While comparing the four "Preview Lookup Equivalents" I noticed there is essential data missing from all except the JSON format. In the others I neglected to retain the grammatical sections of the sentences.

Please have the Table, the DWD Array and the DWD Coordinates store every sentence with JSON object segmentation, for example:

{"determiner":"This","noun":"banking institution or microfinance banking institution, or its controlling company","predicate_verb":"is liable to","attribute":"penalties ","description":"contemplated in section 94(1)","past_participle_verb":"as determined by the Central Bank."}

Since the natural language in this field needs to be parsed anyways, I think there's little lost and processing power gained by including this part of the JSON. Also it makes sense to borrow the JSON syntax rather than create a novel method, so that it is already rountine.