Table 1.
Comparison Table between different embedding models.
Fig 1.
Word2Vec models [3].
Fig 2.
The proposed framework.
Table 2.
Number of recognised words.
Fig 3.
Words similar to “Graduate” suggested by the three models.
Fig 4.
Words similar to “Benghazi” suggested by the AraVec Model and Arab2Vec Model.
Fig 5.
Words similar to emoji for AraVec and Arab2Vec model.
Fig 6.
Words that are similar to the emoji as produced by the Arab2Vec model.
Recall that “hh” is often used in Arabic to denote amusement or laughter.
Table 3.
Set of positive and negative words.
Fig 7.
Words clustering from the AraVec model.
The positioning of the words indicates two clear clusters.
Fig 8.
Word clustering Arab2Vec model.
The arrangement of the words indicates that no meaningful clustering has taken place.
Fig 9.
Clustering of named entities by the Arab2Vec model.
Fig 10.
Clustering of named entities AraVec model.
Table 4.
Named entities table.
Table 5.
Deep learning model parameters of LSTM.
Table 6.
F1 score of different ML algorithms on COVID-19 test dataset.
Table 7.
F1 score of different models on the COVID-19 test data set using machine learning (ML) and deep learning (DL). The overall best performer, Arab2vec Skip-gram NS, is in italics.
Table 8.
Distribution of Arabic tweets in the ASTD dataset.
Table 9.
F1 score of different ML algorithms on the ASTD test dataset.
Table 10.
F1 score of different models on ASTD test dataset.