Document Embedding Enhanced Event Detection with Hierarchical and Supervised Attention解讀
創新點 本文引入了文檔特徵來豐富token的信息，單純依靠本句子的信息有時候未必足夠，可以藉助其他句子的信息來進行補充，主要就是結合文檔級別的信息，加入到原本的ED任務中。具體建模方法如圖1，紅框就是補充的文檔特徵信息，最後與其他信息一起concat到LSTM模型中。但從實驗結果來看（如圖2），在ACE 2005上是所有方法中效果最好的。 模型 1.The EDODEL Module 通過設計良好
相關文章
- 1. 【論文讀後感】《Self-Attention Enhanced CNNs and Collaborative Curriculum Learning for Distantly Supervised》
- 2. Hierarchical Attention Networks for Document Classification
- 3. Hierarchical Attention Network for Document Classification閱讀筆記
- 4. Rumor Detection with Hierarchical Social Attention Network，2018，CIKM
- 5. HAN(Hierarchical Attention Networks for Document Classification)
- 6. attention model -- Hierarchical Attention Networks for Document Classification
- 7. 【EMNLP2019】Event Representation Learning Enhanced with External Commonsense Knowledge
- 8. Event Detection and Domain Adaptation with Convolutional Neural Networks
- 9. Hierarchical Attention Based Semi-supervised Network Representation Learning
- 10. Exploiting Argument Information to Improve Event Detection via Supervised Attention Mechanisms
- 更多相關文章...
- • C# 事件（Event） - C#教程
- • XSLT document() 函數 - XSLT 教程
- • JDK13 GA發佈：5大特性解讀
- • RxJava操作符（七）Conditional and Boolean
相關標籤/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
歡迎關注本站公眾號,獲取更多信息
相關文章
- 1. 【論文讀後感】《Self-Attention Enhanced CNNs and Collaborative Curriculum Learning for Distantly Supervised》
- 2. Hierarchical Attention Networks for Document Classification
- 3. Hierarchical Attention Network for Document Classification閱讀筆記
- 4. Rumor Detection with Hierarchical Social Attention Network，2018，CIKM
- 5. HAN(Hierarchical Attention Networks for Document Classification)
- 6. attention model -- Hierarchical Attention Networks for Document Classification
- 7. 【EMNLP2019】Event Representation Learning Enhanced with External Commonsense Knowledge
- 8. Event Detection and Domain Adaptation with Convolutional Neural Networks
- 9. Hierarchical Attention Based Semi-supervised Network Representation Learning
- 10. Exploiting Argument Information to Improve Event Detection via Supervised Attention Mechanisms