面向作文自动评分的优美句识别

Elegart Sentence Recognition for Automated Essay Scoring

语言优美是学生写作能力中重要的一部分。该文提出一个面向作文自动评分的作文优美句识别任务,主要识别中学生中文作文中的优美句。相比传统文本分类任务,优美句识别更加难以用特征工程的方式解决。因此,该文提出一种基于卷积神经网络(CNN)和双向长短时记忆(BiLSTM)网络的混合神经网络结构进行优美句识别,并和CNN、BiLSTM网络进行了对比。实验证明,混合神经网络的准确率最高,达到89.23%,F1值与BiLSTM相当,达到75.39%。此外,该文将优美句子特征用于作文自动评分任务,可使计算机评分和人工评分的大分差比例下降21.41%。

This paper proposs the task of elegant sentence recognition in Chinese essays of high school students for Automated Essay Seoring（AES）. To deal withthis task clellenging the classical text classification plus feature engi- neering ,this paper presents a deep neural network combining Convolutional Neural Network（CNN） and Bi direction- al Long Short Term Memory（BiLSTM） networks to recognize grace sentences. The experiment results show that our joint neural network ranks to in precision （89.23%）,with a comparable F1 score to BiLSTM（75.39%）. We fi- nally apply the graceful sentence features to the AES task, which can reduce the large-margin prediction error by 21.41%.