开源技术 * IBM 微讲堂:Kubeflow 系列(观看回放 | 下载讲义) 了解详情

IBM Debater® 标记演讲中强调的词

概述

IBM Debater® 标记演讲中强调的词(IBM Debater® Labeled Emphasized Words in Speech)数据集包含 2485 个段落,由 2012 年维基百科所编制的语音文件(面向听众)中的 4002 个语句组成。

这些段落将提供给一位以美国英语为母语的女性专业人士,指导她以生动且极具说服力的语调朗读出来。继而,每句话都由4名专业的标记者为所强调的词进行注释。这些标注员会按照以下准则来处理原始文本和录音数据。

I) 重点词是相对于语句中的大多数词而言,在语音中着重强调的词。

II) 仅根据语音来标注,而不是根据词在文本中的重要程度来标注。

数据集元数据

格式 许可 领域 记录数 大小 初始发布日期
TXT
CC-BY-SA 3.0 自然语言处理 2485 个含重点词的语句
3.3 MB 2018-09-02

记录示例

Exposure to <style=emphasized_word> violent </style=emphasized_word> video games causes at least a temporary increase in <style=emphasized_word> aggression </style=emphasized_word> and this exposure correlates with <style=emphasized_word> aggression </style=emphasized_word> in the real world.    Exposure to <style=emphasized_word> violent </style=emphasized_word> video games causes at least a temporary increase in <style=emphasized_word> aggression </style=emphasized_word> and this exposure correlates with <style=emphasized_word> aggression </style=emphasized_word> in the real world.A single child would be left with having to provide support for his or her <style=emphasized_word> two </style=emphasized_word> parents and <style=emphasized_word> four </style=emphasized_word> grandparents.The pursuit of <style=emphasized_word> doping </style=emphasized_word> athletes has turned into a modern day <style=emphasized_word> witch </style=emphasized_word>-<style=emphasized_word> hunt </style=emphasized_word>.
A single child would be left with having to provide support for his or her <style=emphasized_word> two </style=emphasized_word> parents and <style=emphasized_word> four </style=emphasized_word> grandparents.    
The pursuit of <style=emphasized_word> doping </style=emphasized_word> athletes has turned into a modern day <style=emphasized_word> witch </style=emphasized_word>-<style=emphasized_word> hunt </style=emphasized_word>.

相关链接

  • IBM Project Debater Project Debater 是首款能够就复杂主题与人类展开辩论的 AI 系统。其目的是帮助人们建立具有说服力的论据并做出明智的决策。此数据集有助于训练 Project Debater 中的模型。

本文翻译自:IBM Debater® Labeled Emphasized Words in Speech(2019-07-29)