Update README.md

README.md

@@ -21,9 +21,11 @@ maker	game	charname	char_in_vndb	vndb_char_id	va_credit_name	va_in_vndb	vndb_sta
 0verflow	Cross Days	乙女	true	c1603	松永 雪希	true	s452	松永 雪希
 ```
 
+Here `va_normalized_name` is the most name most often appeared in the credit of that `vndb_staff_id` among various aliases.
+
 - [`voice_actor_alias.json`](voice_actor_alias.json): Contains the alias of voice actors apeared in the above tsv:
 
-- ```json
+```json
   {
   "s5": {
     "alias_map": {
@@ -50,9 +52,11 @@ maker	game	charname	char_in_vndb	vndb_char_id	va_credit_name	va_in_vndb	vndb_sta
 ```
 
 ## Note (for voice actors info)
-- Some multiple `charname` corresponds to the same charactor (e.g. it occurs when the same charactor is called in many ways).
+
+- **`va_normalized_name` is not checked to be unique (colision may happen)**, so you should use `vndb_staff_id` as an identifier.
 - For char or voice actors not in vndb database, its `vndb_char_id` and `vndb_staff_id` is empty
   - For e.g. speaker verification, you may remove rows with `vndb_staff_id` empty, or manually add unique id to such rows
+- Some multiple `charname` corresponds to the same charactor (e.g. it occurs when the same charactor is called in many ways).
 - Currently I'm manually updating [`voice_actor_info.tsv`](voice_actor_info.tsv), so may be updated
 
 ## Some notes on the dataset which you may find useful