diff --git a/_typos.toml b/_typos.toml
index 8b281b8f886..b7fd3fdd637 100644
--- a/_typos.toml
+++ b/_typos.toml
@@ -19,6 +19,7 @@ CANN = "CANN"
 Clas = "Clas"
 arange = "arange"
 unsupport = "unsupport"
+Nervana = "Nervana"
 
 # These words need to be fixed
 Accuray = "Accuray"
@@ -41,8 +42,6 @@ Interfce = "Interfce"
 Kernal = "Kernal"
 Learing = "Learing"
 Moible = "Moible"
-Nervana = "Nervana"
-Numberic = "Numberic"
 Operaton = "Operaton"
 Optimizaing = "Optimizaing"
 Optimzier = "Optimzier"
@@ -169,9 +168,6 @@ metrices = "metrices"
 muliply = "muliply"
 mulitplying = "mulitplying"
 mutbale = "mutbale"
-neeeded = "neeeded"
-nessesary = "nessesary"
-numberic = "numberic"
 occurence = "occurence"
 opeartor = "opeartor"
 opeartors = "opeartors"
diff --git a/docs/design/network/deep_speech_2.md b/docs/design/network/deep_speech_2.md
index 10c14746650..5497c022bd4 100644
--- a/docs/design/network/deep_speech_2.md
+++ b/docs/design/network/deep_speech_2.md
@@ -137,7 +137,7 @@ Key ingredients about the layers:
     - Bi-direcitional + without row convolution: if we don't care about the inference latency.
 - **Row convolution**:
     - For looking only a few steps ahead into the feature, instead of looking into a whole sequence in bi-directional RNNs.
-    - Not nessesary if with bi-direcitional RNNs.
+    - Not necessary if with bi-direcitional RNNs.
     - "**Row**" means convolutions are done within each frequency dimension (row), and no convolution kernels shared across.
 - **Batch Normalization Layers**:
    - Added to all above layers (except for data and loss layer).
diff --git a/docs/dev_guides/custom_device_docs/memory_api_en.md b/docs/dev_guides/custom_device_docs/memory_api_en.md
index 0d6595ae4db..5fe6bcf9c21 100644
--- a/docs/dev_guides/custom_device_docs/memory_api_en.md
+++ b/docs/dev_guides/custom_device_docs/memory_api_en.md
@@ -232,7 +232,7 @@ dst - the address of destination device memory
 
 src - the address of source host memory
 
-size - the size of memory neeeded to be copied (in byte)
+size - the size of memory needed to be copied (in byte)
 
 ## async_memory_copy_d2h 【optional】
 
diff --git a/docs/dev_guides/style_guide_and_references/type_annotations_specification_cn.md b/docs/dev_guides/style_guide_and_references/type_annotations_specification_cn.md
index 341d02da61e..6bf88ef1e2f 100644
--- a/docs/dev_guides/style_guide_and_references/type_annotations_specification_cn.md
+++ b/docs/dev_guides/style_guide_and_references/type_annotations_specification_cn.md
@@ -176,7 +176,7 @@ if TYPE_CHECKING:
     from paddle import Tensor
 
 # TensorLike 定义如下
-# TensorLike: TypeAlias = Union[npt.NDArray[Any], "Tensor", Numberic]
+# TensorLike: TypeAlias = Union[npt.NDArray[Any], "Tensor", Numeric]
 
 def add(a: Tensor, b: TensorLike) -> Tensor:
     if isinstance(b, np.ndarray):
@@ -184,10 +184,10 @@ def add(a: Tensor, b: TensorLike) -> Tensor:
     elif isinstance(b, paddle.Tensor):
         return dispatch_paddle_add(a, b)
     else:
-        return dispatch_numberic_add(a, b)
+        return dispatch_numeric_add(a, b)
 ```
 
-这里 `TensorLike` 是一个通用类型，包含了 `np.ndarray`、`paddle.Tensor`、`Numberic` 三种类型，因此在实现中也应该考虑到这三种类型的输入。
+这里 `TensorLike` 是一个通用类型，包含了 `np.ndarray`、`paddle.Tensor`、`Numeric` 三种类型，因此在实现中也应该考虑到这三种类型的输入。
 
 ### 使用更加明确的类型以提供更好的提示效果