This paper presents work on using continuousudrepresentations for authorship attribution.udIn contrast to previous work,udwhich uses discrete feature representations,udour model learns continuous representationsudfor n-gram features via a neuraludnetwork jointly with the classificationudlayer. Experimental results demonstrateudthat the proposed model outperforms theudstate-of-the-art on two datasets, while producingudcomparable results on the remainingudtwo.
展开▼
机译：本文介绍了使用连续 udrepresentation进行作者身份归属的工作。 ud与以前的工作相反， ud使用离散特征表示， udour模型通过神经 udnetwork结合分类 udlayer来学习n-gram特征的连续表示 ud 。实验结果证明，所提出的模型在两个数据集上的表现优于最新技术，而在其余两个数据集上却可以产生可比的结果。
展开▼
