ResearchBib Share Your Research, Maximize Your Social Impacts
Sign for Notice Everyday Sign up >> Login


Journal: International Journal of Advanced Research in Engineering and Technology (IJARET) (Vol.13, No. 05)

Publication Date:

Authors : ;

Page : 55-64

Keywords : BERT; Deep Learning; Natural Language Processing; Spam; DistilBERT; ALBERT;

Source : Downloadexternal Find it from : Google Scholarexternal


As a consequence of technological advancements, the use of messages on phones, laptops, desktops, and other devices has increased dramatically, resulting in a high number of spam messages and texts. Identifying spam URLs in social media is a major duty for protecting consumers from links associated with fraud and spam. Twitter allows all users to freely generate and consume large volumes of data, regardless of their merits. Though people and corporations use this data to gain a competitive advantage, spam and fraudulent users generate a substantial amount of data. Modern academics have used certain technological aspects to classify them as ham or spam. Stop words have been found despite the implementation of machine learning and deep learning models. An ideal model is created utilizing transformer architecture to decrease this. For spam detection, BERT (Bidirectional Encoder Representations from Transformers) is used, which uses a pre-trained model with fine-tuning. The approach creates embeddings by taking into consideration every word's context, including its neighbors on both the left and right sides of a sequence, and then categorizes as spam or ham. For greater performance, DistilBERT employs Knowledge distillation to detect the messages, whereas ALBERT needs fewer parameters than BERT

Last modified: 2022-07-04 15:45:46