Hybrid-MELAu: A Hybrid Mixing Engineered Linguistic Features Based on Autoencoder for Social Bot Detection.

Social bots are defined as computer algorithms that generate massive amounts of obnoxious or meaningful information. Most bot detection methods leverage multitudinous characteristics, from network features, temporal dynamics activities and sentiment features. However, there has been fairly lower work exploring lexicon measurement linguistic indicators to detect bots. The main purpose this research is recognize the social through their writing style. Thus, we carried out an exploratory study on effectiveness only a set features (17 features) ex- ploitable for detection, without need resort other types And develop novel framework in hybrid fashion Mixing Engineered Linguistic based Autoencoders (Hybrid-MELAu). semi-supervised Hybrid-MELAu frame- composed two essential constituents: learner predictors. We establish innovated powerful structures: a) first Deep dense Autoencoder fed by Lexical Syntactic content (DALS) represents high order lexical syntactic latent space, b) second one Glove-BiLSTM autoencoder, which sculpts semantic features; subsequently, elite elements pre-trained encoder part each space with transfer learning. consider sample 1 Million Cresci datasets conduct our analysis comparison between style humans With dataset, observe bot’s textual diversity median greater than human speech-tagging shows creative behavior Finally, test model’s robustness several public dataset (celebrity, pronbots-2019, political bots). proposed achieves good accuracy 92.22%. Overall, results shown paper, related discussion, argue it possible discern differences humans’ bots’ styles efficient deep framework.