KgPLM: Knowledge-guided Language Model Pre-training via Generative and Discriminative Learning
This work addresses the problem of enhancing factual knowledge and performance in pre-trained language models for researchers and practitioners in NLP, offering strong specific gains on established benchmarks.
This paper introduces KgPLM, a language model pre-training framework that integrates factual knowledge completion and verification using both generative and discriminative learning. The model demonstrates richer factual knowledge on LAMA and achieves state-of-the-art performance on MRQA, with significant F1 improvements of +1.26 on NewsQA and +1.56 on TriviaQA over RoBERTa.
Recent studies on pre-trained language models have demonstrated their ability to capture factual knowledge and applications in knowledge-aware downstream tasks. In this work, we present a language model pre-training framework guided by factual knowledge completion and verification, and use the generative and discriminative approaches cooperatively to learn the model. Particularly, we investigate two learning schemes, named two-tower scheme and pipeline scheme, in training the generator and discriminator with shared parameter. Experimental results on LAMA, a set of zero-shot cloze-style question answering tasks, show that our model contains richer factual knowledge than the conventional pre-trained language models. Furthermore, when fine-tuned and evaluated on the MRQA shared tasks which consists of several machine reading comprehension datasets, our model achieves the state-of-the-art performance, and gains large improvements on NewsQA (+1.26 F1) and TriviaQA (+1.56 F1) over RoBERTa.