Natural language inference (NLI) is the basic task of many applications such as question answering and paraphrase recognition. Existing methods have solved the key issue of how the NLI model can benefit from external knowledge. Inspired by this, we attempt to further explore the following two problems: (1) how to make better use of external knowledge when the total amount of such knowledge is constant and (2) how to bring external knowledge to the NLI model more conveniently in the application scenario. In this paper, we propose a novel joint training framework that consists of a modified graph attention network, called the knowledge graph attention network, and an NLI model. We demonstrate that the proposed method outperforms the existing method which introduces external knowledge, and we improve the performance of multiple NLI models without additional external knowledge.

The ability to learn incrementally is critical to the long-term operation of AI systems. Benefiting from the power of few-shot class-incremental learning (FSCIL), deep learning models can continuously recognize new classes with only a few samples. The difficulty is that limited instances of new classes will lead to overfitting and exacerbate the catastrophic forgetting of the old classes. Most previous works alleviate the above problems by imposing strong constraints on the model structure or parameters, but ignoring embedding network transferability and classifier adaptation (CA), failing to guarantee the efficient utilization of visual features and establishing relationships between old and new classes. In this paper, we propose a simple and novel approach from two perspectives: embedding bias and classifier bias. The method learns an embedding augmented (EA) network with cross-class transfer and class-specific discriminative abilities based on self-supervised learning and modulated attention to alleviate embedding bias. Based on the adaptive incremental classifier learning scheme to realize incremental learning capability, guiding the adaptive update of prototypes and feature embeddings to alleviate classifier bias. We conduct extensive experiments on two popular natural image datasets and two medical datasets. The experiments show that our method is significantly better than the baseline and achieves state-of-the-art results.

Entity-relation extraction is a basic task in natural language processing, and recently, the use of deep-learning methods, especially the Long Short-Term Memory (LSTM) network, has achieved remarkable performance. However, most of the existing entity-relation extraction methods cannot solve the overlapped multi-relation extraction problem, which means one or two entities are shared among multiple relational triples contained in a sentence. In this paper, we propose a text-generated method to solve the overlapped problem of entity-relation extraction. Based on this, (1) the entities and their corresponding relations are jointly generated as target texts without any additional feature engineering; (2) the model directly generates the relational triples using a unified decoding process, and entities can be repeatedly presented in multiple triples to solve the overlapped-relation problem. We conduct experiments on two public datasets—NYT10 and NYT11. The experimental results show that our proposed method outperforms the existing work, and achieves the best results.

Recent years have witnessed the growth of recommender systems, with the help of deep learning techniques. Recurrent Neural Networks (RNNs) play an increasingly vital role in various session-based recommender systems, since they use the user’s sequential history to build a comprehensive user profile, which helps improve the recommendation. However, a problem arises regarding how to be aware of the variation in the user’s contextual preference, especially the short-term intent in the near future, and make the best use of it to produce a precise recommendation at the start of a session. We propose a novel approach named Attention-based Short-term and Long-term Model (ASLM), to improve the next-item recommendation, by using an attention-based RNNs integrating both the user’s short-term intent and the long-term preference at the same time with a two-layer network. The experimental study on three real-world datasets and two sub-datasets demonstrates that, compared with other state-of-the-art methods, the proposed approach can significantly improve the next-item recommendation, especially at the start of sessions. As a result, our proposed approach is capable of coping with the cold-start problem at the beginning of each session.