[Ligncse256] NaN in default POS tagger output

[Roger Levy]

hbeecher at ling.ucsd.edu wrote:
> Hello,
> 
> Is anyone getting the output message
> 
>   Tag Accuracy: NaN
> 
> when running the test harness with the WSJ data for assignment 2 ??
> 
> I take NaN to indicate some sort of 'divide by 0' error which in the
> POSTaggerTester code may be coming from the evaluateTagger routine having
> 0 in the numTags variable.
> 
> I don't want to assume it is a problem with my platform/configuration, etc
> only to find out it's a data file format problem or something!
> 
> Please give me a shout if you're getting this or something similar.

Hi Henry,

Yes, NaN usually arises from dividing zero by zero.  Here's a website 
with a decent explanation:

   http://www.concentric.net/~Ttwang/tech/javafloat.htm

-- 

Roger Levy                      Email: rlevy at ucsd.edu
Assistant Professor             Phone: 858-534-7219
Department of Linguistics       Fax:   858-534-4789
UC San Diego                    Web:   http://ling.ucsd.edu/~rlevy