[{"data":1,"prerenderedAt":310},["ShallowReactive",2],{"content-query-om0MaVrF3R":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"cover":11,"type":12,"category":13,"body":14,"_type":304,"_id":305,"_source":306,"_file":307,"_stem":308,"_extension":309},"/technology-blogs/zh/3222","zh",false,"","基于昇思MindSpore的能量超球体模型提升以事件为中心的结构化预测","作者：李锐锋 来源：知乎","2024-07-11","https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/07/12/3b267b7c3d354f9da89c158a54750538.png","technology-blogs","大V博文",{"type":15,"children":16,"toc":301},"root",[17,25,44,49,54,59,64,69,80,85,94,99,107,115,123,128,133,138,143,151,159,169,177,185,190,197,202,207,215,223,231,238,246,253,261,268,275,283,291,296],{"type":18,"tag":19,"props":20,"children":22},"element","h1",{"id":21},"基于昇思mindspore的能量超球体模型提升以事件为中心的结构化预测",[23],{"type":24,"value":8},"text",{"type":18,"tag":26,"props":27,"children":28},"p",{},[29,31,37,39],{"type":24,"value":30},"**作者：**",{"type":18,"tag":32,"props":33,"children":34},"strong",{},[35],{"type":24,"value":36},"李锐锋",{"type":24,"value":38}," ",{"type":18,"tag":32,"props":40,"children":41},{},[42],{"type":24,"value":43},"来源：知乎",{"type":18,"tag":26,"props":45,"children":46},{},[47],{"type":24,"value":48},"论文标题",{"type":18,"tag":26,"props":50,"children":51},{},[52],{"type":24,"value":53},"SPEECH: Structured Prediction with Energy-Based Event-Centric Hyperspheres",{"type":18,"tag":26,"props":55,"children":56},{},[57],{"type":24,"value":58},"论文来源",{"type":18,"tag":26,"props":60,"children":61},{},[62],{"type":24,"value":63},"ACL 2023",{"type":18,"tag":26,"props":65,"children":66},{},[67],{"type":24,"value":68},"论文链接",{"type":18,"tag":26,"props":70,"children":71},{},[72],{"type":18,"tag":73,"props":74,"children":78},"a",{"href":75,"rel":76},"https://aclanthology.org/2023.acl-long.21/",[77],"nofollow",[79],{"type":24,"value":75},{"type":18,"tag":26,"props":81,"children":82},{},[83],{"type":24,"value":84},"代码链接",{"type":18,"tag":26,"props":86,"children":87},{},[88],{"type":18,"tag":73,"props":89,"children":92},{"href":90,"rel":91},"https://github.com/mindspore-lab/models/tree/master/research/ZJU/speech",[77],[93],{"type":24,"value":90},{"type":18,"tag":26,"props":95,"children":96},{},[97],{"type":24,"value":98},"昇思MindSpore作为开源的AI框架，为产学研和开发人员带来端边云全场景协同、极简开发、极致性能、安全可信的体验，支持国内高校/科研机构发表1000+篇AI顶会论文。本文是昇思MindSpore AI顶会论文系列第41篇，我选择了来自浙江大学的张宁豫和新加坡国立大学Bryan Hooi老师团队发表于ACL的一篇论文解读，感谢各位专家教授同学的投稿，更多精彩的论文精读文章和开源代码实现请访问Models。",{"type":18,"tag":26,"props":100,"children":101},{},[102],{"type":18,"tag":103,"props":104,"children":106},"img",{"alt":7,"src":105},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/07/12/bf1c2e95b93441dfb572f324cf45dec2.png",[],{"type":18,"tag":26,"props":108,"children":109},{},[110],{"type":18,"tag":32,"props":111,"children":112},{},[113],{"type":24,"value":114},"01",{"type":18,"tag":26,"props":116,"children":117},{},[118],{"type":18,"tag":32,"props":119,"children":120},{},[121],{"type":24,"value":122},"研究背景",{"type":18,"tag":26,"props":124,"children":125},{},[126],{"type":24,"value":127},"以事件为中心的结构化预测旨在输出事件的结构化组件。事件结构往往比较复杂，而且结构间存在多种依赖，比如Token之间的长距离依赖、触发词和事件类别之间的关联、以及事件类别和事件关系之间的依赖。",{"type":18,"tag":26,"props":129,"children":130},{},[131],{"type":24,"value":132},"为了解决这些问题，本文提出了基于能量的超球体（SPEECH）以解决以事件为中心的结构化预测任务。SPEECH模型的设计灵感主要来源于万有引力定律和能量函数。如果把抽象的事件类别看成电子核/恒星，其对应的样本就可看成核外电子/行星。因此本文提出利用基于能量的超球体来建模事件类别及其样本。模型主要与 token、sentence、document三种维度的能量相关联。",{"type":18,"tag":26,"props":134,"children":135},{},[136],{"type":24,"value":137},"SPEECH使用基于能量的建模对事件结构化组件之间的复杂依赖进行建模，并使用简单但有效的超球体来表示事件类别。在MAVEN-ERE和OntoEvent-Doc两个事件数据集上进行的实验表明，SPEECH模型在事件检测和事件关系抽取任务上表现不俗。",{"type":18,"tag":26,"props":139,"children":140},{},[141],{"type":24,"value":142},"为了进一步优化SPEECH模型的性能，我们将其在昇思MindSpore平台上进行了实现与测试。昇思MindSpore作为一个新兴的深度学习框架，具有高效、灵活和易于部署的特点, 并提供了丰富的工具集来加速模型的开发和优化。利用昇思MindSpore的模型优化工具，我们对SPEECH模型的结构和参数进行了多次迭代和调优，进一步提升了模型在事件检测和事件关系抽取任务上的性能。",{"type":18,"tag":26,"props":144,"children":145},{},[146],{"type":18,"tag":32,"props":147,"children":148},{},[149],{"type":24,"value":150},"02",{"type":18,"tag":26,"props":152,"children":153},{},[154],{"type":18,"tag":32,"props":155,"children":156},{},[157],{"type":24,"value":158},"团队介绍",{"type":18,"tag":26,"props":160,"children":161},{},[162,167],{"type":18,"tag":32,"props":163,"children":164},{},[165],{"type":24,"value":166},"张宁豫",{"type":24,"value":168},"，浙江大学副教授，浙江大学启真优秀青年学者，在高水平国际学术期刊和会议上发表多篇论文，代表工作有KnowPrompt、DeepKE、EasyEdit、OceanGPT（沧渊）等，获浙江省科技进步二等奖，IJCKG最佳论文/提名2次，CCKS最佳论文奖1次, 担任ACL、EMNLP领域主席、ARR Action Editor、IJCAI 高级程序委员。",{"type":18,"tag":26,"props":170,"children":171},{},[172],{"type":18,"tag":32,"props":173,"children":174},{},[175],{"type":24,"value":176},"03",{"type":18,"tag":26,"props":178,"children":179},{},[180],{"type":18,"tag":32,"props":181,"children":182},{},[183],{"type":24,"value":184},"论文简介",{"type":18,"tag":26,"props":186,"children":187},{},[188],{"type":24,"value":189},"SPEECH 模型的设计灵感主要来源于万有引力定律和能量函数。如果把抽象的事件类别看成电子核/恒星，其对应的样本就可看成核外电子/行星。因此本文提出利用基于能量的超球体来建模事件类别及其样本。模型主要与 token、sentence、document 三种维度的能量相关联。",{"type":18,"tag":26,"props":191,"children":192},{},[193],{"type":18,"tag":103,"props":194,"children":196},{"alt":7,"src":195},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/07/12/28347ec08f794d4599b34f7796a4eb4e.png",[],{"type":18,"tag":26,"props":198,"children":199},{},[200],{"type":24,"value":201},"本文主要在三类以事件为中心的结构化预测任务上进行了实验，用到的数据集是最新发布的 MAVEN-ERE 以及 OntoEvent-Doc。由于 MAVEN-ERE 没有直接发布测试集标签，本文的实验是在 MAVEN-ERE 的验证集以及 OntoEvent-Doc 的测试集上进行的。",{"type":18,"tag":26,"props":203,"children":204},{},[205],{"type":24,"value":206},"主要根据触发词分类，事件分类，时间关系抽取三类任务进行实验。",{"type":18,"tag":26,"props":208,"children":209},{},[210],{"type":18,"tag":32,"props":211,"children":212},{},[213],{"type":24,"value":214},"04",{"type":18,"tag":26,"props":216,"children":217},{},[218],{"type":18,"tag":32,"props":219,"children":220},{},[221],{"type":24,"value":222},"实验结果",{"type":18,"tag":26,"props":224,"children":225},{},[226],{"type":18,"tag":32,"props":227,"children":228},{},[229],{"type":24,"value":230},"4.1 触发词分类（针对token）",{"type":18,"tag":26,"props":232,"children":233},{},[234],{"type":18,"tag":103,"props":235,"children":237},{"alt":7,"src":236},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/07/12/0704232ef2604819a20544699d2ab8c0.png",[],{"type":18,"tag":26,"props":239,"children":240},{},[241],{"type":18,"tag":32,"props":242,"children":243},{},[244],{"type":24,"value":245},"4.2 事件分类（针对event mention）",{"type":18,"tag":26,"props":247,"children":248},{},[249],{"type":18,"tag":103,"props":250,"children":252},{"alt":7,"src":251},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/07/12/0d808ccff2e747ffb49d0761bb5df4ee.png",[],{"type":18,"tag":26,"props":254,"children":255},{},[256],{"type":18,"tag":32,"props":257,"children":258},{},[259],{"type":24,"value":260},"4.3 事件关系抽取（针对event mention pair）",{"type":18,"tag":26,"props":262,"children":263},{},[264],{"type":18,"tag":103,"props":265,"children":267},{"alt":7,"src":266},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/07/12/a677fdbb44e54c25b1dc27613973e6cb.png",[],{"type":18,"tag":26,"props":269,"children":270},{},[271],{"type":18,"tag":103,"props":272,"children":274},{"alt":7,"src":273},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/07/12/14f89c9610324f518ab48f8bb73f2890.png",[],{"type":18,"tag":26,"props":276,"children":277},{},[278],{"type":18,"tag":32,"props":279,"children":280},{},[281],{"type":24,"value":282},"05",{"type":18,"tag":26,"props":284,"children":285},{},[286],{"type":18,"tag":32,"props":287,"children":288},{},[289],{"type":24,"value":290},"总结与展望",{"type":18,"tag":26,"props":292,"children":293},{},[294],{"type":24,"value":295},"本文提出了一种基于能量的超球体模型（SPEECH），用于解决以事件为中心的结构化预测任务。针对事件结构复杂且组件之间存在多种依赖关系的问题，SPEECH 模型通过能量函数和超球体表示方法，成功捕捉了事件结构化组件之间的复杂依赖。在 MAVEN-ERE 和 OntoEvent-Doc 两个数据集上的实验结果表明，SPEECH 模型在事件检测和事件关系抽取任务上均展示了显著的性能和精度提升。",{"type":18,"tag":26,"props":297,"children":298},{},[299],{"type":24,"value":300},"随着人工智能和深度学习技术的不断发展，昇思MindSpore 作为一个高效、灵活且强大的框架，具有广阔的应用前景。昇思MindSpore 框架具有以下几个显著优势：高效的计算加速：与硬件深度整合，充分利用计算资源，显著提高了模型训练速度。自动混合精度：通过在训练过程中自动选择适当的数值精度，降低了内存使用，提高了计算效率。在未来，相信昇思MindSpore生态系统将不断扩展，覆盖更多的行业应用。",{"title":7,"searchDepth":302,"depth":302,"links":303},4,[],"markdown","content:technology-blogs:zh:3222.md","content","technology-blogs/zh/3222.md","technology-blogs/zh/3222","md",1776506127541]