引导式测验 · 中等 等级
从文本测验中提取关系
测试关系提取如何将文本转化为结构化事实,以及其常见管道和监督方法如何失败。
问题 1 的 8
关系提取系统的典型输出是什么?
预览每个问题8 问题
当您参加测验时,就会出现答案和解释。
- 关系提取系统的典型输出是什么?
- 通过关系连接两个实体的结构化三元组
- 句子的语法分析树
- 情绪评分从消极到积极
- 将句子翻译成另一种语言
- 在关系提取的通用管道中,什么可以在对分类之前识别实体提及?
- 文本转语音合成
- 命名实体识别
- 图像分割
- Audio source separation
- 为什么远程监督会产生嘈杂的关系提取标签?
- Knowledge bases contain no structured facts
- It requires every sentence to be hand-labeled
- A sentence may mention two linked entities without stating their knowledge-base relation
- It removes entity mentions before training
- 许多基于 Transformer 的 RE 模型如何指示要分类的实体对?
- 首先翻译句子
- 通过删除实体之一
- 通过将文本转换为音频
- 通过在两个实体周围插入特殊标记令牌
- “开放”关系提取与标准 RE 有何区别?
- 它直接从文本中提取关系短语,无需预定义一组关系类型
- 它只适用于数字
- 它比任何其他方法需要更多的标记数据
- 每个文档只能找到一个关系
- Why include a no-relation label when classifying candidate entity pairs?
- To translate the sentence into another language
- To represent pairs for which the text does not assert a target relation
- To replace named entity recognition
- To guarantee that every pair has a relation
- What does document-level relation extraction allow a system to use?
- Only the spelling of a single entity
- Audio features from a spoken recording
- Evidence that may be spread across multiple sentences
- A fixed relation without reading text
- How does open relation extraction differ from a closed relation schema?
- It cannot use text as input
- It always requires a knowledge base to label every sentence
- It only identifies entity boundaries
- It can express relation phrases without choosing only from a predefined label inventory