Fine-grained Knowledge Fusion for Sequence Labeling Domain Adaptation

TL;DR

This paper introduces a fine-grained knowledge fusion approach for sequence labeling domain adaptation that considers sample-level and element-level domain relevance, improving adaptation performance.

Contribution

It proposes a novel multi-level domain relevance modeling scheme to better balance source and target domain knowledge during adaptation.

Findings

Outperforms strong baselines in three sequence labeling tasks.

Effectively models multi-level domain relevance discrepancies.

Achieves state-of-the-art results in sequence labeling domain adaptation.

Abstract

In sequence labeling, previous domain adaptation methods focus on the adaptation from the source domain to the entire target domain without considering the diversity of individual target domain samples, which may lead to negative transfer results for certain samples. Besides, an important characteristic of sequence labeling tasks is that different elements within a given sample may also have diverse domain relevance, which requires further consideration. To take the multi-level domain relevance discrepancy into account, in this paper, we propose a fine-grained knowledge fusion model with the domain relevance modeling scheme to control the balance between learning from the target domain data and learning from the source domain model. Experiments on three sequence labeling tasks show that our fine-grained knowledge fusion model outperforms strong baselines and other state-of-the-art sequence labeling domain adaptation methods.