Large-scale pre-trained language models have attracted extensive attentions in the research community and shown promising results on various tasks of natural language processing. However, the attention maps, which record the attention scores between tokens in self-attention mechanism, are sometimes ineffective as they are learned implicitly without the guidance of explicit semantic knowledge. Thus, we aim to infuse explicit external knowledge into pre-trained language models to further boost their performance. Existing works of knowledge infusion largely depend on multi-task learning frameworks, which are inefficient and require large-scale re-training when new knowledge is considered. In this paper, we propose a novel and generic solution, KAM-BERT, which directly incorporates knowledge-generated attention maps into the self-attention mechanism. It requires only a few extra parameters and supports efficient fine-tuning once new knowledge is added. KAM-BERT achieves consistent improvements on various academic datasets for natural language understanding. It also outperforms other state-of-the-art methods which conduct knowledge infusion into transformer-based architectures. Moreover, we apply our model to an industry-scale ad relevance application and show its advantages in the real-world scenario.

Dialogue safety problems severely limit the real-world deployment of neural conversational models and have attracted great research interests recently. However, dialogue safety problems remain under-defined and the corresponding dataset is scarce. We propose a taxonomy for dialogue safety specifically designed to capture unsafe behaviors in human-bot dialogue settings, with focuses on context-sensitive unsafety, which is under-explored in prior works. To spur research in this direction, we compile DiaSafety, a dataset with rich context-sensitive unsafe examples. Experiments show that existing safety guarding tools fail severely on our dataset. As a remedy, we train a dialogue safety classifier to provide a strong baseline for context-sensitive dialogue unsafety detection. With our classifier, we perform safety evaluations on popular conversational models and show that existing dialogue systems still exhibit concerning context-sensitive safety problems.

Discriminative pre-trained language models, such as ELECTRA, have achieved promising performances in a variety of general tasks. However, these generic pre-trained models struggle to capture domain-specific knowledge of domain-related tasks. In this work, we propose a novel domain-adaptation method for ELECTRA, which can dynamically select domain-specific tokens and guide the discriminator to emphasize them, without introducing new training parameters. We show that by re-weighting the losses of domain-specific tokens, ELECTRA can be effectively adapted to different domains. The experimental results in both computer science and biomedical domains show that the proposed method can achieve state-of-the-art results on the domain-related tasks.

Large pretrained language models can easily produce toxic or biased content, which is prohibitive for practical use. In order to detect such toxic generations, existing methods rely on templates, real-world data extraction, crowdsourcing workers or automatic generation to construct adversarial contexts that are likely to induce toxic generations. However, what type of context is more likely to induce unsafe responses is still under-explored. In this paper, we identify that context toxicity and context category (e.g., profanity, insult, drugs, etc.) are two important factors to cause safety issues in response generation. Hence, we propose a method called reverse generation to construct adversarial contexts conditioned on a given response, with the flexibility to control category, toxicity level and inductivity of the generated contexts. Via reverse generation, we augment the existing BAD dataset and construct a new dataset BAD+ which contains more than 120K diverse and highly inductive contexts in 12 categories. We test three popular pretrained dialogue models (Blender, DialoGPT and Plato2) and find that BAD+ can largely expose their safety problems. Furthermore, we show that BAD+ can greatly enhance the safety of generation, and we reveal the key factors of safety improvement. Our code and dataset is available at https://github.com/thu-coai/Reverse_Generation.

Offensive language detection is increasingly crucial for maintaining a civilized social media platform and deploying pre-trained language models. However, this task in Chinese is still under exploration due to the scarcity of reliable datasets. To this end, we propose a benchmark –COLD for Chinese offensive language analysis, including a Chinese Offensive Language Dataset –COLDATASET and a baseline detector –COLDETECTOR which is trained on the dataset. We show that the COLD benchmark contributes to Chinese offensive language detection which is challenging for existing resources. We then deploy the COLDETECTOR and conduct detailed analyses on popular Chinese pre-trained language models. We first analyze the offensiveness of existing generative models and show that these models inevitably expose varying degrees of offensive issues. Furthermore, we investigate the factors that influence the offensive generations, and we find that anti-bias contents and keywords referring to certain groups or revealing negative attitudes trigger offensive outputs easier.