• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>International symposium on integrated uncertainty in knowledge modelling and decision making >Averaged Logits: An Weakly-Supervised Approach to Use Ratings to Train Sentence-Level Sentiment Classifiers
【24h】

Averaged Logits: An Weakly-Supervised Approach to Use Ratings to Train Sentence-Level Sentiment Classifiers

机译：平均Logits：使用评分来训练句子级情感分类器的弱监督方法

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

As an important aspect of review mining, sentence-level sentiment classification has received much attention from both academia and industry. Many recently developed methods, especially the ones based on deep learning models, have centred around the task. In a majority of the existing methods, training sentence-level sentiment classifiers require sentence-level sentiment labels, that are usually expensive to obtain. In this research, we propose a novel approach, named 'Averaged logits', that uses the prevalently available ratings, instead of sentence-level sentiment labels to train the classifiers. In the approach, the rating of a review is assumed to be the 'average' of the sentiments of the individual sentences. We experiment with this idea under the framework of the recurrent neural network model. The results show that, the performance of the proposed approach is close to that of the traditional SVM and Naive Bayes classifiers trained by labelled sentences when their training sizes are approximately equal, and close to that of the neural network based classifiers trained by labelled sentences when the proposed approach uses approximately 5 times more training samples.
机译：作为评论挖掘的重要方面，句子层次的情感分类受到了学术界和业界的广泛关注。许多新近开发的方法，尤其是基于深度学习模型的方法，都围绕着这项任务。在大多数现有方法中，训练句子级情感分类器需要句子级情感标签，而这些标签通常价格昂贵。在这项研究中，我们提出了一种称为“平均对数”的新颖方法，该方法使用普遍可用的等级，而不是句子级的情感标签来训练分类器。在这种方法中，评论的等级被认为是各个句子的情感的“平均”水平。我们在递归神经网络模型的框架下尝试这种想法。结果表明，所提方法的性能与传统的支持向量机和朴素贝叶斯分类器的训练大小近似相等时的性能接近，而与基于神经网络的经标记句子训练的分类器的性能接近。建议的方法使用的训练样本大约多5倍。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号