Table 1.

Features used for the logistic regression model.

Thread-level features	Description
NumPost	Number of posts in the thread
NumUser	Number of authors participating in the thread discussion
AvgLen	Average length of post (by word numbers) in the thread
Post-level features	Description
NumName	Number of mentions of other authors’ names
NumNeg	Number of negative sentiment words
NumPos	Number of positive sentiment words
NumCAM	Number of CAM related keywords
NumOverlap	Number of words that also occur in previous post
Num?	Number of question marks
Num!	Number of exclamation marks
TimeDif	Time difference between current and previous post in thread
Sig	If the author has a signature profile
NAgree	Number of “agree”s
NDisagree	Number of “disagree”s
Lexical features	Description
LDA	Topic modeling
LDA-sim	cosine similarity between LDA of current and previous post
W2V	Word embedding
W2V-sim	cosine similarity between W2V of current and previous post