diff --git a/doc/fluid/advanced_usage/paddle_slim/paddle_slim_en.md b/doc/fluid/advanced_usage/paddle_slim/paddle_slim_en.md
index e63212f8de6ba337b9942a0c89c9259c65ba5816..923cfd0d1f564d160424ee275a0d28a29814848e 100644
--- a/doc/fluid/advanced_usage/paddle_slim/paddle_slim_en.md
+++ b/doc/fluid/advanced_usage/paddle_slim/paddle_slim_en.md
@@ -64,7 +64,7 @@ The overall principle of the model compression tool is briefly introduced here t
 <strong>figure 1</strong>
 </p>
 
-As **figure 1** shows, the top purple module is user interface. When calling the model compression function in Python scripts, we only need to construct a Compressor object. Please refer to [Use Document](https://github.com/PaddlePaddle/models/blob/develop/PaddleSlim/docs/usage.md) for more details.
+As **figure 1** shows, the top purple module is the user interface. When calling the model compression function in Python scripts, we only need to construct a Compressor object. Please refer to [Documentation](https://github.com/PaddlePaddle/models/blob/develop/PaddleSlim/docs/usage.md) for more details.
 
 We call every compression algorithms compression strategies, which are called during the process of training models iteratively to complete model compression, as is shown in **figure 2**. The logic is packaged in the tool for model compression, and users only need to provide network structures, data and optimizers for model training. Please refer to [Use Document](https://github.com/PaddlePaddle/models/blob/develop/PaddleSlim/docs/usage.md) for more details.