search

google-code-export / dkpro-tc

Automatically exported from code.google.com/p/dkpro-tc
Other
1 stars 0 forks source link

Handling "empty" predictions in the evaluation module #231

Open GoogleCodeExporter opened 9 years ago

GoogleCodeExporter commented 9 years ago 
In several places, the evaluation module currently assumes that, for each 
instance, there is one or more labels (strings or integers) for both gold 
standard and prediction. Due to particularities in the machine learning 
frameworks (see Issue 228), this might not always be the case.
One solution might be to check for empty predictions right when the evaluation 
file (id2outcome) is first read, and replace empty strings with a dedicated 
constant and integer (e.g. "-1").

Original issue reported on code.google.com by daxenber...@gmail.com on 26 Jan 2015 at 8:48