资源描述
单击此处编辑母版标题样式,编辑母版文本样式,第二级,第三级,第四级,第五级,10/20/2018,#,自然语言处理,国际前沿动态综述,CCL2018,孙薇薇,北京大学,CCL,每年都有都有国际前沿动态,报告。所以,,我重点介绍过去一年的最新,热点,Best,paper,或提名,Presentation,的时候人山人海,汤,亦或石头,两种论文,简单问题复杂化,复杂问题简单化,这,才是,人民群众喜闻乐见的,技术,哪都能用,哪用哪好,包治百病的,ELMO,NAACL 2018 Best Paper,ELMO,前身,ELMo,本尊,ELMo,本尊,各种各样的,task,不仅仅是序列标注,保卫地球大作战,Chris Dyer,Emily Bender,Jason Eisner,Mark Johnson,保卫地球大作战,Is linguistic information useful for neural network architectures,:,can,it improve state of the art neural architectures,and how should it be used?,Does,it help in building models that transfer better to new domains,new languages,new tasks,or for other limited annotated data scenarios?,Are there any better implicit representations that neural networks can extract,whether similar or not to linguistic structures,that can be transferred or shared across tasks and,hence,serve as core language representation layers?,LISA:Linguistically-Informed Self-Attention,EMNLP 2018 Best Paper,LISA,LISA,结论是:,ELMo,很强大,结论是:,LISA,在背起手的时候很不错,句法分析就这点能耐?,句法分析显著提升了系统鲁棒性,拥抱喵星人,ACL 2018 Best Paper,提名,QA-SRL,“标准”,SRL,:,Arg0/Agent,Arg1/Patient,QA-SRL,优点,降低了对标注者“专业性”的要求,使得普通人也可以充当标注者,有利于扩大数据量,大佬说,ACL2017,的论文:找点儿新数据,定义个新任务,然后扔,LSTM,里,2018,仍然如此,Transfer,Learning,ACL 2017 Keynote,Multitask Learning,What is Computational linguistics?,https:/,Computational linguistics,is,the scientific study of language from a computational perspective,.,Work,in computational linguistics is in some cases motivated from a scientific perspective in that one is,trying to provide a computational explanation for a particular linguistic or psycholinguistic phenomenon,;and in other cases the motivation may be more purely technological in that one wants to,provide a working component of a speech or natural language system,.,Computational explaination,ACL 2018 Best Paper,谢谢,大家的耐心聆听,数据很重要,数据很重要,数据很重要,折腾数据很重要,无标注数据(,LM,),异质数据(,Multitask Learning,),相关数据(,MT,平行语料),写,在“可解释性”之前,
展开阅读全文