Neural Symbolic Machine 讲解.pdf

发布时间：2022-05-31 发布人：admin 分类：说明书资料大小：2.36M 资料格式：pdf 举报版权申诉

guotong1988-10781273-4744302542853669956.pdf-第1页.png

第1页 / 共77页

guotong1988-10781273-4744302542853669956.pdf-第2页.png

第2页 / 共77页

guotong1988-10781273-4744302542853669956.pdf-第3页.png

第3页 / 共77页

guotong1988-10781273-4744302542853669956.pdf-第4页.png

第4页 / 共77页

guotong1988-10781273-4744302542853669956.pdf-第5页.png

第5页 / 共77页

guotong1988-10781273-4744302542853669956.pdf-第6页.png

第6页 / 共77页

guotong1988-10781273-4744302542853669956.pdf-第7页.png

第7页 / 共77页

guotong1988-10781273-4744302542853669956.pdf-第8页.png

第8页 / 共77页

文本预览

Neural Symbolic Machines Semantic Parsing on Freebase with Weak Supervision Chen Liang, Jonathan Berant, Quoc Le, Kenneth Forbus, Ni Lao

Overview ● Motivation: Semantic Parsing and Program Induction ● Neural Symbolic Machines ○ Key-Variable Memory ○ Code Assistance ○ Augmented REINFORCE ● Experiments and analysis

Semantic Parsing: Language to Programs Natural Language Question/Instruction LATEN T Program / Logical Form Full supervision (hard to collect) Answer Goal [Berant, et al 2013; Liang 2013] Weak supervision (easy to collect)

Question Answering with Knowledge Base Largest city in US? GO (Hop V1 CityIn) (Argmax V2 Population) RETURN NYC Freebase, DBpedia, YAGO, NELL 1. Compositionality 2. Large Search Space Freebase: 23K predicates, 82M entities, 417M triplets

WebQuestionsSP Dataset ● 5,810 questions Google Suggest API & Amazon MTurk1 ● Remove invalid QA pairs2 ● 3,098 training examples, 1,639 testing examples remaining ● Open-domain, and contains grammatical error ● Multiple entities as answer => macro-averaged F1 Grammatical error Multiple entities • What do Michelle Obama do for a living? writer, lawyer • What character did Natalie Portman play in Star Wars? Padme Amidala • What currency do you use in Costa Rica? Costa Rican colon • What did Obama study in school? political science • What killed Sammy Davis Jr? throat cancer [Berant et al, 2013; Yih et al, 2016]

(Scalable) Neural Program Induction ● Impressive works to show NN can ● The learned operations are not as learn addition and sorting, but... scalable and precise. [Reed & Freitas 2015] ● Why not use existing modules that are scalable, precise and interpretable? [Zaremba & Sutskever 2016]

Overview ● Motivation: Semantic Parsing and Program Induction ● Neural Symbolic Machines ○ Key-Variable Memory ○ Code Assistance ○ Augmented REINFORCE ● Experiments and analysis

Neural Symbolic Machines Weak supervision Manager Neural Programmer Question Answer Program Output Symbolic Computer Knowledge Base Predefined Functions Abstract Scalable Precise Non-differentiable

分享到：

赞收藏

资料库

Neural Symbolic Machine 讲解.pdf

相关推荐

人工智能

热门标签

最新资料