Early Rumor Detection Based on Deep Recurrent Q-Learning

Table 2

Features of text.

Feature name	Description
word_count	Number of words in the sentence
is_question	Describe whether a question mark appears in the sentence
is_exclamation	Describe whether an exclamation mark appears in the sentence
negation_count	Number of negative words, such as “no, nothing”
badword_count	Number of bad words, such as “fuck, bitch”
similar_to_src	Cosine similarity to the original tweet