This new study proposes a novel approach to hate speech detection on Twitter by using a combination of textual, social, user, and language features. This is an interesting development in the field of automated hate speech detection. The results of the analyses presented in the article show that the proposed approach outperformed all the baselines, with a gain in accuracy of up to 6% by adding social graph encoder features. Importantly, the study found that the significance of relationships among users on Twitter is directly proportional to the efficacy of any downstream task, and that the retweet relationship represents a stronger social connection than the follower-followee relationship. The study also observed that tweet length and social network structure are crucial factors in hate speech detection. The proposed approach has the potential to improve the detection of hateful content online, and can help conducting better research on the motivations behind hate speech, leading to the development of targeted strategies to reduce its prevalence. In conclusion, the study presents a promising solution towards creating a more inclusive and respectful online community by detecting and curbing hate speech on Twitter.