作者:刘洋 页数:102 出版社:清华大学 |
基于深度学习的机器 翻译 刘洋 清华大学 1 机器翻译 • 目标:利用计算机实现自然语言的动翻译 目标:利用计算机实现自然语言的动翻译 目标:利用计算机实现自然语言的动翻译 目标:利用计算机实现自然语言的动翻译 目标:利用计算机实现自然语言的动翻译 目标:利用计算机实现自然语言的动翻译 2 布什 与 沙龙 举行 了 会谈 Bush held a talk with with Sharon Sharon 发展历史 • 趋势:让机器更“自主”地学习如何翻译 趋势:让机器更“自主”地学习如何翻译 趋势:让机器更“自主”地学习如何翻译 趋势:让机器更“自主”地学习如何翻译 趋势:让机器更“自主”地学习如何翻译 趋势:让机器更“自主”地学习如何翻译 3 规则 机器翻译 统计 机器翻译 神经 机器翻译 1980 1990 2013 数据驱动 机器翻译 1990 罗塞塔石碑 罗塞塔石碑 4 平行语料库是破解言奥秘的钥匙 5 Garcia y Garcia y Garcia y asociados asociados . los clients y clients y clients y clients y los asociados asociados son enemigos enemigos . sus asociados asociados no son fuertes fuertes. Garcia and associates . Garcia and associates . Garcia and associates . Garcia and associates . Garcia and associates . Garcia and associates .Garcia and associates . the clients and associates are enemies . the clients and associates are enemies . the clients and associates are enemies . the clients and associates are enemies . the clients and associates are enemies . the clients and associates are enemies . the clients and associates are enemies . the clients and associates are enemies . the clients and associates are enemies .the clients and associates are enemies . the clients and associates are enemies . the clients and associates are enemies .the clients and associates are enemies . his associates are not strong . his associates are not strong . his associates are not strong . his associates are not strong . his associates are not strong .his associates are not strong . his associates are not strong . his associates are not strong . his associates are not strong . Garcia y Garcia y Garcia y sus asociados asociados no son enemigos enemigos . Garcia and his associates are not enemies . Garcia and his associates are not enemies . Garcia and his associates are not enemies . Garcia and his associates are not enemies . Garcia and his associates are not enemies .Garcia and his associates are not enemies . Garcia and his associates are not enemies . Garcia and his associates are not enemies . Garcia and his associates are not enemies . 西班牙语 西班牙语 英语 Garcia Garcia Garcia Garcia y and asociados asociados associates associates . . los los the the clients clients son are enemigos enemigos enemigos enemies enemies sus hishis no not fuertes fuertes strong strong 数据驱动的机器翻译 数据驱动的机器翻译 数据驱动的机器翻译 统计机器翻译 统计机器翻译 • 核心问题:如何为翻译过程建立概率模型? 核心问题:如何为翻译过程建立概率模型? 核心问题:如何为翻译过程建立概率模型? 核心问题:如何为翻译过程建立概率模型? 核心问题:如何为翻译过程建立概率模型? 核心问题:如何为翻译过程建立概率模型? 6 布什 与 沙龙 举行 了 会谈 Bush held a talk with with Sharon Sharon 基于短语的统计机器翻译 基于短语的统计机器翻译 基于短语的统计机器翻译 • 短语翻译模型:以隐结构为基本单元 短语翻译模型:以隐结构为基本单元 短语翻译模型:以隐结构为基本单元 短语翻译模型:以隐结构为基本单元 短语翻译模型:以隐结构为基本单元 短语翻译模型:以隐结构为基本单元 7 布什 与 沙龙 举行 了 会谈 布什 与 沙龙 举行 了 会谈 Bush with Sharon ith Sharon ith Sharon held a talk eld a talk eld a talk Bush with Sharon ith Sharon held a talk eld a talk Bush held a talk with with Sharon Sharon (Koehn et al., 2003) (Koehn et al., 2003)(Koehn et al., 2003)(Koehn et al., 2003) (Koehn et al., 2003) (Koehn et al., 2003) 统计 机器翻译 • 隐变量对数线性模型:在式语言结构上设计特征 隐变量对数线性模型:在式语言结构上设计特征 隐变量对数线性模型:在式语言结构上设计特征 隐变量对数线性模型:在式语言结构上设计特征 隐变量对数线性模型:在式语言结构上设计特征 隐变量对数线性模型:在式语言结构上设计特征 8 布什 与 沙龙 举行 了 会谈 Bush held a talk with with Sharon Sharon (Och and Ney, 2002) (Och and Ney, 2002) (Och and Ney, 2002)(Och and Ney, 2002) (Och and Ney, 2002) (Och and Ney, 2002) 例子 9 例子 10 例子
empty