While creating training data using rasa-ui, If there are multiple occurrences of a word in a sentence, then the first word occurrence "start" and "end" values are assigned to all the other word occurrences.
This is creating issues when training using ner_crf as it uses them as it features.
Example:
This is the json format data obtained where two INR values are having same "start" and "end" values:
Example:
This is the json format data obtained where two INR values are having same "start" and "end" values: