Identifying single words in a dictionary database

I am reworking a Marathi-English dictionary to be out on open-source. My dictionary has the Headword in Marathi, followed by its Part of Speech and subsequently by its English glosses as in the examples below;

 v i To contract, shrink.
 a Eleven.
 a Frightful, terrible.
 a Frightful, terrible.
 m A walnut.
 m A walnut.
 a That which is not proper to be done; improper.
 a Incapable, incompetent.
 n Impotence, incapability.
 n A bad action, sin. 
 a Wicked.
 a Wicked.
 a Intransitive or neuter.
 a Exempt from stain, blemish.
 f Figure-writing.
  m A wiseacre.
 a Inconceivable; undevisable
 a Unexpected; unthought of.
 a Sinless.
 n Infelicity; injury.
 m Malice, spite, grudge.
 a Spiteful, malicious
 a Spiteful, malicious
 ad Suddenly, unexpectedly, inconsiderately.
 ad Suddenly, unexpectedly, inconsiderately.

As can be seen the delimiter of the gloss is either a comma or a semi-colon
To make the dictionary more accessible and easily readable and concise, I want to retain only single words or words with the following

to 
A 

At present I have written a macro in Ultraedit: a Word processor to do the job. It uses a regex to identify all glosses with single words (with the condition specified above)
However Macros are slow and since the database is around 400.000+ words,I wonder if there is a means of identifying such single glosses through an awk or perl script. Thus in the example above, only the following will be retained

 v i To contract, shrink.
 a Eleven.
 a Frightful, terrible.
 a Frightful, terrible.
 m A walnut.
 m A walnut.
 a That which is not proper to be done; improper.
 a Incapable, incompetent.
 n Impotence, incapability.
 n A bad action, sin. 
 a Wicked.
 a Wicked.
 a Intransitive,neuter.
 a Exempt from stain, blemish.
 f Figure-writing.
  m A wiseacre.
 a Inconceivable; undevisable
 a Unexpected; unthought of.
 a Sinless.
 n Infelicity; injury.
 m Malice, spite, grudge.
 a Spiteful, malicious
 a Spiteful, malicious
 ad Suddenly, unexpectedly, inconsiderately.
 ad Suddenly, unexpectedly, inconsiderately.

I normally have the painful task of further removing cases where a word has a single gloss as well as a long definition as in

 a That which is not proper to be done; improper.

which is reduced to

 a improper.

If both stages can be handled, it would really help a lot.
I work in a Windows environment. Many thanks for any help given which will be gratefully acknowledged by the community once this simplified dictionary is put on line.

This is a *nix forum. What *nix tools would be available on your windows environment?
The only difference I see in the output file is

 a Intransitive,neuter.

. What exactly is the operation of the script to be?

I have Awk/Gawk and Perl.

May be I did not explain myself clearly. What I need is to remove all glosses which have two or more words and retain only the single words.

This implies a two stage operation. In Stage 1 at present I use a regex to identify such unique words within each string and store the string in a separate file. But then it can so happen that within the string there could also be glosses containing more than one word.

 a That which is not proper to be done; improper.

In stage 2 I write a second regex to identify the gloss delimited by

, ; .

resulting in

 a improper.

and which contains more than one word.
It works but the two stage operation is long and tedious and I was wondering if an Awk or Perl script could do the trick.
Thanks a lot