Optimal Transport-based Alignment of Learned Character Representations for String Similarity

TL;DR

This paper introduces STANCE, a novel learned string similarity model that uses optimal transport and neural networks to improve alias detection and downstream coreference tasks.

Contribution

The paper presents STANCE, a new method combining optimal transport and neural encoding for string similarity, outperforming existing models on multiple datasets.

Findings

STANCE outperforms state-of-the-art models on four datasets

Constructed five new alias detection datasets for evaluation

Improves cross-document coreference by 2.8 B^3 F1 points

Abstract

String similarity models are vital for record linkage, entity resolution, and search. In this work, we present STANCE --a learned model for computing the similarity of two strings. Our approach encodes the characters of each string, aligns the encodings using Sinkhorn Iteration (alignment is posed as an instance of optimal transport) and scores the alignment with a convolutional neural network. We evaluate STANCE's ability to detect whether two strings can refer to the same entity--a task we term alias detection. We construct five new alias detection datasets (and make them publicly available). We show that STANCE or one of its variants outperforms both state-of-the-art and classic, parameter-free similarity models on four of the five datasets. We also demonstrate STANCE's ability to improve downstream tasks by applying it to an instance of cross-document coreference and show that it leads to a 2.8 point improvement in B^3 F1 over the previous state-of-the-art approach.