RASA NLU Training Data - Entity Position is not correct

The entity position returned by the RASA model is not correct.

For example, if I pass inpu text ‘how many mss bugs’ to the model, it would return below output:

entities: [{‘start’: 9, ‘end’: 31, ‘value’: ‘mss’, ‘entity’: ‘cluster_name’, ‘confidence’: 0.9998497947778505, ‘extractor’: ‘ner_crf’}]

As above shows, the ‘end’ position is 31 which is larger than the length of the whole input text. It’s really wierd. Could someone help here?

Well. I figured it out. It’s due to my own fault in defining training data