Search for: All records

Translation to or from low-resource languages (LRLs) poses challenges for machine translation in terms of both adequacy and fluency. Data augmentation utilizing large amounts of monolingual data is regarded as an effective way to alleviate these problems. In this paper, we propose a general framework for data augmentation in low-resource machine translation that not only uses target-side monolingual data, but also pivots through a related highresource language (HRL). Specifically, we experiment with a two-step pivoting method to convert high-resource data to the LRL, making use of available resources to better approximate the true data distribution of the LRL. First, we inject LRL words into HRL sentences through an induced bilingual dictionary. Second, we further edit these modified sentences using a modified unsupervised machine translation framework. Extensive experiments on four low-resource datasets show that under extreme low-resource settings, our data augmentation techniques improve translation quality by up to 1.5 to 8 BLEU points compared to supervised back-translation baselines. 

There is significant interest in the utility of asymmetric nanoaperture arrays as substrates for the surface-enhanced detection, fluorescence, and imaging of individual molecules. This work introduces obliquely-cut, out-of-plane, coaxial layered structures on an aperture edge. We refer to these structures as nanofingernails, which emphasizes their curved, oblique, and out-of-plane features. Broadband coupling into chiral hybrid plasmon modes and helicity-dependent near-field scattering without circular dichroism are demonstrated. The unusually-broadband, multipolar modes of nanofingernail micropore structures exhibit phase retardation effects that may be useful for achieving spatial overlap at different frequencies. The nanofingernail geometry shows new potential for simultaneous polarization-enhanced hyperspectral imaging on apertured, plasmonic surfaces. 

Limited approaches exist for imaging and recording spectra of individual nanostructures in the midinfrared region. Here we use infrared photothermal heterodyne imaging (IR-PHI) to interrogate single, high aspect ratio Au nanowires (NWs). Spectra recorded between 2,800 and 4,000 cm⁻¹for 2.5–3.9-μm-long NWs reveal a series of resonances due to the Fabry–Pérot modes of the NWs. Crucially, IR-PHI images show structure that reflects the spatial distribution of the NW absorption, and allow the resonances to be assigned to them= 3 andm= 4 Fabry–Pérot modes. This far-field optical measurement has been used to image the mode structure of plasmon resonances in metal nanostructures, and is made possible by the superresolution capabilities of IR-PHI. The linewidths in the NW spectra range from 35 to 75 meV and, in several cases, are significantly below the limiting values predicted by the bulk Au Drude damping parameter. These linewidths imply long dephasing times, and are attributed to reduction in both radiation damping and resistive heating effects in the NWs. Compared to previous imaging studies of NW Fabry–Pérot modes using electron microscopy or near-field optical scanning techniques, IR-PHI experiments are performed under ambient conditions, enabling detailed studies of how the environment affects mid-IR plasmons.