diff --git a/python/tvm/relay/op/op.py b/python/tvm/relay/op/op.py
index 9a158b2027f3..ccf011819a97 100644
--- a/python/tvm/relay/op/op.py
+++ b/python/tvm/relay/op/op.py
@@ -440,7 +440,8 @@ def register_fake_quantization_to_integer(op_name, func=None, level=10):
     """Register quantize function for an op
 
     Given an op and Affine Types on it's inputs, this function should return the op
-    in affine space/integer operators and the new type of the output
+    in affine space/integer operators and the new type of the output, where affine
+    denotes the transformation x_real = (x_affine - zero_point) * scale
 
     Parameters
     ----------
diff --git a/src/relay/transforms/fake_quantization_to_integer.cc b/src/relay/transforms/fake_quantization_to_integer.cc
index 329800125601..1a3c459967bc 100644
--- a/src/relay/transforms/fake_quantization_to_integer.cc
+++ b/src/relay/transforms/fake_quantization_to_integer.cc
@@ -51,7 +51,8 @@
  *
  * The second pass is an ExprVisitor that recursively searches for subgraphs leading to the
  * quantize for subtraphs bounded by dequantize operations. This pass extracts the affine
- * types of the inputs for later processing
+ * types of the inputs for later processing, where affine denotes the transformation
+ * x_real = (x_affine - zero_point) * scale
  *
  * The third pass is an ExprMutator that recursively rewrites the subgraphs using packed funcs
  * registered with the FTVMFakeQuantizationToInteger attribute. These packed funcs rewrite