Trading accuracy for faster entity linking

Named entity linking (NEL) can be applied to documents such as financial reports, web pages and news articles, but state of the art disambiguation techniques are currently too slow for web-scale applications because of a high complexity with respect to the number of candidates. In this paper, we accelerate NEL by taking two successful disambiguation features (popularity and context comparability) and use them to reduce the number of candidates before further disambiguation takes place. Popularity is measured by in-link score, and context similarity is measured by locality sensitive hashing. We present a novel approach to locality sensitive hashing which embeds the projection matrix into a smaller array and extracts columns of the projection matrix using feature hashing, resulting in a lowmemory approximation. We run the linker on a test set in 63% of the baseline time with an accuracy loss of 0.72%.