py
index 1cc73e1..132593d 100644
--- a/tester.py
+++ b/tester.py
@@ -86,12 +86,12 @@ def meta_evaluate(dataset, label_map, bert_model, clf_head, config):
 
 def init_args():
     parser = argparse.ArgumentParser(description="Test POS tagging on various UD datasets")
-    parser.add_argument("--test_path", dest="test_path", type=str, help="Datasets to test on", required=True)
+    parser.add_argument("--test_path", dest="test_path", type=str, help="Dataset to test on", required=True)
     parser.add_argument("--model_path", dest="model_path", type=str, help="Path of the model to load", required=True)
     parser.add_argument(
         "-e",
         "--eval_type",
-        help="Type of evaluation (meta/regular)",
+        help="Type of evaluation to perform",
         choices=["meta", "full", "both"],
         default="both",
     )