02.05.2014 Views

Proceedings - Österreichische Gesellschaft für Artificial Intelligence

Proceedings - Österreichische Gesellschaft für Artificial Intelligence

Proceedings - Österreichische Gesellschaft für Artificial Intelligence

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

0.7<br />

0.7<br />

0.6<br />

0.6<br />

0.5<br />

0.5<br />

F-measure<br />

0.4<br />

0.3<br />

F-measure<br />

0.4<br />

0.2<br />

0.3<br />

0.1<br />

IG<br />

CRF weights mean<br />

CRF mixture weight<br />

0<br />

10 1 10 2 10 3 10 4 10 5 10 6<br />

# of features<br />

0.2<br />

IG<br />

CRF weights mean<br />

CRF mixture weight<br />

0.1<br />

10 1 10 2 10 3 10 4 10 5 10 6<br />

# of features<br />

(a) bc-msnbc<br />

(b) bn-cnn<br />

0.8<br />

0.7<br />

0.7<br />

0.6<br />

0.6<br />

0.5<br />

F-measure<br />

0.5<br />

0.4<br />

F-measure<br />

0.4<br />

0.3<br />

0.3<br />

0.2<br />

IG<br />

CRF weights mean<br />

CRF mixture weight<br />

0.1<br />

10 1 10 2 10 3 10 4 10 5 10 6<br />

# of features<br />

0.2<br />

IG<br />

CRF weights mean<br />

CRF mixture weight<br />

0.1<br />

10 1 10 2 10 3 10 4 10 5 10 6<br />

# of features<br />

(c) mz-sinorama<br />

(d) nw-xinhua<br />

0.7<br />

0.6<br />

0.5<br />

F-measure<br />

0.4<br />

0.3<br />

0.2<br />

IG<br />

CRF weights mean<br />

CRF mixture weight<br />

0.1<br />

10 1 10 2 10 3 10 4 10 5 10 6<br />

# of features<br />

(e) wb-eng<br />

Figure 1: Feature selection on different training data. The test data is mz-sinorama. The Y-axis on<br />

all charts is F 1 -measure. The X-axis is the number of features. Lines with crosses stand for feature<br />

selection based on information gain (IG). Lines with stars stand for feature selection based on the<br />

feature weight in a mixture of domains (CRF mixture weight). Lines with circles stand for feature<br />

selection based on the feature weights mean (CRF weights mean). One can see that the latter lines start<br />

at higher values of F 1 and most of the time reach flat part of the chart faster than the lines corresponding<br />

to other methods.<br />

251<br />

<strong>Proceedings</strong> of KONVENS 2012 (Main track: poster presentations), Vienna, September 20, 2012

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!