Efficient segmentation-free keyword spotting in historical document collections

作者：

Highlights：

• We present a query-by-example keyword spotting method for historical collections.

• The method is segmentation-free and avoids any pre-processing step.

• We use a compact and efficient vectorial representation to index large collections.

• We outperform the recent state-of-the-art keyword spotting approaches.

摘要

Highlights•We present a query-by-example keyword spotting method for historical collections.•The method is segmentation-free and avoids any pre-processing step.•We use a compact and efficient vectorial representation to index large collections.•We outperform the recent state-of-the-art keyword spotting approaches.

论文关键词：Historical documents,Keyword spotting,Segmentation-free,Dense SIFT features,Latent semantic analysis,Product quantization

论文评审过程：Received 23 April 2014, Revised 17 June 2014, Accepted 20 August 2014, Available online 28 August 2014.

论文官网地址：https://doi.org/10.1016/j.patcog.2014.08.021