This is simply a step-by-step outline of how I set up to run CorpusSearch on my PC. I'm no windows guru and I can't guarentee what works for me will work for you. But it probably will. If you have problems, email me (email@example.com) and I'll try to help.
The biggest problem people have working with CorpusSearch under Windows is converting back and forth from a word processor format to dos/ascii. You can avoid this by using a MS-DOS editor to write your query files. With a small amount of effort in setting up your desktop before you start, CorpusSearch can be run extremely easily and efficiently under Windows.
C:\WINDOWSChange to your qq folder/directory
edit filename.qA window will open in which you can type your query. It has the usual obvious pull-down menus for saving and exiting and so on. When you're finished writing your query, leave the MS-DOS window open.
qq\filename.qFor the source file just use the files in the psd directory (it's shorter to type than search-files and there's no need to move files unless you want to search a combination of files that you can't specify using * as a wild card).
psd\cm*.m2.psd (for example)
psd\*m2* (even shorter)
For me, using Word Perfect, when I double-click the file it opens as a WP file, I hit RETURN to convert it from dos/ascii to WP so I can read it. When I'm finished I quit without saving and WP leaves the file in dos/ascii format (since I didn't save it). This means it's still searchable by CS. I don't use Word but I expect it will act much the same. Since the output file is in dos/ascii format, you can also read it using "edit" (just use "edit filename.out" as we did above). Often these utilities don't support very large files, however, so if you're doing large searches this might not work. If you're doing a lot of searching, before you open the first output file, set the default initial font on your word processor to a fixed-width font like courier new and the smallest font size you can read (I use 10). This avoids the skewing of the parses that occurs when the fixed-width font of the corpus is converted to a variable-width font like Times and gets as much of the parse on the screen as possible without wrapping.