Leveraging Discourse Information Effectively for Authorship Attribution
This work addresses authorship attribution for forensic or literary analysis, but appears incremental as it focuses on optimizing existing discourse features.
The authors tackled authorship attribution by embedding discourse features into a Convolutional Neural Network, achieving state-of-the-art results with a substantial performance margin.
We explore techniques to maximize the effectiveness of discourse information in the task of authorship attribution. We present a novel method to embed discourse features in a Convolutional Neural Network text classifier, which achieves a state-of-the-art result by a substantial margin. We empirically investigate several featurization methods to understand the conditions under which discourse features contribute non-trivial performance gains, and analyze discourse embeddings.