文章摘要
一种基于Bi-LSTM的多层面隐喻识别方法
A Multi-level Metaphor Detection Method Based on Bi-LSTM
投稿时间:2019-09-22  修订日期:2019-12-11
DOI:
中文关键词: 自然语言理解  隐喻识别  CNN  Bi-LSTM  依存关系  
英文关键词: natural language understanding  metaphor detection  CNN  Bi-LSTM  dependence relationship  
基金项目:国家自然科学基金项目(面上项目,重点项目,重大项目)
作者单位E-mail
朱嘉莹 杭州电子科技大学认知与智能计算研究所 zhujy@hdu.edu.cn 
王荣波 杭州电子科技大学认知与智能计算研究所 wangrongbo@hdu.edu.cn 
黄孝喜 杭州电子科技大学认知与智能计算研究所  
谌志群 杭州电子科技大学认知与智能计算研究所  
摘要点击次数: 41
全文下载次数: 0
中文摘要:
      本文以双向长短期记忆网络(Bi-LSTM)为核心,结合多层卷积神经网络,以及单向长短期记忆网络构建多层面的隐喻识别模型。基于多特征协同作用的思想,利用依存关系特征,语义特征,词性特征等多特征融合输入的方法丰富模型的学习信息。为降低信息干扰,利用基于统计学的规范化文本输入方法提升模型的识别效果。在英文语料的词层面和句层面的实验中,各个特征均表现出明显的正向作用。两种方法在英文语料的词层面研究中F1-score分别提升2.5%和5.1%,在句层面的研究分别提升3.1%和1.9%。的在中文语料句层面的实验中,最优效果的F1值可达88.8%。
英文摘要:
      This paper takes bi-directional long short-term memory network (Bi-LSTM) as the core, combining with multi convolutional neural network layers and unidirectional long short-term memory network to build a multi-level metaphor recognition model. Based on the idea of multi-feature synergism, we enrich the learning information of the model by using the methods of inputting multi-feature such as dependency feature, semantic feature and part-of-speech feature in parallel. In order to reduce information interference, a standardized text input method based on statistics is used to improve the recognition effect of the model. In the experiments on word level and sentence level of English corpus, each feature has obvious positive effect. The F1-scores of the two methods increased by 2.5% and 5.1% respectively in the word level research of English corpus, and 3.1% and 1.9% respectively in the sentence level research. In the experiment at the level of Chinese corpus, the F1-score of the optimal effect can reach 88.8%.
View Fulltext   查看/发表评论  下载PDF阅读器
关闭