A Comprehensive Study of Learning Approaches for Author Gender Identification

In recent years, author gender identification is an important yet challenging task in the fields of information retrieval and computational linguistics. this paper, different learning approaches are presented to address problem for Turkish articles. First, several classification algorithms applied list representations based on paradigms: fixed-length vector such as Stylometric Features (SF), Bag-of-Words (BoW) distributed word/document embeddings Word2vec, fastText Doc2vec. Secondly, deep architectures, Convolution Neural Network (CNN), Recurrent (RNN), special kinds RNN Long-Short Term Memory (LSTM) Gated Unit (GRU), C-RNN, Bidirectional LSTM (bi-LSTM), GRU (bi-GRU), Hierarchical Attention Networks Multi-head (MHA) designated their comparable performances evaluated. We conducted a variety experiments achieved outstanding empirical results. To conclude, ML with BoW have promising fast-Text also probably suitable between embedding models. This comprehensive study contributes literature utilizing ways representations. It first attempt identify applying SF, DNN architectures language.