4.2 Evaluation Measures
replace the edge connection z
t
= (y
t−2
, y
t−1
, y
t
)
When taking the summarization as a sequential bi-
of order-2 Markov model by z
t
= (y
N
t
, y
t−1
, y
t
) ,
classification problem, we can make use of the usual
where y
N
t
represents the label at the source of the
precision, recall and F1 measures (Shen et al., 2007)
non-local edge. Although it is an approximation of
for classification accuracy evaluation.
the exact inference, we will see that it works well for
In our experiments, we also compare the preci-
our answer summarization task in the experiments.
sion, recall and F1 score in the ROUGE-1, ROUGE-
2 and ROUGE-L measures (Lin , 2004) for answer
4 Experimental Setting
summarization performance.
5 Experimental Results
Bạn đang xem 4. - BÁO CÁO KHOA HỌC: "COMMUNITY ANSWER SUMMARIZATION FOR MULTI-SENTENCE QUESTION WITH GROUP L1 REGULARIZATION" PDF