Position-Aware Self-Attention based Neural Sequence Labeling

TL;DR

This paper introduces a position-aware self-attention model for sequence labeling tasks that effectively captures both successive and discrete token dependencies, outperforming existing models without external knowledge.

Contribution

The paper proposes a novel position-aware self-attention mechanism and context fusion layer to improve sequence labeling by modeling complex token relations.

Findings

Outperforms state-of-the-art models on POS tagging, NER, and phrase chunking

Effectively captures both continuous and discrete token dependencies

No external knowledge required for high performance

Abstract

Sequence labeling is a fundamental task in natural language processing and has been widely studied. Recently, RNN-based sequence labeling models have increasingly gained attentions. Despite superior performance achieved by learning the long short-term (i.e., successive) dependencies, the way of sequentially processing inputs might limit the ability to capture the non-continuous relations over tokens within a sentence. To tackle the problem, we focus on how to effectively model successive and discrete dependencies of each token for enhancing the sequence labeling performance. Specifically, we propose an innovative attention-based model (called position-aware selfattention, i.e., PSA) as well as a well-designed self-attentional context fusion layer within a neural network architecture, to explore the positional information of an input sequence for capturing the latent relations among tokens. Extensive experiments on three classical tasks in sequence labeling domain, i.e., partof-speech (POS) tagging, named entity recognition (NER) and phrase chunking, demonstrate our proposed model outperforms the state-of-the-arts without any external knowledge, in terms of various metrics.