Research Article

A Supervised Approach to Predict the Hierarchical Structure of Conversation Threads for Comments

Table 5

Experimental results with interrogative evaluation metrics.

DatasetsMethods AccedgeAccpath

ThestandardSLARTS0.4669
±0.009
0.3059 
±0.013
0.6927 
±0.018
0.7450 
±0.010
0.7179 
±0.013
Seo et al.’s [2]0.3941 
±0.005
0.3413 
±0.006
0.8303 
±0.009
0.6563 
±0.005
0.7331 
±0.003
First0.3630 
±0.010
0.3630
±0.010
1.0000
±0.000
0.5846 
±0.009
0.7379
±0.007
Last0.1824 
±0.006
0.0495 
±0.004
0.2214 
±0.008
1.0000
±0.000
0.3625 
±0.010

ENENewsSLARTS0.4837
±0.018
0.3695 
±0.027
0.7625 
±0.026
0.7770 
±0.013
0.7697 
±0.017
Seo et al.’s [2]0.4661 
±0.017
0.4134 
±0.021
0.8720 
±0.011
0.7086 
±0.013
0.7818 
±0.011
First0.4524 
±0.018
0.4524
±0.018
1.0000
±0.000
0.6634 
±0.016
0.7976
±0.011
Last0.2090 
±0.016
0.0596 
±0.003
0.2319 
±0.011
1.0000
±0.000
0.3764 
±0.015

CourantblogsSLARTS0.6843
±0.028
0.6453 
±0.033
0.9127 
±0.013
0.8619 
±0.014
0.8865 
±0.013
Seo et al.’s [2]0.6700 
±0.040
0.6628
±0.043
0.9667 
±0.014
0.8250 
±0.018
0.8902
±0.015
First0.6490 
±0.036
0.6490 
±0.036
1.0000
±0.000
0.8001 
±0.019
0.8889 
±0.012
Last0.2037 
±0.017
0.1082 
±0.008
0.3183 
±0.011
1.0000
±0.000
0.4829 
±0.012

RussianblogSLARTS0.7463
±0.017
0.6904
±0.017
0.8741 
±0.008
0.8787 
±0.006
0.8764
±0.007
Seo et al.’s [2]0.5446 
±0.010
0.5174 
±0.011
0.9293 
±0.003
0.7821 
±0.004
0.8494 
±0.003
First0.4968 
±0.006
0.4968 
±0.006
1.0000
±0.000
0.7134 
±0.004
0.8327 
±0.003
Last0.3632 
±0.018
0.1549 
±0.012
0.3695 
±0.015
1.0000
±0.000
0.5395 
±0.016

AlefSLARTS0.7753 
±0.017
0.7641 
±0.018
0.9579 
±0.006
0.9033 
±0.008
0.9298 
±0.007
Seo et al.’s [2]0.7802 
±0.019
0.7800 
±0.019
0.9950 
±0.003
0.8828 
±0.010
0.9355 
±0.006
First0.7853
±0.018
0.7853
±0.018
1.0000
±0.000
0.8826 
±0.010
0.9376
±0.006
Last0.0993 
±0.005
0.0822 
±0.005
0.2526 
±0.012
1.0000
±0.000
0.4032 
±0.015

The bold style shows the best result in each metric.