diff --git a/README.md b/README.md
index 9be27091207993d7e0ed02f8dfc6219287d521e6..637e0c80cf00311e2353b0ba4c390e4f7b9fbd08 100644
--- a/README.md
+++ b/README.md
@@ -10,6 +10,8 @@ Expected inputs are a ground truth and a prediction BIOES/BILOU files without an
 
 ## Usage
 
+### Installation
+
 After cloning the repository, install the package with:
 
 ```
@@ -23,12 +25,28 @@ $ pip3 install tox
 $ tox
 ```
 
+### Usage
+
 You can now use Nerval in command line :
+
 ```
-$ nerval -a/--annot <annot_file.bio> -p/--predict <predict-file.bio>
+$ nerval -a/--annot <annot_file.bio> -p/--predict <predict-file.bio> \
+		 [-t/--threshold <threshold_value>]
 ```
 
-To use the demo files :
+The threshold value should be between 0 and 1. It designates the acceptable number of characters differing between an annotated and a predicted entity - over the number of characters in the annotated entity - to consider it as a match. Default value is 0.30. 0 would impose perfect matches, 1 would allow completely different strings to be considered as a match.
+
+For instance, if we consider the following case:
+
+| Annotation        | Prediction        |
+| ----------------- | ----------------- |
+| Hugone B-PERS     | Hugone B-PERS     |
+| Montiniaci I-PERS | Montiniaci I-PERS |
+| domino I-PERS     | domino O          |
+
+Counting the spaces, 7 characters differ over 24 characters in the reference entity: a threshold of 0.30 would accept the match but a lower one would not.
+
+### Demo
 
 ```
 $ nerval -a demo/demo_annot.bio -p demo/demo_predict.bio