Rule-based data cleaning is very important in rule-match and fuzzy-match algorithm.

Everything is classification, so classification solves all. Also, everything is mapping to another dimension to be understand.

What is mapping? Mapping solves all:
text-mapTo-API,
text-mapTo-classifyLabel,
text-mapTo-itsTranslate,
text-mapTo-itsAnswer.

There is no substitute for carefully labeling the data one by one, but a pre-trained LLM allows us to label less data.