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 significant margin. We empirically investigate several featurization methods to understand the conditions under which discourse features contribute non-trivial performance gains, and analyze discourse embeddings.

PDF

In In Proceedings of the 8th International Joint Conference on Natural Language Processing (IJCNLP-17), pp. 584–593, Taipei, Taiwan, November 2017.

@inproceedings{ferracane:ijcnlp17, title={Leveraging Discourse Information Effectively for Authorship Attribution}, author={Elisa Ferracane and Su Wang and Raymond J. Mooney}, booktitle={In Proceedings of the 8th International Joint Conference on Natural Language Processing (IJCNLP-17)}, month={November}, address={Taipei, Taiwan}, pages={584–593}, url="http://www.cs.utexas.edu/users/ai-labpub-view.php?PubID=127666", year={2017} }

Slides (PDF) Video