Search

In the literature, most previous studies on English implicit inter-sentence relation recognition only focused on semantic interactions, which could not exploit the syntactic interactive information in Chinese due to its complicated syntactic structure characteristics. In this paper, we propose a novel and effective model DSGCN-RoBERTa to learn the interaction features implied in sentences from both syntactic and semantic perspectives. To generate a rich contextual sentence embedding, we exploit RoBERTa, a large-scale pre-trained language model based on the transformer unit. DSGCN-RoBERTa consists of two key modules, the syntactic interaction and the semantic interaction modules. Specifically, the syntactic interaction module helps capture the depth-level structure information, including non-consecutive words and their relations, while the semantic interaction module enables the model to understand the context from the whole sentence to the local words. Furthermore, on top of such multi-perspective feature representations, we design a strength-dependent matching strategy that is able to adaptively capture the strong relevant interactive information in a fine-grained level. Extensive experiments demonstrate that the proposed method achieved state-of-the-art results on benchmarks Chinese compound sentence corpus CCCS and Chinese discourse corpus CDTB datasets. We also achieve comparable performance on the English corpus PDTB that demonstrates the superiority of our method. [ABSTRACT FROM AUTHOR]

We argue the agent's low generalization problem for searching target object in challenging visual navigation could be solved by "how" and "where" allowing the agent utilizing the scene priors. Although, recent works endow scene priors as fixed spatial features to provide good generalization in novel environment. However, these priors cannot adapt to new scenes. How to build scene priors and where to use the priors in visual navigation has not been well explored. We propose visual relationship detection module to adaptively build relational scene graph as priors. Besides, in order to use priors, we propose Graph attention Markov logical inference Network (GMN) module, which encodes the scene priors and performs precise action inference. GMN updates the graph structure in an unknown scene and estimates the shortest path in scene graph, whose emission probabilities from path to actions are pointwised by action samples in reinforcement learning to get optimal navigation policy. The whole navigation framework is driven by unsupervised reinforcement learning (RL) to exploit the environment. We conduct experiments on the AI2THOR virtual environment, and the results outperform the current most state-of-the-art both in SPL (Success weighted by Path Length) and success rate. [ABSTRACT FROM AUTHOR]