Extracting relational triples from unstructured text is crucial for large-scale knowledge graph construction. However, few existing works excel in solving the overlapping triple problem where multiple relational triples in the same sentence share the same entities. In this work, we introduce a fresh perspective to revisit the relational triple extraction task and propose a novel cascade binary tagging framework (CASREL) derived from a principled problem formulation. Instead of treating relations as discrete labels as in previous works, our new framework models relations as functions that map subjects to objects in a sentence, which naturally handles the overlapping problem. Experiments show that the CAS-REL framework already outperforms state-ofthe-art methods even when its encoder module uses a randomly initialized BERT encoder, showing the power of the new tagging framework. It enjoys further performance boost when employing a pre-trained BERT encoder, outperforming the strongest baseline by 17.5 and 30.2 absolute gain in F1-score on two public datasets NYT and WebNLG, respectively. In-depth analysis on different scenarios of overlapping triples shows that the method delivers consistent performance gain across all these scenarios. The source code and data are released online 1 .
Context. Several models compete to explain the abundance properties of stellar populations in globular clusters. One of the main constraints is the present-day ratio of first-and second-generation stars that are currently identified based on their sodium content. Aims. We propose an alternative interpretation of the observed sodium distribution, and suggest that stars with low sodium abundance that are counted as members of the first stellar generation could actually be second-generation stars. Methods. We compute the number ratio of second-generation stars along the Na distribution following the fast rotating massive star model using the same constraints from the well-documented case of NGC 6752 as in our previous developments. Results. We reproduce the typical percentage of low-sodium stars usually classified as first-generation stars by invoking only secondary star formation from material ejected by massive stars and mixed with original globular cluster material in proportions that account for the Li-Na anti-correlation in this cluster. Conclusions. Globular clusters could be totally devoid of first-generation low-mass stars today. This can be tested with the determination of the carbon isotopic ratio and nitrogen abundance in turn-off globular cluster stars. Consequences and related issues are briefly discussed.
High-quality nonmagnetic half-Heusler ScPtBi single crystals were synthesized by a Bi self-flux method. This compound was revealed to be a hole-dominated semimetal with a large low-field magnetoresistance up to 240% at 2K in a magnetic field of 1T. Magneto-transport measurements demonstrated that the large low-field magnetoresistance effect resulted from the coexistence of field-induced metal-semiconductor transition and weak-antilocalization effect. Moreover, Hall measurements indicated that ScPtBi single crystal showed a high mobility over a wide temperature region even up to room temperature (4050 cm 2 V -1 s -1 at 2K -2016 cm 2 V -1 s -1 at 300K).These findings not only suggest the nonmagnetic ScPtBi semimetal a potential material candidate for applications in high-sensitivity magnetic sensors, but also are of great significance to comprehensively understand the rare-earth based half-Heusler compounds.Recently, searching for new materials with semimetallic features in electronic structure protected by the interplay of symmetry and topology has attracted increasing attention, following the observation of large non-saturation magnetoresistance (MR) in a variety of nonmagnetic semimetals such as Cd 3 As 2 , 1 TaAs, 2 NbP, 3 and WTe 2 . 4 Electronic transport in heavy half-Heusler compounds is distinguished from others by the fact that the strong spin-orbit coupling (SOC) induces many fascinating phenomena, such as the heavy fermion behavior, 5 noncentersymetry superconductors, 6-8 possible topological insulators, 9-13 large positive MR, [14][15][16] and extremely high mobility, 17 which are of great significance to the applied physics field.According to band structure calculations, half-Heusler ScPtBi is predicted to show similar band inversion to that of HgTe. 9, 10, 13 Its topological properties can be created by applying strain or designing an appropriate quantum well structure. 9 Moreover, this material possesses nearly linear dispersion relation of bulk band along with zero gap, 13 which is usually associated with large MR effect and high mobility behavior. [1][2][3][4]14,15,17,18 Although the growth of many heavy half-Heusler compounds (such as LuPdBi, 6,16, 19 LuPtBi, 7 LuPtSb,15,20,21 and YPtBi 8 ) have been previously reported, the synthesis of ScPtBi (no matter polycrystal or single crystal) is unsuccessful thus far since this compound is proposed to show a weaker thermodynamic stability. 22,23 In this work, we report on the growth and magneto-transport properties of high-quality nonmagnetic ScPtBi single crystal. Intriguingly, the metal-semiconductor transition coexists with the weak antilocalization (WAL) effect in this material, which results in a large low-field positive MR (240% at 2K -150% at 100K) in a magnetic field of 1T.Single crystals of ScPtBi were synthesized by using excess Bi as the flux in a molar ratio of Sc:Pt: Bi = 1:1:10. The starting materials were mixed together and placed in an alumina crucible with the one having higher melting temperatures on the bottom. This process was handle...
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.