Looks like number string parsing may be of enough importance to warrant token look-ahead recursive sub-parsing, then use the most "promising" token-ization; th

Looks like number string parsing may be of enough importance to warrant token look-ahead recursive sub-parsing, then use the most "promising" token-ization; the one that generates the highest yielding probability tree following that number string.