Shallow and deep learning for event relatedness classification
作者：
Highlights：
• ML techniques applied to determine event pair relatedness (event linking) based on event templates automatically extracted from online news.
• Research driven by a real-world need to develop functionalities to reduce otherwise intractable event search space for intelligence gathering.
• Performance of shallow learning methods compared to a deep learning approach based on long short-term memory (LSTM) recurrent neural network.
• Focus on using linguistically lightweight features which are easily portable across languages.
• Practical application of machine learning techniques falling into the subfields of NLP, information engineering, event extraction and linking.
摘要
•ML techniques applied to determine event pair relatedness (event linking) based on event templates automatically extracted from online news.•Research driven by a real-world need to develop functionalities to reduce otherwise intractable event search space for intelligence gathering.•Performance of shallow learning methods compared to a deep learning approach based on long short-term memory (LSTM) recurrent neural network.•Focus on using linguistically lightweight features which are easily portable across languages.•Practical application of machine learning techniques falling into the subfields of NLP, information engineering, event extraction and linking.
论文关键词：Event relatedness,Event linking,Natural language processing,Shallow and deep machine learning,Feature engineering
论文评审过程：Received 6 November 2019, Revised 2 May 2020, Accepted 27 July 2020, Available online 19 August 2020, Version of Record 20 October 2020.
论文官网地址：https://doi.org/10.1016/j.ipm.2020.102371
