| 本文已被：浏览 3168次 下载 5458次 |
码上扫一扫！ |
|
|
| 一种神经范畴标注模型 |
|
吴惠甲1,2, 张家俊1,2, 宗成庆1,3,2
|
|
1.模式识别国家重点实验室(中国科学院 自动化研究所), 北京 100190;2.中国科学院大学, 北京 100190;3.中国科学院 脑科学与智能技术卓越创新中心, 北京 100190
|
|
| 摘要: |
| 范畴标注是组合范畴语法解析中的子任务之一，可用于提高解析器的效率和性能.传统的最大熵模型需要手工定义特征模板，神经网络则通过隐含层学习到离散特征的分布式表示，从而自动提取分类需要的特征.引入该模型来解决该问题，在原有神经语言模型的基础上加入了向量化的词性表示层和范畴表示层，并通过反向传播自动更新词向量、词性向量和范畴向量，学习到它们的分布式表示.此外，在预测时采用束搜索的序列解码方式来引入标签之间的依赖信息.实验结果表明，这两种改进都能提升模型的性能，使其在范畴标注任务上比传统的最大熵模型效果要好（提升1%）. |
| 关键词: 范畴标注 分布式表示 神经语言模型 束搜索 |
| DOI：10.13328/j.cnki.jos.004873 |
| 分类号: |
| 基金项目:国家自然科学基金（61333018）；中国科学院战略性先导科技专项（XDB02070007） |
|
| Neural-Based Combinatory Categorical Grammar Supertagging |
|
WU Hui-Jia1,2, ZHANG Jia-Jun1,2, ZONG Cheng-Qing1,3,2
|
|
1.National Laboratory of Pattern Recognition(Institute of Automation, The Chinese Academy of Sciences, Beijing 100190, China;2.University of Chinese Academy of Sciences, Beijing 100190, China;3.Center for Excellence in Brain Science and Intelligence Technology, The Chinese Academy of Sciences, Beijing 100190, China
|
| Abstract: |
| As a sub-task for combinatory categorical grammar (CCG) based parsing, categorical tagging can improve parsing efficiency and accuracy. While traditional maximum entropy model solves this problem by designing meaningful feature templates, neural network can extract features automatically based on distributed representations. This paper proposes a neural categorical tagging model with two improvements. First, word embedding layer is extended with a part-of-speech embedding layer and a category embedding layer, which facilitates learning their distributed representations jointly by the back-propagation algorithm. Secondly, a beam search is used in the decoding to capture the dependencies among tags. These two improvements make the proposed model more accurate than the state-of-art maximum entropy based tagger (up to 1%). |
| Key words: categorical tagging distributed representation neural language model beam search |
