Dynamic Boundary Time Warping for sub-sequence matching with few examples

作者：

Highlights：

• Varied Information Retrieval problems can be treated as sub-sequence matching.

• A new method for finding a fragment of sequence similar to a few examples is proposed.

• It can retrieve sequential information from an untrimmed stream, e.g., text and video.

• Has good computational complexity and establishes a new direction for few-shot retrieval problems.

摘要

•Varied Information Retrieval problems can be treated as sub-sequence matching.•A new method for finding a fragment of sequence similar to a few examples is proposed.•It can retrieve sequential information from an untrimmed stream, e.g., text and video.•Has good computational complexity and establishes a new direction for few-shot retrieval problems.

论文关键词：Dynamic Time Warping,Sub-sequence matching,Natural Language Processing,Information Retrieval,Few-shot learning,Semantic retrieval

论文评审过程：Received 30 July 2020, Revised 25 October 2020, Accepted 16 November 2020, Available online 21 November 2020, Version of Record 10 February 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2020.114344