We've looked for trained data for named entity recognition (in particular for company names). The end goal is to feed in a list of entity names (with potential misspellings, acronyms, punctuation etc) and map these to official company names and hence a ticker e.g. BBG ticker.
Searches here found questions asked like Dataset for Named Entity Recognition on Informal Text though this does the tagging of the words, not a mapping to a clean entity name so it isn't what we're looking for.
If this is not readily available, should we be approaching this by scrapping a list of entities from an informal source and feeding that through some of the available APIs e.g. Google Cloud to 'create' a trained dataset?