The enormous amount of data being generated on the web and social media has\nincreased the demand for detecting online hate speech. Detecting hate speech\nwill reduce their negative impact and influence on others. A lot of effort in\nthe Natural Language Processing (NLP) domain aimed to detect hate speech in\ngeneral or detect specific hate speech such as religion, race, gender, or\nsexual orientation. Hate communities tend to use abbreviations, intentional\nspelling mistakes, and coded words in their communication to evade detection,\nadding more challenges to hate speech detection tasks. Thus, word\nrepresentation will play an increasingly pivotal role in detecting hate speech.\nThis paper investigates the feasibility of leveraging domain-specific word\nembedding in Bidirectional LSTM based deep model to automatically\ndetect/classify hate speech. Furthermore, we investigate the use of the\ntransfer learning language model (BERT) on hate speech problem as a binary\nclassification task. The experiments showed that domainspecific word embedding\nwith the Bidirectional LSTM based deep model achieved a 93% f1-score while BERT\nachieved up to 96% f1-score on a combined balanced dataset from available hate\nspeech datasets.\n