diff --git a/br/COMPATIBILITY_TEST.md b/br/COMPATIBILITY_TEST.md
index b5580835baee8..44984ebcd2bfa 100644
--- a/br/COMPATIBILITY_TEST.md
+++ b/br/COMPATIBILITY_TEST.md
@@ -3,7 +3,7 @@
 ## Background
 
 We had some incompatibility issues in the past, which made BR cannot restore backed up data in some situations.
-So we need a test workflow to check the compatiblity.
+So we need a test workflow to check the compatibility.
 
 ## Goal
 
diff --git a/docs/design/2019-11-05-index-advisor.md b/docs/design/2019-11-05-index-advisor.md
index 53abab5ba0d5a..5606d9bd9a942 100644
--- a/docs/design/2019-11-05-index-advisor.md
+++ b/docs/design/2019-11-05-index-advisor.md
@@ -57,7 +57,7 @@ for {
 Note that executing `Swap and Re-evaluate` algorithm is necessary as the `reduced_cost` sometimes is a joint effect of several indexes and it's hard to tell each index's independent contribution to the final `reduced_cost`. For example, assume there is an extremely slow query in input workload and the desired indexes for this query is `a` and `b`. However, the number of allowed recommended indexes for the whole workload is limited and for some reason, `a` ranks top `n` in the final score list while `b` is not. But there are chances that without `b`, `a` can no more optimize that extremely slow query. 
 
 ----------------------------------------------
-### A quick exmaple for single-column index recommendation
+### A quick example for single-column index recommendation
 
 **Workload**:
 
diff --git a/docs/design/2020-01-24-collations.md b/docs/design/2020-01-24-collations.md
index af7c9e30b2f26..d610514ed76d5 100644
--- a/docs/design/2020-01-24-collations.md
+++ b/docs/design/2020-01-24-collations.md
@@ -233,7 +233,7 @@ The following features of the general collation algorithm will be supported:
   * Tertiary Weight i.e. case
   * PAD / NOPAD
 
-All of them are supported by `text/collate` package of Go, so it is possible to map Go collations to some of UCA-based collations in MySQL like `utf8mb4_unicode_ci`/`utf8mb4_0900_ai_ci`, if we ignore the differences between UCA versions: current `text/collate` uses UCA version `6.2.0` and it is not changable. However, the collations in MySQL are with different UCA versions marked in the names, for example, `utf8mb4_0900_ai_ci` uses version `9.0`.
+All of them are supported by `text/collate` package of Go, so it is possible to map Go collations to some of UCA-based collations in MySQL like `utf8mb4_unicode_ci`/`utf8mb4_0900_ai_ci`, if we ignore the differences between UCA versions: current `text/collate` uses UCA version `6.2.0` and it is not changeable. However, the collations in MySQL are with different UCA versions marked in the names, for example, `utf8mb4_0900_ai_ci` uses version `9.0`.
 
 For non-standard UCA implementations in MySQL, i.e. the `utf8mb4_general_ci`. The implementation depends on our choice to the [Compatibility with MySQL](#compatibility-with-mysql) chapter, if a 100% compatibility of `utf8mb4_general_ci` is chosen, we need to implement it by our hands.