We present a Few-Shot Relation Classification Dataset (FewRel), consisting of 70, 000 sentences on 100 relations derived from Wikipedia and annotated by crowdworkers. The relation of each sentence is first recognized by distant supervision methods, and then filtered by crowdworkers. We adapt the most recent state-of-the-art few-shot learning methods for relation classification and conduct thorough evaluation of these methods. Empirical results show that even the most competitive few-shot learning models struggle on this task, especially as compared with humans. We also show that a range of different reasoning skills are needed to solve our task. These results indicate that few-shot relation classification remains an open problem and still requires further research. Our detailed analysis points multiple directions for future research. All details and resources about the dataset and baselines are released on

Distantly supervised relation extraction employs existing knowledge graphs to automatically collect training data. While distant supervision is effective to scale relation extraction up to large-scale corpora, it inevitably suffers from the wrong labeling problem. Many efforts have been devoted to identifying valid instances from noisy data. However, most existing methods handle each relation in isolation, regardless of rich semantic correlations located in relation hierarchies. In this paper, we aim to incorporate the hierarchical information of relations for distantly supervised relation extraction and propose a novel hierarchical attention scheme. The multiple layers of our hierarchical attention scheme provide coarseto-fine granularity to better identify valid instances, which is especially effective for extracting those long-tail relations. The experimental results on a large-scale benchmark dataset demonstrate that our models are capable of modeling the hierarchical information of relations and significantly outperform other baselines. The source code of this paper can be obtained from https://github.com/ thunlp/HNRE.

No abstract

The crosslink density plays a key role in the mechanical response of the amorphous polymers in previous experiments. However, the mechanism of the influence is still not clear. In this paper, the influence of crosslink density on the failure behavior under tension and shear in amorphous polymers is systematically studied using molecular dynamics simulations. The present results indicate that the ultimate stresses and the broken ratios (the broken bond number to all polymer chain number ratios) increase, as well as the ultimate strains decrease with increasing crosslink density. The strain concentration is clearer with the increase of crosslink density. In other words, a higher crosslink density leads to a higher strain concentration. Hence, the higher strain concentration further reduces the fracture strain. This study implies that the mechanical properties of amorphous polymers can be dominated for different applications by altering the molecular architecture.

Previous researches on the impacted composite laminates were mainly carried out according to the macromechanics-based homogenous strength theories, which ignore the local stress nonuniformity and strength difference between the fiber and matrix. In this paper, a new multiscale analysis method which combines the micromechanics of failure (MMF) theory for intralaminar damage and cohesive model for interlaminar failure is proposed. This approach is able to identify the failure modes of fiber and matrix in microscale as well as delamination between laminas. The finite element model of the multidirectional carbon fiber reinforced plastic (CFRP) laminate subjected to low-velocity impact is built on ABAQUS/Explicit platform. User material subroutine VUMAT is deveploed to analyze the micro stresses and determine the possible failure modes of fiber and matrix. Cohesive elements with bi-linear traction-separation law are employed to capture the onset and propagation of delamination. Finally, the structure response, fiber and matrix failure mode and delamination area are compared with experimental data under different impact energies, and the good agreements validate the effectiveness and accuracy of the novel method.