[developers] Parsing linux forum data

[Andrew MacKinlay]

Hi,

Following on from my previous post, my more general task is the  
parsing of linux forum data, as some of you may be aware.

Ann suggested I might probe the Delph-in hive mind for any suggestions  
on dealing with this kind of data - eg recommendations for POS- 
tagging, sentence splitting etc.

Posts are, as you might imagine, wildly varying in quality and contain  
combinations of parseable data (which contains the entiy such as a URL  
which we would like to treat as atomic) and console output. For  
example, here is a randomly selected post that shows some of these  
features:


=============
This is my second run at this problem (history is at  http://www.linuxquestions.org/questi...hreadid=3109) 
.

After recompiling kernel 2.4 and doing all the 'make' options, I ran  
'lilo'.  -
I have one problem and three questions:
-
PROBLEM:
When I run lilo, get the following:
-
  Warning: device 0x0306 exceeds 1024 cylinder limit.
Fatal: Sector 51220658 too large for linear mode
(try 'lba32' instead) .
-
Questions:
1.) How do I find out what device 0x0306 is?
2.) How do I find out what is on sector 51220658 and why the system  
says it is too big?
3.)  I am booting from a floppy, which doesn't seem to care
about device 0x0306 or sector 51220658. Why isn't this a problem using  
a floppy?
Thanks to the guys that helped me the first time around and many  
thanks for taking another look. Hope I don't seem to be stupid or  
thankless, but, as yet, I don't understand.
==============

If anyone has any advice they could offer on the basis of experiences  
they've had with similar data in the past, it would be gratefully  
received.

Thanks,
Andy