基于深度学习的范畴语法系统研究与实现  

Research and implementation of categorical grammar system based on deep learning

作　　者：赵鑫源 童文喜 Zhao Xinyuan;Tong Wenxi(School of Information Engineering,North China University of Water Resources and Electric Power,Zhengzhou 450000,China)

机构地区：[1]华北水利水电大学信息工程学院,郑州450000

出　　处：《现代计算机》2024年第19期65-69,共5页Modern Computer

摘　　要：近年来,深度学习在自然语言处理(NLP)领域获得了很大成功,尤其是语义识别方面优势突出。但是,深度学习在分析句法构成和识别句法成分方面的效果较差。其中序列标注是自然语言处理领域中历史最悠久的研究课题之一,包括词性标签(Part of speech tagging)。对范畴语法标签这一任务进行研究,提出了一些技术,可以让赋予每个输入词的词法类别数目减少。研究目标是开发一个简单而准确的系统模型来解决范畴标签的挑战,同时利用神经网络后向传播算法必要的间接表示以避免复杂的人工特征选择。基于深度学习算法的研究,用Haskell语言设计并实现范畴语法系统,对词嵌入过程的监测,能更好地反映范畴的变化。In recent years,deep learning has achieved great success in the field of natural language processing(NLP),especially in semantic recognition.However,deep learning is not effective in analyzing syntactic composition and identifying syntactic components.Sequence tagging is one of the oldest research topics in natural language processing,including Part of speech tagging.Research into the task of categorical syntax labeling has suggested techniques to reduce the number of lexical categories assigned to each input word.Research object is to develop a simple and accurate system model to solve the category labeling challenge,while utilizing the indirect representation necessary for neural network backpropagation algorithms to avoid complex artificial feature selection.Based on the research of deep learning algorithm,Haskell language is used to design and implement categorical grammar system,which can monitor the process of word embedding and better reflect the change of category.

关 键 词：深度学习 范畴语法 HASKELL语言 后向传播算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP391.1[自动化与计算机技术—控制科学与工程]