Word n-gram attention models for sentence similarity and inference
作者：
Highlights：
• Bag-of-Words models boost their performance when making use of context.
• Context in the form of arbitrary n-grams outperforms informed baselines.
• Trainable attention over arbitrary n-grams further improves results.
• Improvements of up to 41% error reduction in language inference.
• Improvements of up to 38 % error reduction in textual similarity.
摘要
•Bag-of-Words models boost their performance when making use of context.•Context in the form of arbitrary n-grams outperforms informed baselines.•Trainable attention over arbitrary n-grams further improves results.•Improvements of up to 41% error reduction in language inference.•Improvements of up to 38 % error reduction in textual similarity.
论文关键词：Attention models,Deep learning,Natural language understanding,Natural Language Inference,Semantic textual similarity
论文评审过程：Available online 22 April 2019, Version of Record 4 May 2019.
论文官网地址：https://doi.org/10.1016/j.eswa.2019.04.054
