Handling variances in user-entered text data

The database that I'm working with has some fields that are critical for my analysis (e.g. for VLOOKUPs) but come from user-entered free text, so I often need to match something like:
"Montezuma #10-W25"
with entries such as:
"muntezuma 10-W 25"
"Montezuma1025W"
"Monte Zuma 25W #10"
"MZA 10-W25"
To date, I've been handling case-by-case errors using TRIM, CLEAN, UPPER, LEFT, MID, etc, but the dataset is large and variations in data entry abound; I'm looking for a more robust approach.

Any suggestions or recommenations on a neural network add-in or other solution that would promote automatic recognition/correction of these kinds of variances?

Thanks in advance,
Mike