Python NLP

our approach consists of two
main phases: Phase 1:Radical Properties Extraction, where
articles from Dabiq extremist magazines are input into this
step to perform two parallel tasks. In the first task, we
build a language model using (i) Term-Frequency Inverse-
Document-Frequency (TF-IDF) scores of uni-, bi-, and trigrams,
and (ii) Word embeddings generated from a word2vec
model [16]. The output of this task is a radical corpus of
top k-grams, and a word embedding model giving a vector
representation for each word in the corpus. The second task
seeks to create a psychological profile based on the language
used in the extremist propaganda articles, consisting of a set
of emotional and topical categories using LIWC dictionary-based
tool. Phase 2: Tweet classification involves the use of
the models generated from Phase 1 to engineer features related
to radical activities. We identify three groups of features and
then train a binary classifier to detect radical tweets.