Efficient query-by-example spoken document retrieval combining phone multigram representation and dynamic time warping
作者：
Highlights：
• Query-by-example spoken document retrieval (QbESDR) strategies are time-consuming.
• A fast QbESDR strategy using different-sized n-grams and inverted indices is proposed.
• This was used to select candidates for a dynamic time warping (DTW)-based system.
• Score fusion of DTW and the proposed approach is also assessed.
• The paper reports effective and efficient solutions for the QbESDR task.
摘要
•Query-by-example spoken document retrieval (QbESDR) strategies are time-consuming.•A fast QbESDR strategy using different-sized n-grams and inverted indices is proposed.•This was used to select candidates for a dynamic time warping (DTW)-based system.•Score fusion of DTW and the proposed approach is also assessed.•The paper reports effective and efficient solutions for the QbESDR task.
论文关键词：Query-by-example spoken document retrieval,Phone decoding,Phone n-grams,Phone posteriorgrams,Dynamic time warping
论文评审过程：Received 20 February 2018, Revised 31 July 2018, Accepted 7 September 2018, Available online 22 September 2018, Version of Record 22 September 2018.
论文官网地址：https://doi.org/10.1016/j.ipm.2018.09.002
