Although I didn't build it for attribute tables, I've had some success with spellchecking using Microsoft Word through Python. The gist of my process was that I looped through a bunch of metadata files, grabbed a specific subset of metadata elements, wrote each element's text value to a new line in a text file, spellchecked the file in Word, then exported the errors out to a list.
Upsides:
Downsides:
- Not very fast when batch-processing a bunch of different metadata documents
- Occasionally Word would visibly pop-up and stall the process
- Obvious baked-in dependency
I did this as a one-off though, so I didn't really worry about the downsides - if I were to use it frequently I'd really want to nail down the second downside in particular.
Code looked something like this:
import win32com.client
app = win32com.client.gencache.EnsureDispatch("Word.Application")
for metadataFile in metadataFiles:
# Extracted specific metadata elements here, put into list.
# ...
# Dump to text file
with io.open(tempDocPath, "w") as tempFile:
for elementText in elementTextBlocks:
tempFile.write(unicode(elementText))
tempFile.write(u"\n\n")
# Open the doc
doc = app.Documents.Open(tempDocPath)
for error in doc.SpellingErrors:
# Do stuff
# ...
doc.Close()
del doc
app.Quit()
Of course the Close()/del/Quit() stuff was handled in finally blocks and all that, and because I was getting a lot of recurring proper nouns flagged as errors I ended up building in an exclusion process where Line 18 would be, but basically I ended up dumping everything to a CSV with the filename and misspelled word.
