Well, the dictionary program was for an example of getting the data from the site. You'd need to do similar sockopen/sockread things and check to see if it's a noun/adjective by parsing the lines.
The suggestion of putting "known" words into a hash table is a good one as it will speed things up as you start to get more and more known words.
As for submitting the words, yes, you'd submit all of them. If you use the hash table idea, then you'd only submit the words that aren't "known".
You can use the socket part of my script if you want for checking the words. Then, just change the parsing part of it to look just for the noun/adjective info. The rest of the script isn't necessary for what you're doing.
For the hash table part, use a normal hash table setup, where the item name is the word and the data is something simple, such as 1 if it's an adjective/verb, and 0 if it's not. Then, when you loop through your words, you would first check the hash table to see if the itemname (word) exists. If it does, then you won't submit that to be checked because you know if it's an adjective or verb already.