Fix error/typo in docstring of TokenClassificationPipeline (#19798)

src/transformers/pipelines/token_classification.py

@@ -173,7 +173,7 @@ class TokenClassificationPipeline(Pipeline):
             the following keys:
 
             - **word** (`str`) -- The token/word classified. This is obtained by decoding the selected tokens. If you
-              want to have the exact string in the original sentence, use `start` and `stop`.
+              want to have the exact string in the original sentence, use `start` and `end`.
             - **score** (`float`) -- The corresponding probability for `entity`.
             - **entity** (`str`) -- The entity predicted for that token/word (it is named *entity_group* when
               *aggregation_strategy* is not `"none"`.