A Unified Knowledge Graph Augmentation Service for Boosting Domain-specific NLP Tasks

TL;DR

KnowledgeDA is a unified service that enhances domain-specific NLP tasks by automatically augmenting training data with domain knowledge graphs, improving model performance across healthcare and software development domains.

Contribution

It introduces a novel, unified framework for injecting domain knowledge into PLMs during fine-tuning using knowledge graphs and data augmentation techniques.

Findings

Improves domain-specific text classification accuracy.

Enhances QA task performance in healthcare and software domains.

Demonstrates generalizability across different NLP tasks.

Abstract

By focusing the pre-training process on domain-specific corpora, some domain-specific pre-trained language models (PLMs) have achieved state-of-the-art results. However, it is under-investigated to design a unified paradigm to inject domain knowledge in the PLM fine-tuning stage. We propose KnowledgeDA, a unified domain language model development service to enhance the task-specific training procedure with domain knowledge graphs. Given domain-specific task texts input, KnowledgeDA can automatically generate a domain-specific language model following three steps: (i) localize domain knowledge entities in texts via an embedding-similarity approach; (ii) generate augmented samples by retrieving replaceable domain entity pairs from two views of both knowledge graph and training data; (iii) select high-quality augmented samples for fine-tuning via confidence-based assessment. We implement a prototype of KnowledgeDA to learn language models for two domains, healthcare and software development. Experiments on domain-specific text classification and QA tasks verify the effectiveness and generalizability of KnowledgeDA.