Word Segmentation as Graph Partition

TL;DR

This paper introduces a novel unsupervised graph-based method for Chinese word segmentation, modeling sentences as character graphs and applying spectral partitioning to improve segmentation accuracy on specialized and benchmark datasets.

Contribution

It presents a new graph partitioning approach for Chinese word segmentation, utilizing spectral algorithms and unsupervised techniques, with demonstrated effectiveness on diverse corpora.

Findings

Effective segmentation on Chinese health records

Competitive results on standard benchmarks

Unsupervised approach reduces reliance on labeled data

Abstract

We propose a new approach to the Chinese word segmentation problem that considers the sentence as an undirected graph, whose nodes are the characters. One can use various techniques to compute the edge weights that measure the connection strength between characters. Spectral graph partition algorithms are used to group the characters and achieve word segmentation. We follow the graph partition approach and design several unsupervised algorithms, and we show their inspiring segmentation results on two corpora: (1) electronic health records in Chinese, and (2) benchmark data from the Second International Chinese Word Segmentation Bakeoff.